United States v. Oracle Corp.
This text of 331 F. Supp. 2d 1098 (United States v. Oracle Corp.) is published on Counsel Stack Legal Research, covering District Court, N.D. California primary law. Counsel Stack provides free access to over 12 million legal documents including statutes, case law, regulations, and constitutions.
Opinion
FINDINGS OF FACT, CONCLUSIONS OF LAW AND ORDER THEREON
WALKER, Chief Judge.
The government, acting through the Department of Justice, Antitrust Division, and the states of Connecticut, Hawaii, Maryland, Massachusetts, Michigan, Minnesota, New York, North Dakota, Ohio and Texas, First Amended Complaint (FAC) (Doc. # 125) ¶ 3 at 5-6, seek to enjoin Oracle Corporation from acquiring, directly or indirectly, the whole or any part of the stock of PeopleSoft, Inc. Plaintiffs allege that the acquisition would violate section 7 of the Clayton Act, 15 USC § 18. Both companies are publicly traded and headquartered in this district. Jt. Stip Fact (Doc. # 218) at 1-2. The court has subject matter jurisdiction under 15 USC § 25 and 28 USC §§ 1331, 1337(a) and 1345. There is no dispute about the *1101 court’s personal jurisdiction over the defendant.
Oracle initiated its tender offer for the shares of PeopleSoft on June 6, 2003. Jt. Stip of Fact (Doc. # 128) at 2; Ex. P2040. Plaintiffs brought suit on February 26, 2004. Compl. (Doc. # 1). The case was tried to the court on June 7-10, 14-18, 21-25, 28-30 and July 1, 2004, with closing arguments on July 20, 2004, and further evidentiary proceedings on August 13, 2004. Based on the evidence presented and the applicable law, the court concludes that plaintiffs have failed to carry the burden of proof entitling them to relief and, therefore, orders that judgment be entered for defendant and against plaintiffs.
INTRODUCTORY FINDINGS: INDUSTRY OVERVIEW
Products at Issue
Of the many types of computer software, such as operating system software, database software, integration software (sometimes called “middleware” in software parlance) and utilities software, this case involves only one — application software. And within this type, the present case deals with only applications that automate the overall business data processing of business and similar entities; these applications are called “enterprise application software” (EAS). Jt. Definitions (Doc. # 332) at 6. There are three main kinds of EAS. Plaintiffs single out one.
Some EAS programs are mass market PC-based applications of fairly limited “functionality” (meaning capability). Id. (Doc. # 332) at 5. See Daniel E. O’Leary, Enterprise Resource Planning Systems at 19 (Cambridge, 2000). Other EAS programs are developed by or for a specific enterprise and its particular needs; most large organizations had such specially designed EAS (called “legacy software”) pri- or to the advent of the products in suit. Plaintiffs focus their claims on the third, intermediate category of EAS — enterprise resource planning (ERP) system software. Jt. Sub. Definitions (Doc. # 332) at 6. ERP is packaged software that integrates most of an entity’s data across all or most of the entity’s activities. See O’Leary, Enterprise Resource Planning Systems at 27-38. Oracle and PeopleSoft develop, produce, market and service ERP software.
These copyrighted software programs are licensed (“sold” is the term applied to these license transactions) to end users along with a continued right to use license which usually includes maintenance or upgrades of the software. To the customer, the fees to license and maintain ERP software are generally a small part, 10 to 15 percent, of the total cost of the installation and maintenance of an ERP system. Tr. at 133:12-15 (Hatfield); 655:2-4 (Maxwell); 1385:6-11 (Gorriz). An ERP installation, because of its complexity, usually requires substantial and expensive personnel training, consulting and other services to integrate the program into the customer’s preexisting or “legacy” software. Jt. Sub. Definitions (Doc. #332) at 6. See also O’Leary, Enterprise Resource Planning Systems at 19. ERP software vendors often provide some of those services, but they are typically also performed and augmented by the customer’s own staff, obtained from providers other than ERP vendors or both.
Many ERP programs were developed to address the needs of particular industries, such as banking and finance, insurance, engineering, construction, healthcare, government, legal and so forth (in industry lingo, these are called “verticals”). See Martin Campbell-Kelly, From Airline Reservations to Sonic the Hedgehog, at 169-73 (MIT, 2003). Vertical-specific ERP programs, although well suited to the needs of firms engaged in a particular industry, often are not well suited to the *1102 needs of firms in other verticals. An enterprise that relies on vertical-specific ERP software products, but whose operations embrace more than one vertical faces the task of integrating the programs. The largest and most complex organizations face particular difficulty. “[0]nly custom-written software or carefully tailored and integrated cross-industry packages [can] handle larger firms’ historically idiosyncratic accounting systems and diverse overseas operations.” Id. at 172.
ERP programs have been developed to handle the full range of an enterprise’s activities; these include human relations management (HRM), financial management systems (FMS), customer relations management (CRM), supply chain management (SCM), Product Life Cycle Management, Business Intelligence (BI), among many others. These are called “pillars.” Although ERP encompasses many pillars, see Ex. D5572, plaintiffs assert claims with respect to only two pillars, HRM and FMS. FAC (Doc. # 125) ¶ 23 at 12-13.
Within these two pillars, plaintiffs further limit their claims to only those HRM and FMS products able to meet the needs of large and complex enterprises with “high functional needs.” Id. at ¶ 14 at 9. Plaintiffs label HRM and FMS products capable of meeting these high function needs “high function HRM software” and “high function FMS software,” respectively. Id. ¶23&)-&) at 12-13. ERP pillars incapable of meeting these high function needs are called “mid-market” software by plaintiffs. Id. ¶ 13 at 9.
“High function software” is a term adopted by plaintiffs to describe what they contend is the separate and distinct line of commerce in which they- contend competition would be lessened by the proposed acquisition. Id. at ¶23 at 13-14. Plaintiffs apply the term “high function” to both HRM and FMS. “High function software,” as defined by plaintiffs, has no i*eeognized meaning in the industry. See Tr. at 349:7-10 (Bergquist); 2298:6-20 (Elzinga).
Rather, industry participants and software vendors use the terms “enterprise” software, “up-market” software and “Tier One” software to denote ERP that is capable of executing a wide array of business processes at a superior level of performance. See Tr. at 274:24-275:7 (Bergquist); Tr. at 1771:5-1772:1 (Wilmington); Tr. at 1554:25-1555:7 (Wolfe); Tr. at 2180:22-2181:5 (Iansiti). Software vendors use these terms to focus sales and marketing initiatives. Tr. 2816:6-2818:8 (Knowles) (testifying that SAP divided mid-market and large enterprise at $1.5 billion based on SAP’s sales resources and estimated amount of IT “spend” available from those customers).
Each ERP pillar consists of “modules” that automate particular processes or functions. HRM and FMS software each consists of numerous modules. Exs. P3010, P3011. Tr.
Free access — add to your briefcase to read the full text and ask questions with AI
FINDINGS OF FACT, CONCLUSIONS OF LAW AND ORDER THEREON
WALKER, Chief Judge.
The government, acting through the Department of Justice, Antitrust Division, and the states of Connecticut, Hawaii, Maryland, Massachusetts, Michigan, Minnesota, New York, North Dakota, Ohio and Texas, First Amended Complaint (FAC) (Doc. # 125) ¶ 3 at 5-6, seek to enjoin Oracle Corporation from acquiring, directly or indirectly, the whole or any part of the stock of PeopleSoft, Inc. Plaintiffs allege that the acquisition would violate section 7 of the Clayton Act, 15 USC § 18. Both companies are publicly traded and headquartered in this district. Jt. Stip Fact (Doc. # 218) at 1-2. The court has subject matter jurisdiction under 15 USC § 25 and 28 USC §§ 1331, 1337(a) and 1345. There is no dispute about the *1101 court’s personal jurisdiction over the defendant.
Oracle initiated its tender offer for the shares of PeopleSoft on June 6, 2003. Jt. Stip of Fact (Doc. # 128) at 2; Ex. P2040. Plaintiffs brought suit on February 26, 2004. Compl. (Doc. # 1). The case was tried to the court on June 7-10, 14-18, 21-25, 28-30 and July 1, 2004, with closing arguments on July 20, 2004, and further evidentiary proceedings on August 13, 2004. Based on the evidence presented and the applicable law, the court concludes that plaintiffs have failed to carry the burden of proof entitling them to relief and, therefore, orders that judgment be entered for defendant and against plaintiffs.
INTRODUCTORY FINDINGS: INDUSTRY OVERVIEW
Products at Issue
Of the many types of computer software, such as operating system software, database software, integration software (sometimes called “middleware” in software parlance) and utilities software, this case involves only one — application software. And within this type, the present case deals with only applications that automate the overall business data processing of business and similar entities; these applications are called “enterprise application software” (EAS). Jt. Definitions (Doc. # 332) at 6. There are three main kinds of EAS. Plaintiffs single out one.
Some EAS programs are mass market PC-based applications of fairly limited “functionality” (meaning capability). Id. (Doc. # 332) at 5. See Daniel E. O’Leary, Enterprise Resource Planning Systems at 19 (Cambridge, 2000). Other EAS programs are developed by or for a specific enterprise and its particular needs; most large organizations had such specially designed EAS (called “legacy software”) pri- or to the advent of the products in suit. Plaintiffs focus their claims on the third, intermediate category of EAS — enterprise resource planning (ERP) system software. Jt. Sub. Definitions (Doc. # 332) at 6. ERP is packaged software that integrates most of an entity’s data across all or most of the entity’s activities. See O’Leary, Enterprise Resource Planning Systems at 27-38. Oracle and PeopleSoft develop, produce, market and service ERP software.
These copyrighted software programs are licensed (“sold” is the term applied to these license transactions) to end users along with a continued right to use license which usually includes maintenance or upgrades of the software. To the customer, the fees to license and maintain ERP software are generally a small part, 10 to 15 percent, of the total cost of the installation and maintenance of an ERP system. Tr. at 133:12-15 (Hatfield); 655:2-4 (Maxwell); 1385:6-11 (Gorriz). An ERP installation, because of its complexity, usually requires substantial and expensive personnel training, consulting and other services to integrate the program into the customer’s preexisting or “legacy” software. Jt. Sub. Definitions (Doc. #332) at 6. See also O’Leary, Enterprise Resource Planning Systems at 19. ERP software vendors often provide some of those services, but they are typically also performed and augmented by the customer’s own staff, obtained from providers other than ERP vendors or both.
Many ERP programs were developed to address the needs of particular industries, such as banking and finance, insurance, engineering, construction, healthcare, government, legal and so forth (in industry lingo, these are called “verticals”). See Martin Campbell-Kelly, From Airline Reservations to Sonic the Hedgehog, at 169-73 (MIT, 2003). Vertical-specific ERP programs, although well suited to the needs of firms engaged in a particular industry, often are not well suited to the *1102 needs of firms in other verticals. An enterprise that relies on vertical-specific ERP software products, but whose operations embrace more than one vertical faces the task of integrating the programs. The largest and most complex organizations face particular difficulty. “[0]nly custom-written software or carefully tailored and integrated cross-industry packages [can] handle larger firms’ historically idiosyncratic accounting systems and diverse overseas operations.” Id. at 172.
ERP programs have been developed to handle the full range of an enterprise’s activities; these include human relations management (HRM), financial management systems (FMS), customer relations management (CRM), supply chain management (SCM), Product Life Cycle Management, Business Intelligence (BI), among many others. These are called “pillars.” Although ERP encompasses many pillars, see Ex. D5572, plaintiffs assert claims with respect to only two pillars, HRM and FMS. FAC (Doc. # 125) ¶ 23 at 12-13.
Within these two pillars, plaintiffs further limit their claims to only those HRM and FMS products able to meet the needs of large and complex enterprises with “high functional needs.” Id. at ¶ 14 at 9. Plaintiffs label HRM and FMS products capable of meeting these high function needs “high function HRM software” and “high function FMS software,” respectively. Id. ¶23&)-&) at 12-13. ERP pillars incapable of meeting these high function needs are called “mid-market” software by plaintiffs. Id. ¶ 13 at 9.
“High function software” is a term adopted by plaintiffs to describe what they contend is the separate and distinct line of commerce in which they- contend competition would be lessened by the proposed acquisition. Id. at ¶23 at 13-14. Plaintiffs apply the term “high function” to both HRM and FMS. “High function software,” as defined by plaintiffs, has no i*eeognized meaning in the industry. See Tr. at 349:7-10 (Bergquist); 2298:6-20 (Elzinga).
Rather, industry participants and software vendors use the terms “enterprise” software, “up-market” software and “Tier One” software to denote ERP that is capable of executing a wide array of business processes at a superior level of performance. See Tr. at 274:24-275:7 (Bergquist); Tr. at 1771:5-1772:1 (Wilmington); Tr. at 1554:25-1555:7 (Wolfe); Tr. at 2180:22-2181:5 (Iansiti). Software vendors use these terms to focus sales and marketing initiatives. Tr. 2816:6-2818:8 (Knowles) (testifying that SAP divided mid-market and large enterprise at $1.5 billion based on SAP’s sales resources and estimated amount of IT “spend” available from those customers).
Each ERP pillar consists of “modules” that automate particular processes or functions. HRM and FMS software each consists of numerous modules. Exs. P3010, P3011. Tr. at 268:8-269:11, 270:5-271:12 (Bergquist). HRM modules include such functions as payroll, benefits, sales incentives, time management and many others. Ex. P3010. FMS modules include such functions as general ledger, accounts receivable, accounts payable, asset management and many others. Ex. P3011.
“Core” HRM modules are those specific ERP modules that individually or collectively automate payroll, employee tracking and benefits administration. Core FMS modules are those ERP modules that individually or collectively track general ledger, accounts receivable, accounts payable and cash and asset management business processes. Core FMS and HRM modules are offered by all the ERP vendors that have HRM and FMS offerings. Ex. P3179 (Ciandrini 1/16/04 Dep) at Tr. 256:2-257:10. Large enterprise customers rarely, if ever, buy core HRM or FMS modules *1103 in isolation. Tr. at 3461:14-23 (Catz). Customarily, FMS and HRM software are purchased in bundles with other products. Tr. at 3807:21-3808:1 (Hausman). See also Tr. at 3813:12-13 (Hausman). Customers purchase a cluster of products such as Oracle’s E-Business Suite that provide the customer with a “stack” of software and technology, which may include core HRM or FMS applications, add-on modules, “customer-facing” business applications such as CRM software, and the infrastructure components (application servers and database) on which the applicátions run. Tr. at 3461:14-3462:5 (Catz); Tr. at 3807:21-3808:1 (Hausman). See, e.g., Exs. P1000-P1322 (Oracle discount request forms).
ERP vendors, including Oracle and Peo-pleSoft, sell modules individually as well as integrated suite products. Some ERP vendors sell only one or a few modules. Individual modules are referred to as “point solutions” as they address a particular need of the enterprise. ERP vendors that sell products for only one or a limited number pillars are referred to as point solution or “best of breed” providers. A customer licensing a particular module because it fits the specific needs of the enterprise is sometimes said to be seeking a best of breed or point solution. An ERP customer that acquires best of breed or point solutions faces the task of integrating these solutions with one another and with the customer’s existing ERP or legacy footprint.
Although the production cost of ERP applications is negligible, vendors bear significant development and marketing expenses and substantial costs of pre- and postsales support and ongoing maintenance and enhancement. ERP vendors employ and bear substantial costs of account managers, technical sales forces and personnel for user training, product documentation and post-sale support.
Customers at Issue
“Large Complex Enterprises” (LCE) is a term adopted by plaintiffs to describe the ERP customers. that have “high function software” needs. Based on the testimony described hereafter, the court finds that industry participants and software vendors do not typically use this term and it has no widely accepted meaning in the industry.
While many in the software industry differentiate between large customers and mid-market customers, there is no “bright line” test for what is a “large” or “upmarket” customer. Tr. 348:23-349:3 (Bergquist) (acknowledging “different parties tend to define it differently”); Tr. 2033:1-12 (Iansiti); Ex. P3032 (Henley 5/4/04 Dep) at Tr. 98:20-25. Likewise, there is no “bright line” test for what is a “mid-market” customer. TT. at 2820:9-19 (Knowles) (SAP executive noting that the separation between mid-market and large enterprise customers is “not an exact science”); Ex. D7174 (Pollie 5/26/04 Dep) at Tr. 54:14-55:3 (testifying that the meaning of the term mid-market “varies from, from everyone you talk to”); Ex. P3191 (Block 12/16/03 Dep) at Tr. 88:12-21, 94:19-95:3 (noting the term mid-market is used in many different ways by many • different people). ERP vendors, analysts, systems integrators and others in the industry define the mid-market variously. Compare Tr. at 864:19-865:2 (Keating) (noting variability of definitions and that Bearing Point generally refers to mid-market as customers in its General Business Group, which-is synonymous with companies having less than $2 billion in revenue) with Tr. at 1846:17-1847:15 (Wilmington) (PeopleSoft formerly defined mid-market as less than $500 million revenue, but after- acquiring J D Edwards, it raised mid-market to include companies with less than $1 billion revenue).
*1104 Prior to Oracle’s tender offer, People-Soft used a proxy of $500 million in revenue to distinguish mid-market customers from large customers. Tr. at 348:5-18 (Bergquist). SAP defines its “large enterprise” market as companies with more than $1.5 billion in revenues. Tr. at 2819:12-20 (Knowles). Oracle segments the market based on the customers’ revenue level or number of employees. Ex. P3070 (Prestipino 5/18/04 Dep) at Tr. at 21:5-23:11.
Plaintiffs failed to show ERP vendors distinguish mid-market customers from large customers on the amount of money spent in an ERP purchase. Yet, as discussed below, this was the basis on which plaintiffs attempted to quantify the ERP market.
Vendors at Issue
Many firms develop, produce, market and maintain ERP software. Ex. 5543 at 8-17. Some ERP software vendors, notably Oracle, PeopleSoft and a German company, SAP AG, developed cross-industry applications or “suites” of “generalized integrated software that could be customized for virtually any large business,” Campbell-Kelly, From Airline Reservations to Sonic the Hedgehog at 172. It is to the products of these three vendors that plaintiffs direct their allegations. Although not alone in the ERP business, these three firms have the most comprehensive ERP software offerings.
Oracle. Oracle is headquartered in Redwood Shores, California. Oracle has over 41,000 employees and offices in 80 countries and sells product in over 120 countries. Tr. at 3485:10-12, 3486:16-18 (Catz). Oracle’s E-Business suite is a fully integrated suite of more than 70 modules for FMS, internet procurement, BI, SCM, manufacturing, project systems, HRM and sales and service management. Ex. P2209 at xiv. As of December 2002, Oracle had over 5000 customers of its E-Business Suite, Release Hi. Ex. P2208 at ORLIT-EDOC-00244117; Ex. P3038. Oracle’s ERP products have enjoyed success with telecommunications and financial services customers. Oracle is a major producer of relational database software which accounts for a much larger share of its revenue than its ERP business.
PeopleSoft. PeopleSoft is headquartered in Pleasanton, California and has 8300 employees. PeopleSoft sells software “in most major markets.” Ex. 7149 at 7. It has offices in Europe, Japan, Asia-Pacific, Latin America and other parts of the world. Id. PeopleSoft was formed in 1987 to develop an HRM product, and it continues to enjoy widespread customer acceptance of its HRM offerings. PeopleSoft now sells, in addition to HRM products, FMS, SCM and CRM products and related consulting services. Jt. Stip Fact (Doc. # 218) at 2. In 2003, Peoplesoft generated about $1.7 billion in revenue, derived almost entirely from ERP-related business. PeopleSoft v8 is PeopleSoft’s current integrated suite offering. It competes with Oracle’s E-Business suite, Release lli.
SAP. SAP AG is headquartered in Waldorf, Germany. SAP AG has global operations, including major business operations in more than a dozen countries and customers in more than 120 countries around the globe. Tr. at 2805:20-2806:2 (Knowles). SAP AG has over 30,000 employees and sells a product called MySAP ERP Suite, which includes HRM, FMS, corporate controlling and corporate services. Tr. at 2811:7-13 (Knowles). SAP AG offers a product called All-in-One, which is “essentially a scaled-down version of MySAP ERP with a lot of functionality turned off.” Tr. at 2813:20-2814:2 (Knowles). All-in-One is marketed both through an indirect channel of resellers to the $200 milhon-and-below customer revenue segment and by SAP’s direct sales *1105 force. Tr. at 2813:20-2814:2 (Knowles). SAP AG also offers a product called Business One, which is a “packaged software offering” targeting the $200 million-and-below customer revenue segment and sold through an indirect channel of resellers. Tr. at 2813:10-17 (Knowles). SAP has six sales regions worldwide. SAP America, Inc is responsible for sales in the United States and Canada. Tr. at 2808:16-19 (Knowles). SAP America sells software solutions created by SAP AG. Tr. at 2808:8-15, 2806:16-17 (Knowles). In addition to selling software solutions created by SAP AG, the largest price discounts offered by SAP America must be approved by SAP AG. Tr. 2836:22-24 (Knowles). SAP products have won wide acceptance in the aerospace and petroleum industries. Tr. at 899:9-900:19, 947:10-21 (Keating).
Lawson. Lawson is headquartered in Saint Paul, Minnesota and has 1700 employees. Lawson was founded in the mid-1970s and has 2000 customers, mostly in North America and Europe. Lawson offers FMS, HRM, procurement products, merchandising products, enterprise performance management (EPM), service automation and a unique function called surgery instrument management. Tr. at 3591:5-10 (Coughlan). In 2003, Lawson generated more than $360 million in annual revenue. Tr. at 3589:19 (Coughlan). Lawson has tended to do extremely well in the healthcare and retail verticals. Tr. at 3591:1-2 (Coughlan). As Professor Jerry Hausman testified, and the court will hereafter find, although Lawson does not now compete in all the industry verticals in which Oracle, PeopleSoft and SAP compete, Lawson has sufficient resources and capabilities to reposition to any industry vertical it so chooses. Tr. at 3841:3-13 (Hausman).
AM'S. AMS is an ERP vendor that was recently acquired by CGI, headquartered in Montreal, Quebec, with.offices in North America, Europe and Asia-Pacific. As an ERP vendor, AMS offers FMS, HRM, procurement, tax and revenue software, CRM, CMS, environmental compliance software, performance management and budgeting and contracting software to government entities. See P3034 (Morea 5/7/04 Dep) at Tr. 14:19-23. AMS has been successful in its sales to state and federal governmental agencies, often competing head to head with commercial ERP vendors. Tr. at 972:6-15 (Keating) (agreeing that AMS is a “viable'competitor-for large and complex federal procurements”). In fact, only a short time after this action was initiated, the United States Department of Justice chose AMS FMS over the FMS offerings of Oracle, PeopleSoft and SAP. See D7166 (Morea 5/7/04 Dep) at Tr. 21:22-22:7.
Microsoft. Microsoft is headquartered in Redmond, Washington, and sells a wide range of software products. In 2001 Microsoft acquired Great Plains Software and renamed it Microsoft Great Plains. Microsoft now has a division called Microsoft Business Solutions (MBS), which was created in 2002 when Microsoft Great Plains acquired the Danish software company Navison. Tr. at 2972:19-2973:9, 2973:8-9 (Burgum). MBS has four existing ERP product lines: Navison, Great Plains, Axapta and Solomon. Tr. at 2996:16 (Burgum). Great Plains offers FMS, HRM, E-commerce, retail management, CRM, analytics and reporting. See http://www. micro soft. com/BusinessSo-luions/GreatPlains/default, aspx. Solomon provides FMS only. Tr. at 2998:4 (Bur-gum). Navison offers FMS, SCM, CRM and E-commerce. See http://mvw.micro-soft.com/BusinessSolutions/Navi-son/default.aspx. Finally Axapta offers FMS, HRM, SCM, E-commerce, CRM and analytics. See http://www.micro-soft. com/BusinessSolutions/Axap-ta/default.aspx.
*1106 Best of breed vendors. Ninety percent of ERP sales are purchases of software “bundles” containing several pillars; rarely does a consumer purchase a single pillar. Tr. at 3815:10-13 (Hausman). FMS and HRM pillars typically are sold in a bundle along with additional kinds of ERP, such as CRM or SCM. Further, the discounts that are offered to potential consumers are based on the value of the entire bundle, not simply based upon the presence of an HRM or FMS pillar. Tr. at 3813:23-3814:1 (Hausman). Accordingly, when Oracle or PeopleSoft offers a discount on a bundle, it is doing so in order to ensure that the customer purchases all the pillars from Oracle or PeopleSoft, rather than turn to a best of breed vendor that specializes in selling a single kind of pillar. One best of breed vendor, Siebel, sells individual pillars of CRM. Testimony suggests Siebel is recognized industry-wide as selling high-quality CRM, equal to or better than the CRM pillars in Tier One software. Tr. at 3814:15-17 (Hasuman).
Outsourcing. Because of the extensive amount of training and maintenance involved in implementing ERP packages purchased from ERP vendors, some companies have chosen an alternative solution — outsourcing. Outsourcing occurs when a company hires another firm to perform business functions, often HRM functions. Tr. at 2198:15-2198:3 (Elzinga). A company may outsource a single HRM function, such as benefits, pensions or payroll, or it may choose to outsource its entire continuum of HRM needs. Tr. at 1648:14-22 (Bass). Many firms have outsourcing capabilities. Some of the out-sourcers discussed at trial include: Accenture, Fidelity, ADP, Mellon, Exult, Hewitt, Aon and Convergys. Outsourcing firms may process a company’s HR data using HRM software manufactured by an ERP vendor, such as Oracle, but some outsourcing firms use internally created HRM software (such as Fidelity using HR Access). Tr. at 3152:18-3153:23 (Sternklar).
In addition to individual vertical success, ERP vendors have tended to enjoy varying degrees of success in different geographic regions. SAP, for example, has been more successful at selling ERP to financial institutions in Europe than in North America. Tr. at 996:20-997:15 (Keating).
The FMS and HRM software sold to large customers is the same as that sold to mid-market customers. Tr. at 819:8-11 (Allen); Tr. at 1787:25-1788:2 (Wilmington); Tr. at 3436:24-3437:11 (Catz); Ex. D7166 (Morea 5/17/04 Dep) at Tr. 18:15-19:2(AMS); Ex. P3179 (Ciandrini 1/16/04 Dep) at Tr. 235:15-22. All the vendors— including Oracle, SAP, and PeopleSoft— have a single product “and that one product is sold up and down the line” to customers of all sizes. Ex. P3171 (Ellison 1/20/04 Dep) at Tr. 148:10-151:15. While some ERP vendors have introduced special licensing packages of FMS and HRM that are marketed to smaller customers, the actual software code in the FMS and HRM products sold to both large and mid-market customers is not different. Ex. P3070 (Prestipino 5/18/04 Dep) at Tr. 35:19-36:10 (Oracle); Tr. at 3437:5-9 (Catz). Oracle has recently launched its E-Business Suite Special Edition to appeal to its smallest customers — those who can use only 50 seats or less. It contains the same code as the software sold to the largest and middle-sized customers, but it arrives pre-eonfigured by the consulting organization. Tr. at 3436:24-3438:5 (Catz). It contains a subset of the modules found in Oracle’s E-Business Suite, including FMS but excluding HRM. Tr. at 3437:5-11 (Catz); Ex. P3070 (Prestipino 5/18/04 Dep) at Tr. 25:5-22, 32:19-33:19.
Despite the identity of code in each company’s ERP packaged product, ERP product offerings are not homogeneous. Whlie *1107 the ERP products offered by Oracle and PeopleSoft and other vendors perform the same or similar functions, these products are not uniform in their architecture, scalability, functionality or performance characteristics. Tr. at 897:23-899:3, 899:9-900:19, 901:6-902:15, 903:6-15, 946:18-20, 947:4-9, 992:23-993:7, 993:16-994:2, 996:20-997:15 (Keating). The product of each vendor possesses certain features or qualities so that none is a perfect substitute for any other. As the testimony indicated, and the court finds, no vendor is capable of meeting all of the high function needs, as defined by plaintiffs, of all customers. Tr. at 2085:3-5 (Iansiti).
Furthermore, because each packaged ERP product must be customized and configured to fit the software footprint of the customer, a packaged ERP product may, as fitted to one customer’s information technology footprint, differ significantly from the same packaged ERP product fitted to another customer’s footprint. Because of these facts, the court finds the ERP products in suit to be differentiated products.
The court also finds that ERP software is highly durable and, therefore, regarded by customers as a capital good. Campbell demo # 5,6,19; see also Tr. at 189:12-18 (Hatfield); Tr. at 1107:16-19 (Cichnowicz); Tr. at 1572:14-18 (Wolfe).
Customers almost always purchase a cluster of products such as Oracle’s E~ Business Suite that provide the customer with a stack of software and technology, which may include core HRM or FMS applications, add-on modules, customer-facing business applications such as CRM software and the infrastructure components (application servers and database) on which the applications run. Tr. at 3461:14-3462:5 (Catz); Tr. at 3807:21-3808:1 (Hausman). See, e.g., Exs. P1000-P1322 (Oracle discount request forms).
Plaintiffs’ Claim of Threatened Injury to Competition
Plaintiffs allege that the HRM and FMS sold by Oracle, PeopleSoft and SAP are the only HRM and FMS products that can appropriately be deemed “high function HRM and FMS.” FAC (Doc. # 125) ¶ 9 at 8.
Plaintiffs allege that these “high function” HRM and FMS products have the “scale and flexibility to support thousands of simultaneous users and many tens of thousands of simultaneous transactions and the ability to integrate seamlessly into bundles or ‘suites’ of associated HRM and FMS functions.” Id. ¶ 14 at 9. Plaintiffs allege that “high function” HRM and FMS products compete in a market that is separate and distinct from that of all other ERP products, such as SCM, CRM or mid-market HRM and FMS, the latter being HRM or FMS products designed for organizations having less demanding needs. These mid-market products include Oracle’s E-Business Suite Special Edition, SAP’s MySAP and All-in-One, PeopleSoft’s PeopleSoft EnterpriseOne and the products of ERP vendors such as Lawson and AMS.
Moreover, plaintiffs allege that this competition is geographically confined to the United States. Id. at ¶¶ 24, 26 at 13. Within this narrowly defined product and geographic market, plaintiffs allege that with limited and specially explained exceptions, only Oracle, PeopleSoft and, to a lesser degree, SAP’s United States arm, SAP America, are in effective competition. The proposed merger would therefore, in plaintiffs’ view, constrict this highly concentrated oligopoly to a duopoly of SAP America and a merged Oraele/PeopleSoft.
Oracle, predictably enough, contends that plaintiffs’ market definition is legally and practicably too narrow. Oracle contends that (1) “high function” HRM and *1108 FMS software does not exist; “high function” is simply a label created by plaintiffs; (2) there is just one market for all HRM and FMS ERP products; (3) many firms other than the three identified by plaintiffs compete in the business of developing, producing, marketing and maintaining HRM and FMS ERP software; (4) this competition plays out in many more products than those in the HRM and FMS pillars; (5) price competition comes from sources in addition to ERP software vendors and includes competition from firms that provide outsourcing of data processing, the integration layer of the “software stack” and from the durability and adaptability of enterprises’ installed base or legacy systems; (6) the geographic area of competition is worldwide or, at the very least, the United States and Europe; (7) the knowledgeable and sophisticated customers of ERP software would impede the exercise of any market power by a merged Oracle/People-Soft; and (8) potential entrants are poised to enter into competition, so that the proposed merger will not have an anticompeti-tive effect.
Taking up this dispute, the court first discusses the applicable law and economic principles that underlie its decision and then describes the parties’ contentions and evidence along with the court’s resolution of the disputed factual issues not previously discussed. This begins with the parties’ sharply differing definitions of the product and geographic markets and whether there is a level of concentration sufficient to trigger the presumption under United States v. Philadelphia Nat. Bank, 374 U.S. 321, 83 S.Ct. 1715, 10 L.Ed.2d 915 (1963), that the proposed transaction will lead to a substantial lessening of competition under the principles set forth in the Department of Justice and Federal Trade Commission Horizontal Merger Guidelines (Apr 2, 1992, as revised Apr 8, 1997) (“Guidelines”). The court then turns to an efficiency defense offered by Oracle before setting forth its conclusions of law.
In brief summary, for the reasons explained at length herein, the court’s findings and conclusions are as follows:
• plaintiffs have not proved that the product market they allege, high function HRM and FMS, exists as a separate and distinct line of commerce;
• plaintiffs have not proved the geographic market for the products of the merging parties is, as they allege, confined to the United States alone;
• plaintiffs have not proved that a post-merger Oracle would have sufficient market shares in the product and geographic markets, properly defined, to apply the burden shifting presumptions of Philadelphia Nat Bank;
• plaintiffs have not proved that the post-merger level of concentration (HHI) in the product and geographic markets, properly defined, falls outside the safe harbor of the Horizontal Merger Guidelines (Guidelines);
• plaintiffs have not proved that the ERP products of numerous other vendors, including Lawson, AMS and Microsoft, do not compete with the ERP products of Oracle, PeopleSoft and SAP and that these other vendors would not constrain a small but significant non-transitory increase in price by a post-merger Oracle;
• plaintiffs have not proved that outsourcing firms, such as Fidelity and ADP, would not constrain a small but significant non-transitory increase in price by a post-merger Oracle;
• plaintiffs have not proved that the ability of systems integrators to adapt, configure and customize competing ERP vendors’ products to the needs of the group of customers that plaintiffs contend constitute a separate and dis *1109 tinct product market would not constrain a small but significant non-transitory increase in price by a post-merger Oracle;
• plaintiffs have not proved that a post-merger Oracle and SAP would likely engage in coordinated interaction as the products of Oracle and SAP are not homogeneous, but are differentiated products, and that the pricing of these products is not standardized or transparent;
• plaintiffs have not proved localized product or geographic competition between Oracle and PeopleSoft that will be lessened as a result of the proposed merger as the merger would not create a dominant firm occupying a product or geographic space in which there is no serious competition;
• assuming that localized product or geographic competition exists between Oracle and PeopleSoft, plaintiffs have not proved that SAP, Microsoft and Lawson would not be able to reposition themselves in the market so as to constrain an anticompetitive price increase or reduction in output by a post-merger Oracle;
• plaintiffs have proved that products in the integration layer of the computer software industry and the presence of incumbent ERP systems would not constrain anticompetitive conduct on the part of a post-merger Oracle;
• Oracle has not proved efficiencies from the proposed merger sufficient to rebut any presumption of anticompeti-tive effects; should the court’s principal findings and' its conclusion that plaintiffs have not proved the proposed merger will likely lead to a substantial lessening of competition not be upheld on appeal, Oracle’s efficiency defense should not require further trial court proceedings.
■HORIZONTAL MERGER ANALYSIS
Section 7 of the Clayton Act prohibits a person “engaged in commerce or in any activity affecting commerce” from acquiring “the whole or any part” of a business’ stock or assets if the effect of the acquisition “may be substantially to lessen competition, or to tend to create a monopoly.” 15 USC § 18. The United States is authorized to seek an injunction to block the acquisition, 15 USC § 25, as are private parties and the several states, California v. American Stores Co., 495 U.S. 271, 110 S.Ct. 1853, 109 L.Ed.2d 240 (1990); Hawaii v. Standard Oil Co. of Cal., 405 U.S. 251, 258-59, 92 S.Ct. 885, 31 L.Ed.2d 184 (1972), and district courts have jurisdiction over such actions. 15 USC § 25; 28 USC § 1337(a). Plaintiffs have the burden of proving a violation of section 7 by a preponderance of the evidence.
To establish a section 7 violation, plaintiffs must show that a pending acquisition is reasonably likely to cause anti-competitive effects. See United States v. Penn-Olin Chem. Co., 378 U.S. 158, 171, 84 S.Ct. 1710, 12 L.Ed.2d 775 (1964) (noting that a section 7 violation is established when “the ‘reasonable likelihood’ of a substantial lessening of competition in the relevant market is shown”); United States v. Marine Bancorp., Inc., 418 U.S. 602, 622-23, 94 S.Ct. 2856, 41 L.Ed.2d 978 (1974); FTC v. H J Heinz Co., 246 F.3d 708, 713, 719 (D.C.Cir.2001). “ ‘Congress used the words “may be substantially to lessen competition” (emphasis supplied) to indicate that its concern was with probabilities, not certainties.’ ” Id. at 713 (quoting Brown Shoe Co. v. United States, 370 U.S. 294, 323, 82 S.Ct. 1502, 8 L.Ed.2d 510 (1962)). “Section 7 does not require proof that a merger or other acquisition [will] cause higher prices, in the affected market. All that is necessary is that the merger create an appreciable danger of such con *1110 sequences in the future.” Hospital Corp. of Am. v. FTC, 807 F.2d 1381, 1389 (7th Cir.1986). Substantial competitive harm is likely to result if a merger creates or enhances “market power,” a term that has specific meaning in antitrust law. See Eastman Kodak Co v. Image Tech. Services Inc., 504 U.S. at 451, 464 (1992); Rebel Oil Co. v. Atlantic Richfield Co., 51 F.3d 1421, 1434 (9th Cir.1995).
Market Definition
In determining whether a transaction will create or enhance market power, courts historically have first defined the relevant product and geographic markets within which the competitive effects of the transaction are to be assessed. This is a “necessary predicate” to finding anticompetitive effects. United States v. E. I. du Pont & Co., 353 U.S. 586, 593, 77 S.Ct. 872, 1 L.Ed.2d 1057 (1957). Market definition under the case law proceeds by determining the market shares of the firms involved in the proposed transaction, Philadelphia Nat. Bank, 374 U.S. 321, 83 S.Ct. 1715, 10 L.Ed.2d 915, the overall concentration level in the industry and the trends in the level of concentration. United States v. Aluminum Co. of Am., 377 U.S. 271, 277-79, 84 S.Ct. 1283, 12 L.Ed.2d 314 (1964); United States v. Von’s Grocery Co., 384 U.S. 270, 272-74, 86 S.Ct. 1478, 16 L.Ed.2d 555 (1966). A significant trend toward concentration creates a presumption that the transaction violates section 7. United States v. Baker Hughes Inc., 908 F.2d 981, 982-83 (D.C.Cir.1990) (Thomas, J.). See also United States v. Citizens & Southern Nat. Bank, 422 U.S. 86, 120-22, 95 S.Ct. 2099, 45 L.Ed.2d 41 (1975). In other words, plaintiffs establish a prima facie case of a section 7 violation by “show[ing] that the merger would produce a firm controlling an undue percentage share of the relevant market, and [would] result [ ] in a significant increase in the concentration of firms in that market.’ ” Heinz, 246 F.3d at 715 (quoting Philadelphia Nat. Bank, 374 U.S. at 363, 83 S.Ct. 1715) (alterations in original). Under Philadelphia Nat Bank, a post-merger market share of 30 percent or higher unquestionably gives rise to the presumption of illegality. 374 U.S. at 364, 83 S.Ct. 1715 (“Without attempting to specify the smallest market share which would still be considered to threaten undue concentration, we are clear that 30% presents that threat.”).
To rebut this presumption, defendant may “show that the market-share statistics give an inaccurate account of the merger’s probable effects on competition in the relevant market.” Heinz, 246 F.3d at 715 (internal quotation marks and alterations omitted). See also Baker Hughes, 908 F.2d at 987; California v. Am. Stores Co., 872 F.2d 837, 842-42 (9th Cir.1989), rev’d on other grounds, 495 U.S. 271, 110 S.Ct. 1853, 109 L.Ed.2d 240 (1990); FTC v. Warner Communications, 742 F.2d 1156, 1164 (9th Cir.1984); Olin Corp. v. FTC, 986 F.2d 1295, 1305-06 (9th Cir.1993). Arguments related to efficiencies resulting from the merger may also be relevant in opposing plaintiffs’ case. See FTC v. Tenet Health Care Corp., 186 F.3d 1045, 1054-55 (8th Cir.1999); FTC v. Staples, Inc., 970 F.Supp. 1066, 1088 (D.D.C. 1997). “ ‘If the defendant successfully rebuts the presumption [of illegality], the burden of producing additional evidence of anticompetitive effects shifts to [plaintiffs], and merges with the ultimate burden of persuasion, which remains with the government at all times.’” Heinz, 246 F.3d at 715 (quoting Baker Hughes, 908 F.2d at 983) (first alteration in original).
An application of the burden-shifting approach requires the court to determine (1) the “line of commerce” or product market in which to assess the transaction; (2) the “section of the country” or geographic market in which to assess the *1111 transaction; and (3) the transaction’s probable effect on competition in the product and geographic markets. See Marine Bancorporation, 418 U.S. at 618-23, 94 S.Ct. 2856; FTC v. Harbour Growp Investments LP, 1990 WL 198819 at *2 n. 3 (D.D.C.1990). See also FTC v. Swedish Match, 131 F.Supp.2d 151, 156 (D.D.C. 2000); FTC v. Cardinal Health, Inc., 12 F Supp 2d 34, 45 (D.D.C.1998); Staples, 970 F.Supp. at 1072.
Both the Supreme Court and appellate courts acknowledge the need to adopt a flexible approach in determining whether anticompetitive effects are likely to result from a merger. Reflecting their “generality and adaptability,” Appalachian Coals v. United States, 288 U.S. 344, 360, 53 S.Ct. 471, 77 L.Ed. 825 (1933), application of the antitrust laws to mergers during the past half-century has been anything but static. Accordingly, determining the existence or threat of anticompetitive effects has not stopped at calculation of market shares. In Hospital Co'i'p of Am the court upheld the FTC’s challenge to the acquisition of two hospital chains, but noted that “the economic concept of competition, rather than any desire to preserve rivals as such, is the lodestar that shall guide the contemporary application of the antitrust laws, not excluding the Clayton Act.” 807 F.2d at 1386. Hence, the court held that it was appropriate for the FTC to eschew reliance solely on market percentages and the “very strict merger decisions of the 1960s.” Id. at 1386. In addition to market concentration, probability of consumer harm in that case was established by factors such as legal barriers to new entry, low elasticity of consumer demand, inability of consumers to move to distant hospitals in emergencies, a history of collusion and cost pressures creating an incentive to collude. 807 F.2d at 1388-89.
In United States v. Waste Management, 743 F.2d 976 (2d Cir.1984), the court of appeals reversed a finding of a section 7 violation based on market shares and pri-ma facie illegality under Philadelphia Nat Bank, one made even though there were few barriers to new entry into the market. The trial court had erroneously ignored the Supreme Court’s holding in United States v. General Dynamics, 415 U.S. 48&, 94 S.Ct. 1186, 39 L.Ed.2d 530 (1974), that a prima facie case may still be rebutted by proof that the merger will not have anti-competitive effects. A finding of market shares and consideration of the Philadelphia Nat Bank presumptions should not end the court’s inquiry.
The trend in these cases away from the “very strict merger decisions of the 1960s,” Hospital Corp. of Am., 807 F.2d at 1386, is also reflected in the Guidelines. The Guidelines view statistical and non-statistical factors as an integrated whole, avoiding the burden shifting presumptions of the case law. The Guidelines define market power as “the ability profitably to maintain prices above competitive levels for a significant period of time.” Guidelines § 0.1. Five factors are relevant to the finding of market power: (1) whether the merger would significantly increase concentration and would result in a concentrated market, properly defined; (2) whether the merger raises concerns about potential adverse competitive effects; (3) whether timely and likely entry would deter or counteract anticompetitive effects; (4) whether the merger would realize efficiency gains that cannot otherwise be achieved; and (5) whether either party would likely fail in the absence of the merger. Guidelines, § 0.2.
In defining the market, the Guidelines rely on consumer responses. Starting with the smallest possible group of competing products, the Guidelines then ask “whether ‘a hypothetical monopolist over that group of products would profitably impose at *1112 least a “small but significant and nontran-sitory” [price] increase [“(SSNIP)”],’ ” generally deemed to be about five percent lasting for the foreseeable future. United States v. Sungard Data Sys., Inc., 172 F Supp 2d 172, 182 (D.D.C.2001) (quoting Guidelines § 1.11). If a significant number of customers respond to a SSNIP by purchasing substitute products having “a very considerable degree of functional interchangeability” for the monopolist’s products, then the SSNIP would not be profitable. E.I. du Pont, 351 U.S. at 399, 76 S.Ct. 994. See Guidelines § 1.11. Accordingly, the product market must be expanded to encompass those substitute products that constrain the monopolist’s pricing. The product market is expanded until the hypothetical monopolist could profitably impose a SSNIP. Id. § 1.11. Similarly, in defining the geographical market, the Guidelines hypothesize a monopolist’s ability profitably to impose a SSNIP, again deemed to be about five percent, in the smallest possible geographic area of competition. Id. § 1.21. If consumers respond by buying the product from suppliers outside the smallest area, the geographic market boundary must be expanded. Id.
Once the market has been properly defined, the Guidelines set about to identify the firms competing in the market and those likely to enter the market within one year. Guidelines § 1.32. Following these steps, the Guidelines calculate the market share of each participant, followed by the Herfindahl-Hirschman Index (HHI) concentration measurement for the market as a whole. Guidelines § 1.5. The HHI is calculated by squaring the market share of each participant, and summing the resulting figures. Id. The concentration standards in the Guidelines concern the (1) pre-merger HHI (HHI^, (2) the post-merger HHI (HHI2) and (3) the increase in the HHI resulting from the merger, termed delta HHI (□HHI). See Andrew I Gavil, William E Kovacic and Jonathan B. Baker, Antitrust Law in Perspective; Cases, Concepts and Problems in Competition Policy, 480-84 (Thomson West, 2002). The Guidelines specify safe harbors for mergers in already concentrated markets that do not increase concentration very much. For example if the post-merger HHI is between 1000 and 1800 (a moderately concentrated market) and the □HHI is no more than 100 points, the merger is unlikely to be presumed illegal. Guidelines § 1.51. Likewise, if the post-merger HHI is above 1800 (a highly concentrated market) and the □HHI is no more than 50 points, the merger will not be presumed illegal. Id.
Notwithstanding these statistical data, the Guidelines next focus on the likely competitive effects of the merger. Guidelines § 2.0; Baker Hughes, 908 F.2d at 984 (“Evidence of market concentration simply provides a convenient starting point for a broader inquiry into future competitiveness * * *.”). The Guidelines recognize that anticompetitive effects may arise in two contexts. First, the Guidelines address the lessening of competition through coordinated interaction between the merged firm and remaining rivals. Guidelines § 2.1. Second, the Guidelines address the anticompetitive effects based on unilateral action. Id. § 2.2.
Anticompetitive Effects
Coordinated Effects
In analyzing potential coordinated effects, a court is concerned that the merger may diminish competition by “enabling the firms * * * more likely, more successfully, or more completely to engage in coordination interaction.” Guidelines § 2.1. This behavior can be express or tacit (implied by silence), and the behavior may or may not be lawful in and of itself. Id. The Guidelines explicitly recognize that successful coordinated interaction “entails *1113 reaching [1] terms of coordination that are profitable to the firms involved and [2] an ability to detect and punish [cheating].” Id. § 2.1. See also FTC v. Elders Grain, Inc., 868 F.2d 901, 905 (7th Cir.1989); Hospital Corp. of Am., 807 F.2d at 1386— 87. Examples of “terms that are profitable” include common pricing, fixed price differentials, stable market shares and customer or territorial restrictions. Guidelines § 2.11
Factors that increase the likelihood of coordination include product homogeneity, pricing standardization and pricing transparency. Brooke Group Ltd. v. Brown & Williamson Tobacco Corp., 509 U.S. 209, 238, 113 S.Ct. 2578, 125 L.Ed.2d 168 (1993); Elders Grain, 868 F.2d at 905. Plaintiffs do not contend that any of those conditions are presented in the proposed merger which must, therefore, be analyzed for unilateral anticompetitive effects.
Unilateral Effects
There is little case law on unilateral effects merger analysis. Few published decisions have even discussed the issue, at least using the term “unilateral effects.” See, e.g., Swedish Match, 131 F Supp 2d at 168; New York v. Kraft Gen. Foods, Inc., 926 F.Supp. 321, 333-35 (S.D.N.Y.1995); Guidelines § 2.2. But, as the court demonstrates below, “unilateral effects” is primarily a new term to address antitrust issues that courts have in other contexts considered for quite some time.
Unilateral effects result from “the tendency of a horizontal merger to lead to higher prices simply by virtue of the fact that the merger will eliminate direct competition between the two merging firms, even if all other firms in the market continue to compete independently.” Carl Shapiro, Mergers with Differentiated Products, 10 Antitrust 23, 23 (Spring 1996). Unilateral effects are thought to arise in primarily two situations, only the second of which is alleged in this case. See Roscoe B Starek III & Stephen Stockum, What Makes Mergers Anticom-petitive?: “Unilateral Effects’’ Analysis Under the 1992 Merger Guidelines, 63 Antitrust LJ 801, 803 (1995); Guidelines §§ 2.21, 2.22; Phillip E Areeda, Herbert Hovenkamp & John L. Solow, 4 Antitrust Law ¶ 910 (Aspen, rev ed 1998) (subdividing unilateral effects theories into four categories).
The first situation involves a “dominant firm and a ‘fringe’ of competitors producing a homogeneous product.” Starek & Stockum, 63 Antitrust LJ at 803. In this situation, the dominant firm has a substantial cost advantage over the fringe competitors and, therefore, can restrict output to obtain an above-marginal cost price.
The second situation, and the one here applicable, concerns differentiated products. Starek & Stockum, 63 Antitrust LJ at 803; Guidelines § 2.21. Competition in differentiated product markets, such as ERP products, is often described as “monopolistic competition.” There is a notable and interesting literature on this subject commencing with the path-breaking and independent insights of two notable economists. See Edward Chamberlin, The Theory of Monopolistic Competition (Harvard, 1933, 1938), Joan Robinson, The Theory of Imperfect Competition (St Martin’s, 1933, 2d ed 1969). The admirably clear exposition found in Paul A. Samuelson & William D. Nordhaus, Economics 187-89 (McGraw-Hill, 17th ed 2001) makes apparent this nomenclature.
The market demand curve shows the quantity of a good that would be purchased in the market at each price, other things being equal. Id. at 760. A seller’s “own,” or “residual,” demand curve shows the quantity of the good offered by the seller that would be purchased from the seller at each price, other things being equal. Under perfect competition, the in *1114 dividual seller faces a horizontal (each additional unit brings the same revenue), or perfectly elastic, demand curve because nothing the seller can do alters demand for the seller’s product. Id. at 148. The seller is a price taker. Because the seller’s demand curve is horizontal, the seller’s marginal revenue curve is also horizontal and the seller continues to produce until its marginal cost is equal to the market price or average revenue and profits, as economists define them, are zero. See id. fig 8-2 and text at 148-50.
[[Image here]]
The adjacent figure, borrowed from Samuelson & Nordhaus, Economics fig 9-4 at 178, illustrates the different picture facing the monopolist. Its demand curve is not horizontal but reflects the inverse relationship between price and the quantity demanded. Because it is the only seller of the product, the pure or natural monopolist faces not the horizontal demand curve of the perfectly competitive firm, but the sloping demand curve of the entire market. In the graph, the monopolist is able to maximize profit at the intersection of marginal cost and revenue by reducing output to 4 and raising the price to $120, which exceeds marginal cost. The monopolist thus derives a “monopoly rent” equal to the number of units sold times the difference between the market price (G) and the monopolist’s average cost (F), algebraically» (G — F) x 4. It is this reduction in output and elevation of price that has been the historic concern of antitrust.
*1115 [[Image here]]
Firms in perfect competition “produce homogeneous product” so that “price is the only variable of interest to consumers, and no firm can raise its price above marginal cost without losing its entire market share.” Jean Tiróle, The Theory of Industrial Organization at 277 (MIT, 1988). Differentiated products are imperfect substitutes representing as they do different features or characteristics that appeal variously to different customers. Because no product is a perfect substitute of another in a differentiated products market, each seller continues to face a downward sloping demand curve. Like a pure monopolist, the seller of a differentiated product, facing a downward sloping, or less than perfectly elastic, demand curve, maximizes its profit by pricing above marginal cost. See Samuelson & Nordhaus, Economics fig 10-3 and text at 188-89.
*1116 [[Image here]]
Like a seller in a perfectively competitive market, however, sellers in a “competitive” differentiated products market do not obtain monopoly rents. In differentiated product markets with few barriers to entry, firms will introduce products that are increasingly close, although not perfect substitutes, for the other products in the market. The introduction of additional products causes the demand curve faced by each seller to shift downward and leftward until, at long run equilibrium, the demand curve intersects the average cost curve of the seller (defined as economists define costs to include a reasonable profit) eliminating the monopolistic rent (ACGB). See id. fig 10-4 and text at 188-89.
Differentiated product markets hence share some characteristics of both a pure monopoly and perfect competition, in that “prices are above marginal costs but economic profits have been driven down to zero.” Id. at 189 (describing “economic profits” as supra-normal profits or monopoly rents). Firms selling differentiated *1117 products have some “market power” in that they are able to exert some control over the prices they obtain although this does not rise to the level of “monopoly power.” See Shapiro, 63 Antitrust LJ 24 n. 4 (citing the economic literature).
The Guidelines provide some instruction on the necessary elements of a unilateral effects claim involving differentiated products under section 7.
Substantial unilateral price elevation in a market for differentiated products requires [1] that there be a significant share of sales in the market accounted for by consumers who regard the products of the merging firms as their first and second choices, and [2] that repositioning of the non-parties’ product lines to replace the localized competition lost through the merger be unlikely.
Guidelines § 2.21.
Although the Guidelines’ discussion quoted above may be a helpful start, the factors described therein are not sufficient to describe a unilateral effects claim. First, the Guidelines’ discussion, at least in section 2.21, emphasizes only the relative closeness of a buyer’s first and second choices. But the relative closeness of the buyer’s other choices must also be considered in analyzing the potential for price increases. The Guidelines later acknowledge as much in section 2.212, which recognizes that if a buyer’s other options include “an equally competitive seller not formerly considered, then the merger is not likely to lead to a unilateral elevation of prices.” Accordingly, a plaintiff must prove not only that the merging firms produce close substitutes but also that other options available to the buyer are so different that the merging firms likely will not be constrained from acting anticompet-itively.
Second, the Guidelines require only a demonstration of some “significant share of sales in the market accounted for by customers” that rank the merging firms first and second. Id. § 2.21. “Measures of the ‘closest substitutes’ or ‘second choices’ of inframarginal purchasers of Product A are only relevant to the degree that infra-marginal and marginal consumers have similar preferences. However, essentially by definition, marginal and inframarginal consumers do not share similar preferences.” Christopher A. Vellturo, Creating an Effective Diversion: Evaluating Mergers with Differentiated Products, 11 Antitrust 16, 18 (Spring 1997); Gregory J. Werden & George A. Rozanski, The Application of Section 7 to Differentiated Products Industries: The Market Definition Dilemma, 8 Antitrust 40, 41 (Summer 1994) (“[T]here is no reason why the shares in any delineated market in a differentiated products industry are indicative of the relative importance of each merging firm as a direct competitor of the other.”).
In sum, it appears that four factors make up a differentiated products unilateral effects claim. First, the products controlled by the merging firms must be differentiated. Products are differentiated if no “perfect” substitutes exist for the products controlled by the merging firms. See Samuelson & Nordhaus, Economics at 187-89; Areeda, Hovenkamp & Solow, 4 Antitrust Law ¶ 914d (“By ‘significant’ we mean product differentiation that goes to fairly fundamental differences in product design, manufacturing costs, technology, or use of inputs.”). Second, the products controlled by the merging firms must be close substitutes. Products are close substitutes if a substantial number of the customers of one firm would turn to the other in response to a price increase. Third, other products must be sufficiently different from the products controlled by the merging firms that a merger would make a small but significant and non-transitory price increase profitable for the merging *1118 firms. Finally, repositioning by the non-merging firms must be unlikely. In other words, a plaintiff must demonstrate that the non-merging firms are unlikely to introduce products sufficiently similar to the products controlled by the merging firms to eliminate any significant market power created by the merger. These four factors substantially track the analysis in Areeda, Hovenkamp and Solow. Areeda, Hoven-kamp & Solow, 4 Antitrust Law ¶ 914f at 68-69.
The essential elements of such a differentiated products unilateral effects claim are quite similar to those in “standard” antitrust analysis. In standard antitrust analysis, the court considers both “demand elasticity” and “supply elasticity” in determining whether anticompetitive effects are likely. Rebel Oil, 51 F.3d at 1436. In other words, courts determine the degree to which price increases will cause marginal buyers to turn to other products or marginal suppliers to increase output of the product. Considerations of demand and supply elasticity also motivate the factors outlined by the court for a differentiated products unilateral effects analysis. The factors considering the relative substi-tutability of the products of the merging and non-merging firms, factors 1 to 3, essentially address demand-side substitut-ability and the repositioning factor, factor 4, essentially addresses supply-side substi-tutability.
Antitrust analysis of differentiated product markets is hardly new. See, e.g., E. I. du Pont, 351 U.S. at 392-93, 76 S.Ct. 994 (describing the concepts of monopolistic competition and differentiated product markets); Areeda, Hovenkamp & Solow, 4 Antitrust Law ¶ 914c (suggesting that early railroad merger cases could be viewed as unilateral effects cases). Indeed, as noted above, defining a geographic market involves exactly same concept of localized competition that motivates differentiated products unilateral effects analysis.
Areeda, Hovenkamp and Solow persuasively contend that “the appropriate conclusion [under a unilateral effects analysis] is that the merger has facilitated the emergence of a new grouping of sales capable of being classified as a relevant market.” Id. ¶ 913b. This “new grouping of sales” is one “in which the merging firms have either a monopoly or else a dominant share.” Id ¶ 914f at 69. In an example of two merging firms, B and C, Areeda, Hoven-kamp and Solow state that “the merger does not create such a market because a cartel of firms B and C would also have been able to increase price profitably, indicating that B and C were already a relevant market.” Id. ¶ 914a at 60. But of course, “before their union, B and C felt one another’s competition, as well as that of other firms, more significantly than after the merger.” Id. Areeda, Hovenkamp and Solow also later note that “the sufficiently similar output of other firms must be included” in the relevant market. Id. ¶ 914f at 70.
In a unilateral effects case, a plaintiff is attempting to prove that the merging parties could unilaterally increase prices. Accordingly, a plaintiff must demonstrate that the merging parties would enjoy a post-merger monopoly or dominant position, at least in a “localized competition” space.
Unilateral effects analysis shares many similarities with standard coordinated effects antitrust analysis. But there are also notable differences.
Relevant markets defined in terms of “localized competition” may be much narrower than relevant markets defined in typical cases in which a dominant position is required. Judicial experience cautions against the use of qualitative factors to define narrow markets. This judicial ex *1119 perience arises, in part, from the rise (and fall) of the “submarkets” doctrine.
In Brown Shoe, the Supreme Court stated that submarkets may constitute relevant product markets. “The outer boundaries of a product market are determined by the reasonable interchangeability of use or the cross-elasticity of demand between the product itself and substitutes for it. However, within this broad market, well-defined submarkets may exist which, in themselves, constitute product markets for antitrust purposes.” Brown Shoe, 370 U.S. at 825, 82 S.Ct. 1502 (citing E. I. du Pont, 353 U.S. at 593-95, 77 S.Ct. 872) (footnote omitted).
Properly construed, Brown Shoe suggests merely that the technical definition of a relevant market in an antitrust case may be smaller than a layperson would normally consider to be a market. The use of the term “submarket” may be useful in “overcoming] the first blush or initial gut reaction” to a relatively narrowly defined market. See Staples, 970 F.Supp. at 1074 (defining the relevant market as “the sale of consumable office supplies through office supply superstores”).
Focusing on “submarkets” may be misleading, however, because “the same proof which establishes the existence of a relevant product market also shows (or * * * fails to show) the existence of a product submarket.” H J. Inc. v. International Tel. & Tel. Corp., 867 F.2d 1531, 1540 (8th Cir.1989). See also Olin, 986 F.2d at 1301. Defining a narrow “submarket” tends to require a relatively long laundry list of factors, which creates the danger of narrowing the market by factors that have little economic basis. Courts and commentators suggest that the use of the sub-markets doctrine has, in fact, misled courts into “identifying] artificially narrow groupings of sales on the basis of noneco-nomic criteria having little to do with the ability to raise price above cost.” Areeda, Hovenkamp & Solow, 4 Antitrust Law ¶ 914a at 60. See also Allen-Myland, Inc. v. IBM, 33 F.3d 194, 208 n. 16 (3d Cir.1994); Satellite Television & Associated Resources v. Continental Cablevision of Va., Inc., 714 F.2d 351, 355 n. 5 (4th Cir.1983).
The similarities between the submarkets doctrine generally and localized competition in unilateral effects cases are difficult to miss. Indeed, commentators have been quick to note the potential for “localized competition” analysis to devolve into an unstructured submarket-type analysis. See Areeda, Hovenkamp & Solow, 4 Antitrust Law ¶ 914a at 60; Starek & Stoc-kum, 63 Antitrust LJ at 814-15 (arguing that the Guidelines’ focus on localized competition should not “be used as a tool for rehabilitating discredited ‘submarket’ analysis”).
Furthermore, judicial rejection of markets narrowly defined to a single manufacturer’s product has been even more pronounced than judicial skepticism about narrowly defined submarkets. See, e.g., E. I. du Pont, 353 U.S. at 592-93, 77 S.Ct. 872 (refusing to define a market limited to cellophane); TV Communs. Network, Inc. v. Turner Network Television, Inc., 964 F.2d 1022, 1025 (10th Cir.1992) (refusing to define a market limited to TNT cable provision in the greater Denver area); Town Sound & Custom Tops, Inc. v. Chrysler Motors Corp., 959 F.2d 468, 479-80 (3d Cir.1992) (en banc) (refusing to define a market limited to Chrysler products); Gall v. Home Box Office, Inc., 1992 WL 230245 at *4 (S.D.N.Y.) (“[T]he natural monopoly every manufacturer has in its own product simply cannot serve as the basis for antitrust liability.”). Cf. Eastman Kodak, 504 U.S. at 481-82, 112 S.Ct. 2072 (upholding denial of summary judgment in an installed base context).
As emphasized in E. I. du Pont:
*1120 [O]ne can theorize that we have monopolistic competition in every nonstandardized commodity with each manufacturer having power over the price and production of his own product. However, this power that, let us say, automobile or soft-drink manufacturers have over their trademarked products is not the power that makes an illegal monopoly. Illegal power must be appraised in terms of the competitive market for the product.
351 U.S. at 393, 76 S.Ct. 994 (footnotes omitted).
Merely demonstrating that the merging parties’ products are differentiated is not sufficient. Instead, a plaintiff must demonstrate product differentiation sufficient to sustain a small but significant and non-transitory price increase.
Additionally, defining markets in terms of “localized competition” may result in markets defined so narrowly that one begins to question whether the market constitutes a “line of commerce” as required by section 7. One concern is that the market is defined so narrowly that it encompasses an insubstantial amount of commerce. In Philadelphia Nat Bank, the Supreme Court found a “workable compromise” between a geographic market narrowly defined in terms of bank offices in the immediate neighborhood or more expansively defined to include the banks available only to large borrowers. 374 U.S. at 360-61, 83 S.Ct. 1715. Another concern is that the market is defined so narrowly it fails to capture the potential effects of the merger. For example, it might be inappropriate to focus on a single city in analyzing the effects of a merger between sellers who compete on a much larger scale. Cf. Staples, 970 F.Supp. at 1073 nn. 5-6 (analyzing the likelihood of anticompetitive effects in forty-two metropolitan areas).
Even if a narrow market definition would be appropriate, it may be more difficult to identify “clear breaks in the chain of substitutes” sufficient to justify bright-line market boundaries in differentiated products unilateral effects cases. The conventional ideal market boundary divides products within the market, which are freely substitutable with one another, from products outside the market, which are poor substitutes for the products within the market. See United States v. Rockford Memorial Corp., 717 F.Supp. 1251, 1260 (N.D.Ill.1989) (emphasis added), aff'd, 898 F.2d 1278 (7th Cir.1990). In differentiated products unilateral effects cases, a “spectrum” of product differences, inside and outside the market boundary, is more likely. In re Super Premium Ice Cream Distribution Antitrust Litig., 691 F.Supp. 1262 (N.D.Cal.1988), aff'd. sub. nom., Haagen-Dazs Co. v. Double Rainbow Gourmet Ice Creams, Inc., 895 F.2d 1417, 1990 WL 12148 (9th Cir.1990) (table). In discussing unilateral effects, Shapiro has written:
[A]ny attempt to make a sharp distinction between products “in” and “out” of the market can be misleading if there is no clear break in the chain of substitutes: if products “in” the market are but distant substitutes for the merging products, their significance may be overstated by inclusion to the full extent that their market share would suggest; and if products “out” of the market have significant cross-elasticity with the merging products, their competitive significance may well be understated by their exclusion.
Shapiro, 10 Antitrust at 28. See also Edward Chamberlin, Product Heterogeneity and Public Policy, 40 Am. Econ Rev. (Papers & Procs.) 85, 86-87 (1950).
Additionally, to the extent that clear breaks are difficult to identify, attempts to create defensible market boundaries are likely to be based on relatively vague prod *1121 uct characteristics. Product characteristics that are too vague do not meet section 7’s requirement that the relevant market be “well-defined.” See Tenet Health Care, 186 F.3d at 1052.
A closer look at product differentiation demonstrates further difficulties in defining the relevant market in differentiated product unilateral effects cases. Price is one, but only one, of many ways in which to differentiate a product. A market of homogeneous goods can be seen as a market in which sellers have only one dimension in which to differentiate their product. One expects sellers in such a market to “differentiate” their products by lowering the price until price equals marginal cost. On the other hand, a differentiated product “market” is a market in which sellers compete along more dimensions than price. As a result, products competing against one another in a differentiated product market may have widely different prices. That products with widely different prices may, in fact, be in the same market complicates market definition considerably.
The “Cellophane fallacy” may complicate matters even further. This phenomenon takes its name from an error in the Supreme Court’s logic E.I. du Pont. In E. I. du Pont, the plaintiff was the primary manufacturer of cellophane. The Supreme Court held that the relevant market included “all flexible wrappings” because cross-price elasticities of demand indicated that an increase in the price currently charged for cellophane would cause a significant number of purchasers to turn to other flexible wrapping products.
The error in the logic of E. I. du Pont is that “ ‘[t]he existence of significant substitution in the event of further price increases or even at the current price does not tell us whether the defendant already exercises significant market power.’ ” Eastman Kodak, 504 U.S. at 471, 112 S.Ct. 2072 (quoting Phillip Areeda & Louis Ka-plow, Antitrust Analysis ¶ 340(b) (Aspen, 4th ed 1988)). Stated slightly differently, because a monopolist exercises market power by increasing price until the cross-price elasticity of demand is so high that a further price increase would be unprofitable, a high cross-price elasticity of demand at current prices, by itself, does not demonstrate that the seller lacks market power.
The implications of the Cellophane fallacy on market definition in differentiated product market cases would seem to suggest caution. Courts should be wary of defining markets so broadly that a seller’s existing market power is missed. On the other hand, in differentiated product markets, some measure of market power is inherent and an unduly narrow product market definition proves too much. In merger analysis, the court is concerned primarily with determining whether the merger would enhance market power, not whether market power currently exists.
In sum, defining the relevant market in differentiated product markets is likely to be a difficult task due to the many non-price dimensions in which sellers in such markets compete. Further, it may be difficult to determine currently existing market power and separate this from enhanced market power due to the merger.
The inability clearly to define a market suggests that strong presumptions based on mere market concentration may be ill-advised in differentiated products unilateral effects cases. As noted by Starek and Stockum, “it is generally misleading to suggest that a firm “controls” a certain market share in the absence of an analysis beyond market concentration.” Starek & Stockum, 63 Antitrust LJ at 804. See also Jerry A. Hausman & Gregory K. Leonard, Economic Analysis of Differentiated Products Mergers Using Real World Data, 5 Geo. Mason L. Rev. 321, 337-39 (1997). *1122 Such a concern applies with equal force to differentiated products unilateral effects claims. Furthermore, in differentiated products unilateral effects cases, the merging parties’ combined market shares relative to competitors may be less relevant than the size of their market shares in determining whether anticompetitive effects are likely. See Gregory J. Werden & Luke M. Froeb, The Effects of Mergers in Differentiated Products Industries: Logit Demand and Merger Policy, 10 J. L. Econ. & Org. 407, 413 (1994).
Accordingly, a strong presumption of anticompetitive effects based on market concentration is especially problematic in a differentiated products unilateral effects context.
Despite the problems with qualitative analyses, modern econometric methods hold promise in analyzing differentiated products unilateral effects cases. Merger simulation models may allow more precise estimations of likely competitive effects and eliminate the need to, or lessen the impact of, the arbitrariness inherent in defining the relevant market. For example, some merger simulation methods compensate for potential errors in market definition. A model advanced by Werden and Froeb uses a set of “inside goods” and a set of “outside goods.” Id. at 410. The model contains a parameter, beta, that controls for the substitutability among the inside goods and another parameter, epsilon, that controls for the substitutability between the inside and outside goods. Id. To the extent the set of goods considered as “inside goods” is defined narrowly, epsilon increases. Id. at 424-25. The increase in epsilon increases the predicted amount of substitution to outside goods. Accordingly, error in defining the product market too narrowly will be offset, at least to some extent, by the increase in epsilon.
In sum, differentiated products unilateral effects analysis shares many similarities to “standard” antitrust analysis. The primary differences are that the relevant market is likely to be smaller and more difficult to define and that quantitative analyses may be robust.
In analyzing antitrust claims, courts have considered both “circumstantial” and “direct” evidence of anticompetitive effects. See Rebel Oil, 51 F.3d at 1434. Even though “direct” evidence of the potential for anticompetitive harm from a merger is not literally available, merger analyses range from highly qualitative (“circumstantial”) to highly quantitative (“direct”), depending on the data available for a particular market. Qualitative analyses of antitrust claims are most often structural. In a structural analysis, anti-competitive effects are presumed if a plaintiff demonstrates undue concentration in a well-defined market. See Philadelphia Nat. Bank, 374 U.S. at 363, 83 S.Ct. 1715; Baker Hughes, 908 F.2d at 982. A relevant market may be defined by reference to Brown Shoe’s “practical indicia.” 370 U.S. at 325, 82 S.Ct. 1502. Once the relevant market is defined, market shares are calculated and inferences are drawn from the degree of concentration.
The Guidelines adopt a structural approach for addressing unilateral effects claims that closely mirrors traditional structural analysis. See Guidelines § 2.211. The biggest weakness in the Guidelines’ approach appears to be its strong reliance on particular market share concentrations. Under the Guidelines, anticompetitive effects are presumed “[w]here market concentration data fall outside the safeharbor regions of Section 1.5, the merging firms have a combined market share of at least thirty-five percent, and where data on product attributes and relative product appeal show that a significant share of purchasers of one merging firm’s product regard the other as *1123 their second choice,” unless “rival sellers likely would replace any localized competition lost through the merger by repositioning their product lines.” Id. at §§ 2.211, 2.212.
A presumption of anticompetitive effects from a combined share of 35% in a differentiated products market is unwarranted. Indeed, the opposite is likely true. To prevail on a differentiated products unilateral effects claim, a plaintiff must prove a relevant market in which the merging parties would have essentially a monopoly or dominant position. In Rebel Oil, the Ninth Circuit noted that a market share of 30% is “presumptively insufficient to establish the power to control price.” 51 F.3d at 1438.
Market definitions, statistical presumptions and likelihood of unilateral anticom-petitive effects are all issues on which the parties contended vigorously and presented much evidence. To these, the court now turns.
CONTENTIONS, EVIDENCE AND FINDINGS
“Defining the relevant market is critical in an antitrust case because the legality of the proposed merger[ ] in question almost always depends upon the market power of the parties involved.” Cardinal Health, 12 F.Supp.2d at 45. Yet the precise characteristics that plaintiffs have used to describe the line of commerce allegedly affected by the proposed transaction changed throughout the course of this litigation. And the evidence of market shares presented to enable the court to apply the Philadelphia Nat Bank presumptions or make the HHI calculations of the Guidelines is, given the mountain of evidence plaintiffs presented, startling sparse.
Plaintiffs’ Proposed Product Market Definition
Plaintiffs offer a product market of high function HRM and FMS and a geographic market of the United States.
Four elements constitute plaintiffs’ definition of high function HRM software as alleged in the FAC: “[1] Human Resource Management (HRM) [2] software and accompanying services [3] that can be integrated into suites of associated functions from a single vendor [4] with performance characteristics that meet the demands of multifaceted organizations with high-level functional needs.” FAC (Doc. # 125) ¶ 23(a) at 12.
Likewise, four elements constitute plaintiffs’ definition of high function FMS software as alleged in the FAC: “[1] Financial Management Services (FMS) [2] software and accompanying services [3] that can be integrated into suites of associated functions from a single vendor [4] with performance characteristics that meet the demands of multifaceted organizations with high-level functional needs.” Id. ¶ 23(b) at 12-13.
The FAC also notes certain performance characteristics of high function software:
Customers with high-level functional needs (“enterprise customers”) require products that can support their ongoing business processes and reporting requirements that may stretch across multiple jurisdictions (often requiring support for foreign languages and reporting requirements), multiple legal entities or divisions within the organization and multiple lines of business. These products must have the scale and flexibility to support thousands of simultaneous users and many tens of thousands of simultaneous transactions, and the ability to integrate seamlessly into bundles or “suites” of associated HRM and FMS functions. Most importantly, these *1124 products must have the flexibility through configuration options or other means to be matched to the administrative and reporting processes of each unique customer.
Id. ¶ 14 at 9. ,
Plaintiffs clarified their allegations at the request of the court during the trial by submitting a statement of definitions, some of which were joined by defendant. Jt. Sub. Definitions (Doc. # 332). In these definitions, plaintiffs omitted “and accompanying services” from the second element alleged in the FAC. Plaintiffs also relegated the FAC’s third element regarding integration to a mere sub-element of the performance characteristics described in the FAC’s fourth element. Finally, plaintiffs describe four “performance capabilities.” Products in, the - market are (1) “highly” configurable, (2) “seamlessly” integrated software products that support (3) “multiple” languages, currencies and legal regimes with (4) “virtually unlimited” scalability. See id. at 2-4 & n. 2.
This definition shifted somewhat in plaintiffs’ proposed findings of fact and conclusions of law. Plaintiffs clarified the definition to include “licensing and maintenance” rather than “licensing and accompanying services,” as alleged in the FAC’s second element. See Pis. Prop FF (Doc. #356) at ¶¶ 3.1.1 — 3.1.2. Plaintiffs also added an element to the formal definition, claiming that high function software “pro-vid[es] robust functionality that allows organizations to go beyond the basics.” Id. at ¶ 3.1.3.4.
Even though not stated as part of the formal definition of high function software, plaintiffs scatter throughout their proposed findings of fact other characteristics of ERP software in an apparent attempt further to narrow the relevant market.
First, plaintiffs point to the claimed strength of high function software in “core” applications. See, e.g., id. at ¶ 3.6.2.1.
Second, plaintiffs emphasize that high function customers purchase high function software. Id. (Doc. # 356).
Third, plaintiffs emphasize the brand value of the software vendor. Factors that promote vendor brand value include previous experience in a particular industry, research and development spending and local sales forces. See, e.g., id. (Doc. # 356) at ¶¶ 3.2.4.3 — 3.2.4.5.
Fourth, plaintiffs note the incumbent advantage software vendors have in competing for further sales with a customer who has that vendor’s product as part of its existing footprint. See, e.g., id. at ¶ 7.3.2.1.18 (pointing to testimony that “Bearing Point has identified more than 1,200 companies that now have an Oracle Financials and PeopleSoft HR footprint”).
Fifth, plaintiffs emphasize the alleged strength of Oracle' and PeopleSoft in certain industry verticals, such as insurance. See, e.g., id. at ¶¶ 7.2.3.10 — 7.2.3.13.
Sixth, plaintiffs describe high-function software as being “Able to Accommodate Rapid Growth, Acquisitions and Reorganizations.” Id. at ¶ 2.2.5.
Seventh, plaintiffs define high function software as allowing users to consolidate data across multiple organizations while still allowing the user to drill down to the original data. Id. at ¶ 2.2.6.
In their post-trial brief too, plaintiffs adjusted their proposed product market definition. They eliminated the “robust functionality” factor and incorporated two of the factors scattered throughout their proposed findings of fact into the more formal definition of high function software. The .newly incorporated factors are that high function software must accommodate rapid growth and complicated business structures.
*1125 At closing argument, plaintiffs disclaimed reliance on high function software’s claimed strength in “core” functionality in defining the relevant market and accused Oracle of creating confusion “by limiting the relevant market to basic ‘core’ functionality.” PL Post Brief (Doc. # 366) at 10 n. 17.
Added together, plaintiffs propose a very restricted product market definition: HRM and FMS integrated suites sold to large complex enterprises (“high function FMS and HRM market”). See id (Doc. #366) at 8. Plaintiffs have defined the asserted relevant product market using a large number of factors. In sum, the competition between Oracle and PeopleSoft that plaintiffs claim will be impaired bears the following characteristics:
Product characteristics:
• Software licensing and maintenance;
• HRM and FMS (as separate markets); Customer characteristics:
• High function needs;
• Oracle or PeopleSoft are major vendors in their software footprint;
Performance characteristics:
• Scalable;
• Highly configurable;
• Seamlessly integratable;
• Able to accommodate rapid growth, acquisitions and reorganizations;
• Able to reflect actual units of business; and
• Able to adapt to industry specific requirements.
Plaintiffs contend that this product market does not include mid-market vendors, best-of-breed solutions, incumbent solutions or outsourcing. Id. (Doc. #366) at 14-19.
Plaintiffs’ Evidence of a High Function HRM & FMS Market
In support of their proposed product market definition and theory of anticom-petitive effects, plaintiffs presented at trial or through deposition ten customer witnesses, five industry witnesses, two systems integration witnesses, three expert witnesses, a few others who appear mostly to have been presented to fill a gap or two in the evidence or, because every trial seems to need some, for spice (e.g., the Ellison and Phillips videotape deposition testimony) and a plethora of exhibits, some of these also for spice (e.g., Ex. P2290). The court will not attempt to recount or even summarize the entire evidentiary record. Given the quantity of evidence, that would be unduly time-consuming and is unnecessary. It suffices to note that the laboring oar of the plaintiffs’ case was pulled by the customer witnesses (whom plaintiffs’ counsel described as their strongest witnesses), by some of the systems integrator and industry witnesses and by the experts.
Customer Witnesses
Michael Gorriz, Vice President of Information Technology Business at Daimler-Chrysler (Daimler), testified about his company’s large and complex needs regarding HRM software. Tr. at 1368 (Gor-riz). Daimler has about 365,000 employees worldwide in about 100 manufacturing facilities. Tr. at 1368:6-13 (Gorriz). Since 1996, Daimler has used SAP as its financial management software. Tr. at 1370:4-10 (Gorriz). Daimler requires highly functional HRM to accommodate its large number of employees and to comply with the differing labor laws and union agreements in different countries. Tr. at 1371:9-12 (Gorriz). For its HRM needs, Daimler currently uses PeopleSoft. Daimler chose PeopleSoft based upon its reputation and the fact that companies of comparable size to Daimler have had success with PeopleSoft HRM. Tr. at 1375:13-21 (Gorriz). But when Daimler was first searching for an HRM vendor in 1996, Gorriz stated that “only SAP, PeopleSoft *1126 or Oracle could serve [Daimler’s] needs for the HR management.” Tr: at 1376:9-11 (Gorriz). Gorriz stated that Daimler considered no other vendors. Tr. at 3716:18-19 (Gorriz). Daimler’s legacy system was “too old” for the company seriously to consider upgrading. Tr. at 1376:24 (Gor-riz). Daimler did not consider outsourcing to be an option because Daimler’s HRM requirements were, Gorriz testified, “too complex.” Tr. at 1377:24-25 (Gorriz). Further, if Oracle, SAP or PeopleSoft were to increase their price for HRM by 10 percent, Gorriz stated that Daimler “would not consider any offer” from any other vendors. Tr. at 1381:16 (Gorriz).
Bob Bullock, Senior Vice President and Chief Information Officer of CH2M Hill, testified about the ERP needs of that civil and environmental engineering firm. CH2M Hill has 14,000 employees, 200 worldwide offices and over $2 billion in annual revenue. CH2M Hill has used Oracle FMS since 1993, but in 2002 the company decided to replace its legacy HRM software. Bullock stated that through consultation with the Gartner Group, CH2M Hill was given a list of HRM vendors. CH2M Hill did not seriously consider SAP, as it “was a very complex product” and had a “reputation for being a costly product.” Tr. at 207:19-20 (Bullock). In Bullock’s opinion, there were only two candidates, Oracle and People-Soft. Id at 208:7-8 (Bullock). CH2M Hill never considered outsourcing, Lawson or remaining on its legacy system. Tr. at 210:8, 211:12, 216:8-9 (Bullock). Oracle and PeopleSoft both offered initial bids between $1.5 and 41.6 million. Bullock stated that if this price had been 10 percent higher, CH2M Hill would not walked away from the deal with Oracle or People-Soft. Tr. at 218-19 (Bullock).
Curtis Wolfe, CIO for the State of North Dakota, testified about the state’s process of picking an ERP vendor. Tr. at 1532 (Wolfe). North Dakota has approximately 10,000 full and part-time employees, 58 state agencies and a budget of $5 billion. Tr. at 1533 (Wolfe). In 2002, the state decided to buy a full ERP program that included FMS and HRM. Tr. at 1534:10-16 (Wolfe). North Dakota had a unique need in that it required that its ERP serve the state’s higher education facilities as well. Id. North Dakota had six vendors submit proposals: Oracle, PeopleSoft, SAP, SCT, Jenzabar (a partner of Lawson) and Microsoft’s Great Plains. Tr. at 1543:21-22 (Wolfe). The state eliminated SAP, Great Plains and Jenzabar almost immediately. SAP was too expensive, while Jenzabar and Great Plains did not have the required functionality. Tr. at 1545-46 (Wolfe). SCT did not make the final round; while SCT met the functionality for the higher education area, it could not do so with state agency needs. Tr. at 1551:1-4 (Wolfe). Oracle and PeopleSoft were in head to head competition and Wolfe testified that he believes that this caused the state to get a $6 to $8 million lower final bid from each vendor. Tr. at 1561:10-11 (Wolfe). If these final offers had been 10 percent higher, Wolfe stated that North Dakota would not have turned to Lawson, Microsoft, SCT, outsourcing or writing its own software. Tr. at 1569-1570 (Wolfe).
Kenneth Johnsen, Chief of Technology for Pepsi Americas, testified as to his concerns about the Oracle/PeopleSoft merger. Pepsi Americas is the second largest bottler of Pepsi-brand soft drinks within the Pepsi system and the third largest bottler worldwide. Tr. at 1723:25-1724:1 (John-sen). Pepsi Americas has over 15,000 employees and annual revenues of about $3.2 billion. Tr. at 1724:5-10 (Johnsen). Pepsi Americas uses PeopleSoft ERP in its North America operations and SAP ERP in its European operations. Tr. at 1727:13-14 (Johnsen). Johnsen testified that he has “a concern” about the impact *1127 of this merger on the long-term effectiveness of the PeopleSoft ERP. Tr. at 1734:23 (Johnsen). Johnsen is concerned that a post-merger Oracle, while agreeing to maintain the PeopleSoft ERP, will not provide enhancements to the functionality of the software (i.e., upgrades). Tr. at 1737: 1-9 (Johnsen). To Johnsen this leaves Pepsi Americas with two options: constantly upgrade with point solutions (not his desired choice) or buy ERP from a new vendor. When asked what vendors he could turn to meet his ERP needs, John-sen claims there are no options outside of Oracle, PeopleSoft and SAP. Tr. at 1739:14 (Johnsen).
Scott Wesson, Senior Vice President and Chief Information Officer of AIMCO, discussed the company’s choices for FMS and HRM software. Tr. at 1126 (Wesson). AIMCO is the largest owner and operator of apartment buildings in the United States. Tr. at 1127:7-8 (Wesson). The company owns approximately 2000 complexes in 47 states and the District of Columbia. AIMCO has over 6,500 employees and an annual revenue of about $1.5 billion. Tr. at 1127:9-24 (Wesson). For its FMS, AIMCO uses PeopleSoft’s financial suite. For its HR payroll systems, AIMCO currently uses Lawson. Tr. at 1129:8,21 (Wesson). In 2002, AIMCO began to reevaluate its HRM options and it hired Towers Perrin consult in this process. Towers Perrin told AIMCO that only three vendors could meet AIMCO’s HRM needs: PeopleSoft, Oracle and SAP. Tr. at 1132:7-8 (Wolfe). (There was no objection to the question that elicited this response). Wesson stated that AIMCO decided not to upgrade to the latest version of Lawson because it would have cost AIMCO “about the same * * * as it would to go with a new system” and also, Lawson “[was] lacking some key features” that AIMCO was looking for. Tr. at 1133:5-11 (Wolfe). AIMCO was deciding between Oracle and PeopleSoft when Oracle first made its tender offer to PeopleSoft. Tr. at 1143: 9-10 (Wolfe). Wesson stated that because of this proposed merger, he believes PeopleSoft gave him a “very good deal” on the HRM. Tr. at 1144:17 (Wolfe). Wesson testified that Oracle agreed to match any price offered by PeopleSoft. Tr. at 1145:5 (Wolfe). Wesson said AIM-CO ultimately chose PeopleSoft because PeopleSoft had guaranteed to pay AIMCO three times the contract price should there be a “change of ownership” at PeopleSoft. Tr. at 1146:14, 1147:6-16 (Wolfe). AIMCO is expecting to implement the PeopleSoft system in late 2004 or early 2005. Tr. at 1148:10 (Wolfe). Moreover, Wesson stated, AIMCO does not consider outsourcing to be a viable option because it is not quick to respond to “last minute changes,” such as new benefits programs. Tr. at 1150:10 (Wolfe). Best of breed solutions are too expensive for AIMCO to consider. Tr. at 1150:22-24 (Wolfe).
Richard Cichanowicz, Vice President of Systems Integration of Nextel, testified about the wireless services company’s ERP needs. Nextel has 13 million subscribers, over $8 billion in annual revenue 17,000 [transcript incorrect] employees. See Tr. at 1052:25-1053:3 (Cichanowicz). Before 2002, Nextel had been using Peo-pleSoft HRM, Oracle FMS and Ariba SCM. Tr. at 1058:9-11 (Cichanowicz). In 2002, however, Nextel determined that using one integrated solution would provide more operational efficiency. Tr. at 1061:7-9 (Cichanowicz). Nextel received advice from six consulting firms, which informed Nextel that Oracle, SAP and PeopleSoft could meet those software needs. Tr. at 1066:13-19 (Cichanowicz). Nextel then sent RFPs to Oracle and Peoplesoft. Tr. at 1067:25-1068:3 (Cichanowicz). Nextel did not seriously consider SAP because it was already using Oracle for FMS and PeopleSoft for HRM and believed that conversion costs and risks for those two *1128 vendors would be lower. Tr. at 1068:4-17 (Cichanowicz). Nextel ultimately chose Peoplesoft, based on its scoring of vendor criteria such as functionality, ease of integration, scalability, audits, costs and relationship confidence. See Tr. at 1071:20-1072-22 (Cichanowicz). Even after it had chosen PeopleSoft, however, Nextel continued to negotiate with Oracle for leverage purposes until the signing of the December 2002 contract with PeopleSoft. Tr. at 1073:11-20 (Cichanowicz). Cichanowicz stated that if the price of the Oracle or PeopleSoft licenses had been 10 percent higher, Nextel would not have considered a best of breed approach, writing or building its own ERP software, outsourcing, staying with its previous system or using SAP or any other United States vendor. Tr. at 1077:16-1080:25 (Cichanowicz).
Mary Elizabeth Glover, Vice President of Information Technology at Greyhound Lines, testified about her company’s foray into the market for HRM software. Greyhound is in the bus transportation business in both the United States and Canada. The company employees some 16,000 people and has annual revenues of around $1.2 billion. Tr. at 1459-1460 (Glover). For its FMS, Greyhound uses Oracle in the United States and J D Edwards in Canada. Tr. at 1464:11-21 (Glover). For its HRM, Greyhound uses a product called HR1 in the United States and HR2000 in Canada. The company outsources its payroll to ADP. Tr. at 1465:11 (Glover). Glover stated that the HR incumbent systems are “very old” and no longer meet the needs of the company. Tr. at 1466:21-25 (Glover). Further, she testified that outsourcing is too expensive for Greyhound. Tr. at 1467:12-15 (Glover). For these reasons, in 2001, Greyhound began a potential procurement process for new HRM software. Tr. at 1468:17-18 (Glover). The company hired CDG & Associates to match Greyhound with potential vendors who met their HRM needs. The firm narrowed the selection down to only four vendors: Oracle, PeopleSoft, Lawson and Ultimate Software. Tr. at 1470:11 (Glover). Greyhound never considered SAP because the consulting firm believed they were too costly. Tr. at 1470:16 (Glover). Ultimate Software was eliminated soon thereafter because of lack of functionality. Tr. at 1470:24-25 (Glover). Greyhound eliminated PeopleSoft as being too costly. Between Oracle and Lawson, Greyhound found that Oracle had more functionality; therefore, Lawson was eliminated. But before Greyhound made a final choice, Glover stated that the company decided to give PeopleSoft a second look. Upon reexamination, Greyhound determined that both Oracle and PeopleSoft could meet the company’s needs, with the company preferring PeopleSoft over Oracle. Tr. at 1483:6-9 (Glover). Unfortunately, the events of September 11, 2001, a new CEO and a decrease in profits caused Greyhound to lose the funds necessary to purchase the software. Tr. at 1490:6-11 (Glover). But Glover stated that should Greyhound ever decide to purchase HRM software, this proposed merger would make the purchase more costly, as Greyhound’s only choices were Oracle and PeopIeSoft. Tr. at 1495:13-21. Without the competition between the two, Glover foresees prices increasing. Tr. at 1495:13-21 (Glover).
Phillip Maxwell, Senior Vice President and Chief Information Officer of the Nei-man Marcus Group (NMG), testified about the ERP needs of the specialty retailer. NMG has properties located throughout the country, approximately 15,000 employees and $3 billion in annual sales. Tr. at 652:3-13 (Maxwell). NMG formerly had used FMS software that was originally from MSA, a vendor purchased by Dun & Bradstreet and then GEAC subsequent to NMG’s installation of the software. Tr. at *1129 655:15-22 (Maxwell). In 2002, NMG decided to replace its FMS software and began conferring with individuals in its business and technology units, three consulting firms and the Gartner Group. See Tr. at 662:1-663:11 (Maxwell). After examining vendors’ functionality, experience in retail, price and size/stability, NMG narrowed its choices to Oracle and PeopleSoft. Tr. at 665:6-20 (Maxwell). NMG did not consider SAP because of SAP’s lack of strong presence in the retail vertical and Maxwell’s opinion that SAP is “very expensive to implement.” Tr. at 669:11-16 (Maxwell). Had the cost of Oracle or Peo-pleSoft FMS software been 10 to 20 percent higher, NMG would not have considered SAP, any other FMS vendor, legacy software or internally developed software. Tr. at 669:17-670:15 (Maxwell). Based on price, a high level comparison and detailed GAP analysis, NMG eventually selected Oracle to provide it with FMS software. See Tr. at 671:8-673:2 (Maxwell).
NMG also began licensing HRM software from Oracle in 2003, though it has not yet begun to implement that software. See Tr. at 674:9-11, 676:14-18 (Maxwell). NMG went through a similar process in evaluating HRM software as it did in evaluating FMS software. Tr. at 684:25-685:20 (Maxwell). As with the FMS software, NMG concluded that Oracle and PeopleSoft were its only viable alternatives. See Tr. at 686:13-16 (Maxwell). NMG did not believe that SAP suited its needs as a retailer. See Tr. at 686:11-18 (Maxwell). Had the cost of the Oracle or PeopleSoft HRM software been 10 to 20 percent higher, NMG would not have considered other HRM vendors, legacy software, internally developed software or outsourcing. Tr. at 686:19-687:13 (Maxwell). NMG eventually selected the Oracle HRM software, but based on a 70 to 80 percent higher target price than previously predicted, NMG has delayed implementation of the Oracle HRM software to look for cost-reducing options. Tr. at 676:19-677:13 (Maxwell). But Maxwell testified that, even with the 80 percent price increase, NMG has not abandoned the Oracle HRM. Tr. at 677:20-25 (Maxwell).
Laurette Bradley, Senior Vice President of Information Technology at Verizon, testified about Verizon’s current procurement of new HRM software. Tr. at 577 (Bradley). Verizon is a telecommunications company with a “majority holding in four of five different countries.” Tr. at 580:22-25 (Bradley). Verizon has minor investments in over 30 countries worldwide with an annual revenue of approximately $66 billion. Id. Bradley testified that 49 percent of Verizon’s labor is unionized worldwide, which places “significant demands upon [the] ERP systems, particularly [the] HR and payroll systems” because each union contract, from each jurisdiction, must be reflected and managed regarding payroll, vacation, absences, and personal days. Tr. at 583:6-15 (Bradley). Prior to October 2003, Verizon had used two different HRM programs, one from PeopleSoft and one from SAP. Tr. at 583:23 (Bradley). The PeopleSoft HRM was used to manage the former BellAtlantic part of the company and SAP HRM was used to manage the former GTE part of the company. Tr. at 584:1^1 (Bradley). The same is true of Verizon’s FMS. But in October 2003, Verizon decided to consolidate the two systems as far as HRM software. Tr. at 584:11-12 (Bradley). Verizon chose PeopleSoft HRM for the entire company and as of the date of the trial, the new software was being implemented. Id. Bradley testified that a merger between Oracle and People-Soft makes her very concerned that Oracle will not be interested in upgrading or further “developing” current PeopleSoft software. Tr. at 592:5, 593:3-10 (Bradley). Bradley does not want to lose the constant “care, feeding, repair, and evolution” that PeopleSoft now offers to its customers. *1130 Tr. at 592:17-18 (Bradley). When asked what other vendors Verizon could turn to in obtaining FMS and HRM that meet Verizon’s complex and international needs, Bradley listed only Oracle, PeopleSoft and SAP. Tr. at 598:7-8 (Bradley). But Bradley did testify that Verizon is “constantly” considering outsourcing its entire HR management, but so far has determined that the risks are just too high. Tr. at 604:20-21 (Bradley).
Bradley admitted that Verizon already outsources its 401(k) stock plans and medical and dental benefits. Tr. at 604:12-14 (Bradley). Finally, Bradley stated that if Oracle, PeopleSoft or SAP increased prices by 10 percent, Verizon would not turn to any other vendors for their FMS and HRM. Tr. at 606:23-25, 607:1-3 (Bradley). Further, Verizon would not use its off-shore information technology staff to develop an in-house FMS or HRM system in response to a 10 percent increase. Tr. at 607:12-15 (Bradley).
Finally, Scott Hatfield, Chief Information Officer of Cox Communications discussed his company’s ERP software needs. Tr. at 87:8-11 (Hatfield). Cox is the third largest cable television operator in the United States, delivering video service to about six and half million households. Tr. at 89:11-14 (Hatfield). Cox has a presence in 30 states and about 21,000 employees. Cox has annual revenues of over $6 billion. Tr. at 89:22-25 (Hatfield). Hatfield testified that Cox uses PeopleSoft HRM to handle payroll, recruitment, benefits programs and training. Tr. at 94:14-19 (Hatfield). In 1995, during the HRM vendor procurement process, Cox only considered Oracle and SAP as other potential vendors of HRM. Tr. at 96:12 (Hatfield). Hatfield testified that while Cox had considered outsourcing its HRM altogether, it had decided against doing so because the company needed to have a “tight integration” between its HRM and CMS, which could not be outsourced. Tr. at 97:17-19 (Hatfield).
Regarding FMS, in 2003, Cox decided to change from J D Edwards to a new vendor. Cox hired Accenture to consult in this process. Tr. at 114:22-25 (Hatfield). Accenture gave Cox a list of three vendors of FMS that could meet Cox’s needs: Oracle, PeopleSoft and SAP. Tr. at 115:9-10 (Hatfield). Hatfield stated that no other firms were “brought to his attention.” Tr. at 121:18 (Hatfield). Cox eliminated SAP because no one in the company had any real experience with SAP and Hatfield did not want to “be starting from scratch.” Tr. at 118:3 (Hatfield). Hatfield stated that Cox wanted Oracle and PeopleSoft to know they were the final two competing for Cox’s FMS business and that Cox asked the two vendors to give their best prices. Tr. at 126:1-3 (Hatfield). Cox ultimately chose Oracle as its FMS vendor based upon highest level of functionality ratings. Tr. at 129:1-5 (Hatfield). Finally, Hatfield stated that if Oracle or People-Soft’s prices had been 10 percent higher, Cox would not have turned to Lawson, Great Plains, best of breed solutions, outsourcing or writing its own FMS software. Tr. at 136:14-138:23.
In the main, and contrary to the characterization of plaintiffs’ counsel before trial, the court found the testimony of the customer witnesses largely unhelpful to plaintiffs’ effort to define a narrow market of high function FMS and HRM. Each of these witnesses had an impressive background in the field of information technology. They appeared knowledgeable and well informed about their employers’ ERP needs and resources. And the court does not doubt the sincerity of these witnesses’ beliefs in the testimony that they gave. What the court questions is the grounds upon which these witnesses offered their opinions on the definition of the product *1131 market and competition within that market.
The test of market definition turns on reasonable substitutability. E. I. du Pont, 351 U.S. 377, 76 S.Ct. 994, 100 L.Ed. 1264. This requires the court to determine whether or not products have “reasonable interchangeability” based upon “price, use and qualities * * Id. at 404, 76 S.Ct. 994. What, instead, these witnesses testified to was, largely, their preferences.
Customer preferences towards one product over another do not negate intei'-changeability. See R R Donnelley & Sons Co., 120 FTC 36, 54 n. 65 (1995) (citing Robert Pitofsky, New Definitions of the Relevant Market and the Assault on Antitrust, 90 Colum. L. Rev. 1805, 1816 (1990) (“There will almost always be classes of customers with strong preferences * * * but to reason from the existence of such classes to a conclusion that each is entitled to * * * a separate narrow market definition grossly overstates the market power of the sellers.”)). The preferences of these customer witnesses for the functional features of PeopleSoft or Oracle products was evident. But the issue is not what solutions the customers would like or prefer for their data processing needs; the issue is what they could do in the event of an anticompetitive price increase by a post-merger Oracle. Although these witnesses speculated on that subject, their speculation was not backed up by serious analysis that they had themselves performed or evidence they presented. There was little, if any, testimony by these witnesses about what they would or could do or not do to avoid a price increase from a post-merger Oracle. To be sure, each testified, with a kind of rote, that they would have no choice but to accept a ten percent price increase by a merged Oraele/PeopleSoft. But none gave testimony about the cost of alternatives to the hypothetical price increase a post-merger Oracle would charge: e.g., how much outsourcing would actually cost, or how much it would cost to adapt other vendors’ products to the same functionality that the Oracle and PeopleSoft products afford.
If backed by credible and convincing testimony of this kind or testimony presented by economic experts, customer testimony of the kind plaintiffs offered can put a human perspective or face on the injury to competition that plaintiffs allege. But unsubstantiated customer apprehensions do not substitute for hard evidence.
While listening to the testimony of these customer witnesses, it became clear to the court that these witnesses represent a group of extremely sophisticated buyers and users of information technology; they have decades of experience in negotiating in this field. This made more evident the failure of these witnesses to present cost/benefit analyses of the type that surely they employ and would employ in assessing an ERP purchase. The evidence at trial established that ERP customers have choices outside the integrated suites of Oracle, PeopleSoft and SAP. Indeed, Glover’s testimony showed that — as Oracle contends- — customers have some leverage by virtue of their existing installed base “to do nothing” and thereby resist anti-competitive price increases by ERP vendors. Although the court is not convinced that this is a long-term option due to the ever changing business and legal environment in which enterprises operate, this option does afford ERP customers some limited protection and leverage. At any rate, plaintiffs’ customer witnesses did not, in their testimony, provide the court with data from actual or probable ERP purchases and installations to demonstrate that the witnesses’ employers would have had no choice but to submit to a SSNIP imposed by a post-merger Oracle.
*1132 The court, therefore, finds that these witnesses did not establish by a preponderance of the evidence that the products offered by Oracle, PeopleSoft and SAP are in a distinct line of commerce or product market from those offered by other ERP vendors. The court finds that these witnesses did not establish that it was more likely than not that customers of a post-merger Oracle would have no choice but to submit to a small but significant non-transitory price increase by the merged entity. These findings do not rest alone on the court’s skepticism about the testimony of plaintiffs’ customer witnesses.
Oracle, too, presented customer witnesses, although a much smaller number of such witnesses. Brian Mearns, Director of Personnel Service Delivery for Bank Of America (BA), testified about BA and Fleet Boston’s (Fleet) needs and decisions regarding HRM and FMS software. Tr. at 3276:2-21 (Mearns). In April 2004, BA acquired Fleet. Tr. at 3276:10 (Mearns). Mearns had held the title of Director of HR Service Delivery at Fleet prior to the acquisition. Mearns stated that Fleet had personnel of over 50,000 worldwide, with investment and mortgage offices in 32 states and throughout South America, Europe and Asia. Tr. at 3280:14-3281:11 (Mearns). Mearns testified that Fleet had used PeopleSoft HRM software since 1996. Tr. at 3286:18-20 (Mearns). In 2002, Fleet sought to upgrade its PeopleSoft HRM software to encompass increased functionality. But the $12 million price tag was too much for Fleet’s appropriation committee and Mearns was told that upgrading PeopleSoft was not an option. Tr. at 3289-3290:11. Based upon this turn of events, Mearns stated that Fleet instead turned to outsourcing to meet its HRM needs. Tr. at 3290:24-25 (Mearns). The search to find an outsourcing firm that could meet all of Fleet’s needs led to five candidates: Mellon, Hewitt, Exult, Accenture and Fidelity. Tr. at 3293:1-2 (Mearns). Fidelity “best met [the] business objectives and selection criteria” that Fleet required. Tr. at 3295:11-12 (Mearns). After implementation of the new outsourcing solution, Mearns stated that Fidelity’s systems were “very configurable to meet [Fleet’s] requirements.” Tr. at 3297:12-14 (Mearns). After BA acquired Fleet, Mearns gave a presentation to BA executives about Fleet’s experience with outsourcing and the capability of Fidelity. Tr. at 3300:14-17 (Mearns). Based upon this presentation, BA decided to outsource all of its HRM functions to Fidelity. Tr. at 3300:20-22 (Mearns).
Charles Peters, Senior Executive Vice President for Emerson Electric Company (Emerson), was also called by Oracle to testify about other viable substitutes to high function ERP. Emerson is a global manufacturing company operating in six industries including climate technologies (air conditioning and heating components), motor and appliance components and components for large industrial equipment. Tr. at 1190-1191:15 (Peters). Emerson’s annual revenue exceeds $15 billion and its workforce includes about 110,000 employees in over 50 countries. Tr. at 1191:18-25 (Peters). Within these six industries, Emerson has over 40 divisions. Tr. at 1193:11 (Peters). Some of these divisions, standing alone, have global operations and revenues in the billions of dollars. Tr. at 1193:19-20 (Peters). Many of these divisions operate their own HRM and FMS software. Tr. at 1198:7-8 (Peters).
One aspect of Peters’ job is to provide “options” to each division regarding their choices for handling FMS and HRM needs. Tr. at 1197:6-18 (Peters). Peters stated these “options” include ERP vendors, outsourcing, best of breed solutions, in-house solutions and extending incumbent systems. Tr. at 1198:7-19 (Peters). Peters testified that one of his divisions *1133 will not implement Oracle ERP because their in-house software fully meets its needs. Tr. at 1211:1-18 (Peters). Further, Peters discussed the increasing role that outsourcing to Asia or the Philippines plays in the HR area of many divisions. Tr. at 1214:7-16 (Peters). Finally, Peters stated that he did not believe that Emerson divisions would have to pay more for Oracle ERP if the proposed merger is consummated. Tr. at 1235:11-14 (Peters).
In so testifying, Peters cited to a recent negotiation he conducted with Oracle concerning ERP for one division. During the negotiations, Peters stated, PeopleSoft was never a contender. Tr. at 1235:16 (Peters). The possibility of using PeopleSoft was not leverage that Peters could use to advantage in seeking to obtain a lower price from Oracle. Emerson still received a competitive price from Oracle. Tr. at 1235:18-24 (Peters). Accordingly, Peters stated that he does not believe that the presence or absence of PeopleSoft is a factor that constrains Oracle pricing. Id.
To be sure, the testimony of the Oracle witnesses, like that of the plaintiffs’ customer witnesses, entailed some speculation about the presence or absence of People-Soft in the market. But the Oracle witnesses testified about concrete and specific actions that they had taken and been able to complete in order to meet their firms’ information processing needs, apart from relying on the three ERP vendors that plaintiffs contend are a market unto themselves. Hence, the court finds on this basis, as well as an assessment of the witnesses’ credibility, that the testimony of the Oracle customer witnesses was more believable than that of the plaintiffs’ witnesses, despite the greater number of the latter.
Plaintiffs’ Expert: Iansiti
In addition to the customer witnesses, plaintiffs presented the expert opinion testimony of Marco Iansiti, a professor of business administration at the Harvard Business School. Iansiti’s expertise lies in operations management and information technology. Iansiti also has experience as a consultant for companies in the “software space.” Tr. at 2020:24 (Iansiti). Iansiti thus brought an academic perspective that basically echoed the testimony of the customer witnesses. The court is satisfied that Iansiti is well qualified to opine on features of ERP products.
Iansiti was asked to identify the vendors whose ERP products would meet the needs of a “large and complex enterprise.” Tr. at 2024:4 (Iansiti). Iansiti examined the product documents and analysts reports of 148 ERP vendors. Tr. at 2025:10 (Iansiti). Iansiti testified that only the products of Oracle, PeopleSoft and SAP possess the functionality adequate to meet the needs of such an enterprise. With regard to Lawson, Iansiti testified that its HRM product can handle only three levels of an organization and its FMS product five levels and thus is wholly inadequate for a large and complex enterprise. Tr. at 2047:3-5 (Iansiti). By contrast the People-Soft and Oracle products can capture “unlimited levels of organization.” Tr. at 2047:17 (Iansiti).
Iansiti testified that Microsoft Business Solutions (MBS) provides four ERP products: Navison, Axapta, Great Plains and Solomon. Tr at 2054:7-8 (Iansiti). But MBS sells exclusively through resellers and thus lacks the kind of direct relationship necessary to furnish the level and specific services required by large and complex enterprises. Tr. at 2054:17-2055:11 (Iansiti). Microsoft will not, in Iansiti’s view, have a single product to “rationalize” its present four ERP products until 2009. Tr. at 2058:25-2061:10 (Iansiti). Iansiti expressed doubts that Microsoft will be able to develop products competitive with those of PeopleSoft, Ora- *1134 ele and SAP because Microsoft’s business model is radically different from that of these three companies. Tr. at 2061:11-2063:15 (Iansiti). Iansiti also saw no developments in internet technology or the integration layer that would likely replace the functionalities of the ERP offerings of PeopleSoft, Oracle and SAP. Tr. at 2077:12-2080:11 (Iansiti).
The court finds that Iansiti’s testimony fails to establish a product market. Iansiti did not claim to have performed an economic study of the ERP industry. Tr. at 2082:5-20 (Iansiti). He conceded that there is not a “clear line or demarcation” to distinguish enterprises that have high functional needs from “lower function or mid-market needs.” Tr. at 2088:7-2090:21 (Iansiti). Furthermore, Iansiti conceded that a number of companies that would appear to meet the criteria of large and complex enterprises have satisfied their ERP requirements with the products of vendors other than PeopleSoft, Oracle and SAP and have satisfied their needs from outsourcing or from their legacy systems. See Tr. at 2091:5-2095:3, 2100:1-2113:15 (Iansiti). Because of his lack of economic analysis and his inability to identify articu-lable product market boundaries (a key issue in a horizontal merger case), the court finds that Iansiti failed to establish a clearly defined product market along the lines alleged by plaintiffs.
Systems Integrators
Plaintiffs presented the testimony of two systems integrator witnesses in an effort to prove the existence of a separate high function ERP market. One of these witnesses, Perry Keating of BearingPoint, however, rebutted as much as supported plaintiffs’ positions regarding market participants and likelihood of entry into the market.
Keating is the Senior Vice President of BearingPoint, one of the largest consulting companies in the world. Tr. at 857:12-15 (Keating). BearingPoint is involved in “management consulting” which includes the “implementation of financial [and] human resource * * * solutions.” Tr. at 858:4-7 (Keating). At the outset of his testimony, Keating made clear that Bear-ingPoint has taken no position either for or against the proposed merger. Tr. at 858:11-18 (Keating). Keating stated that BearingPoint “wishes both [Oracle and PeopleSoft] well.” Id.
Keating started off by supporting plaintiffs’ product market definition, stating that BearingPoint’s “large clients, whether it be commercial or public service * * * predominantly] * * * choose Oracle, Peo-pleSoft and SAP” software. Tr. at 867:10-14 (Keating). Keating called these large customers “Tier 1” customers, describing their needs with regard to multiple currencies, languages and legal systems. Keat-ing stated that “Oracle, PeopleSoft and SAP are the three clear, you know, players in the marketplace.” Tr. at 870:9-10 (Keating).
Further, Keating testified that no other vendor could deliver the degree of functionality that these three vendors deliver. Tr. at 871:17-20 (Keating). In support of these contentions, plaintiffs introduced a questionnaire that BearingPoint had completed for the European Commission’s investigation of the merger at bar. Ex. P203 at 1. Keating was personally involved in preparing the responses to this questionnaire. In one question, the EC asked BearingPoint: “[Is] there a specific market for supplying EAS * * * to large companies, * * * in which only a few vendors are active?” Id. at 11. BearingPoint responded: ‘Tes, there is such a market. The vendors are SAP, Oracle and People-Soft * * *.” Moreover, BearingPoint’s responses also stated that it believed innovation would be slowed in this market as a *1135 result of the proposed merger between Oracle and PeopleSoft. Id. at 16.
Once the topic turned to the likelihood of entry into this marketplace by vendors other than SAP, Oracle or PeopleSoft, Keating’s testimony began to undermine BearingPoint’s response to the EC. Plaintiffs directed the court’s attention to a portion of the EC questionnaire pertaining to ease of entry. When BearingPoint was asked to “indicate at least three companies that are potentially able to enter this [EAS for large companies] market,” Bearing-Point had listed Microsoft, Siebel and IBM. Id at 14. Moreover, the response stated that the only barrier to entry by these three vendors is “self choice.” Id. But when asked at trial by plaintiffs if Keating was surprised by Microsoft’s approach to acquire SAP, Keating responded: “No, * * * Microsoft’s not a company that plays for second.” Tr. at 926:22-24 (Keating).
On cross-examination, Oracle delved deeper. When asked if “there was any question in [his] mind that Microsoft has the ability to develop a scalable product,” Keating replied “no.” Tr. at 940:13-15 (Keating). The following testimony presents a good summary of Keating’s contribution regarding the potential entry of Microsoft into the high function market:
Question (by Oracle Counsel): They’re [Microsoft] coming aren’t they, to the large market space?
Answer (Keating): Monday they were almost there [referring to the SAP acquisition revelation].
Question: Indeed they were.
Answer: I had a conference call with my SAP practice [saying], “you guys might want to get new letterheads.” I don’t mean to be flip, but it was pretty clear they’re coming.
Tr. at 942:14-19 (Keating).
Furthermore, Keating’s testimony makes it appear that BearingPoint is rolling out the red carpet for Microsoft’s arrival. At trial, an “alliance” between Microsoft and BearingPoint came to light under which BearingPoint has agreed to become Microsoft’s “go to partner” in the high function ERP software market for customers that have less than $2 billion in annual revenues. Ex. D5051 at 2.
In the main, the court found Keating’s testimony to be credible. Most particularly, Keating’s testimony of the alliance between his company and Microsoft substantiates Oracle’s contention that Microsoft is a competitor for much ERP business and able to extend its reach into an arena in which plaintiffs contend that only Oracle, PeopleSoft and SAP now compete. Keat-ing’s testimony gives evidence that Microsoft’s entry into competition may be achieved by a business model different from that followed by Oracle, PeopleSoft or SAP. Microsoft’s ERP products through this collaboration with BearingPoint can be customized, configured and adapted to be competitive with the offerings of the three companies that plaintiffs contend make up the market, at least up to a level that is well within the large, complex level of customer demand that plaintiffs contend requires high function ERP.
Nancy Ellen Thomas, the Global and Americas Financial Management Solutions Leader for IBM, also called by plaintiffs, testified about IBM’s role as a consultant to “large, global complex clients” procuring FMS software. Unlike BearingPoint, IBM has publicly stated its opposition to the hostile takeover of PeopleSoft by Oracle. Ex. D5240R at 13 (stating that a “successful Oracle bid” would be a “negative for IBM * * * [with] possible impact on strong PeopleSoft [and IBM] alliance revenue” and also considering taking a “proactive stance against the [Oracle/Peo-pleSoft] deal”). Thomas began by echoing many of the same views that Keating ex *1136 pressed in regard to the ERP needs of large complex customers, including multiple geographies, currencies, languages and regulatory requirements. Tr. at 474:9-12 (Thomas). When asked, based upon her experience, which ERP vendors could offer a product that could satisfy the requirements of these customers across multiple countries, Thomas listed only Oracle, Peo-pleSoft and SAP. Tr. at 475:2 (Thomas). When asked what vendors could support reporting requirements for multiple ranges of legal entities, Thomas listed only Oracle, PeopleSoft and SAP. Tr. at 476:3 (Thomas). The same three vendors were listed when Thomas was asked about supporting multiple lines of business. Tr. at 476:15 (Thomas). Thomas downplayed the role that Lawson plays within this “up market” sector, stating that “the clients * * * we work with are typically not” focusing on Lawson to the extent that they are focusing on Oracle, PeopleSoft and SAP. Tr. at 495:10-15 (Thomas).
Plaintiffs also appeared to use Thomas to bolster their contention on “localization” between Oracle and PeopleSoft by asking Thomas about the banking industry and which firms compete for that business. When asked which vendors she would expect to see in the final scoring and recommendation phase of a banking customer’s selection process, Thomas stated: “primarily Oracle and PeopleSoft.” Tr. at 498:21-25 (Thomas).
When Oracle’s counsel questioned Thomas about the possible bias of IBM, Tr. at 499-503 (Thomas), Thomas admitted that IBM has the “largest PeopleSoft practice of any consulting firm in the world” and that PeopleSoft has “publicly described IBM as PeopleSoft’s strongest partner.” Tr. at 499 (Thomas). Further, IBM has over 150 employees dedicated to consulting and implementing PeopleSoft products, all of whom could lose their jobs if PeopleSoft was merged with Oracle, a company for which IBM has only 75 dedicated consultants. Tr. at 500:20-502:10 (Thomas).
Turning to Lawson, when asked about IBM’s large and complex implementation of Lawson HRM for the State of Arizona, which has over 60,000 employees, Thomas stated that she didn’t have the “Lawson expertise” to talk about that transaction. Tr. at 519:12-13. Further, Thomas “was not aware” of IBM’s implementation of Lawson software at Montgomery County Schools in Maryland, an entity with over 140,000 students. Nor was she “aware” of IBM’s implementation of Lawson for the State of Michigan or IBM’s implementation of Lawson for a large school district in Tampa. Tr. at 520:7-19 (Thomas).
The court first notes a possible IBM bias due to IBM’s potential loss of People-Soft implementation business, a significant source of IBM revenue. Furthermore, the court cannot overlook Thomas’ lack of knowledge about any potential high function implementation of Lawson software. This makes the court reluctant to afford much, if any, weight to her testimony. Thomas seemed not to be able to identify factors that would keep Lawson from competing in the high function sector. Her testimony failed to substantiate plaintiffs’ claim of separate FMS and HRM high function markets.
Industry Witnesses: PeopleSoft and Microsoft
Next, plaintiffs presented the testimony of several industry witnesses in an effort to support the proposed high function ERP market.
Richard Bergquist, Chief Technology Officer, Senior Vice President and People-Soft “Fellow,” explained to the court how PeopleSoft defines a high function customer versus a mid-market customer. Tr. at 255:18-19, 275-276:21 (Bergquist). Not surprisingly, Bergquist’s definition of high *1137 function customers and high function software echoed plaintiffs’ definitions (or, at least, some of them). First, Bergquist stated that a customer cannot be labeled as high function simply based upon its size or revenue. Rather, one “ha[s] to look all the different dimensions” in order properly to distinguish between these two types of customers. Tr. at 276:3 (Bergquist). The “different dimensions” that Bergquist referred to in guiding an explorer through the task of deciding what label to apply to a customer are: functionality, flexibility, scalability, reliability and technology. Tr. at 280-282, 283:18) 289:4-25 (Bergquist). Only after knowing the customer’s needs regarding all of these dimensions, which one must learn “through a series of conversations with the customer,” can one then place a customer in the correct talis-manic column of high function or mid-market. Tr. at 276:11-13 (Bergquist).
A high function customer requires software that is highly functional, highly flexible, contains large scalability and is reliable 24 hours per day, seven days a week. Tr. at 283-289 (Bergquist). Customers who do not need software with such deep functionality, large scalability or high flexibility are mid-market customers who buy mid-market software. Tr. at 300:10-13 (Bergquist). Bergquist succinctly stated that “customers that don’t have the needs of large and complex enterprises, we [Peo-pleSoft] group into the mid-market.” Tr. at 275:1-2 (Bergquist). Bergquist clearly stated that a market exists for the sale of high function software to high function customers, and in this market PeopleSoft competes with only SAP and Oracle. Tr. at 279:17 (Bergquist). Berquist went on to explain that a customer can be high function regardless of its international locations or international currency needs. Tr. at 292:20 (Bergquist). “Internationality” was not a dimension for delineating high function from mid-market, rather international needs simply create the need for more function and - scalability. Nonetheless, multiple currency, language and nationality capabilities are not requirements for a high function.customer, as a customer can be located in the United States only and use only English and still be a high function customer according to Bergquist. Tr. at 292:1-15 (Bergquist).
Questions soon turned to Lawson and its role in this high' function software market. Berquist stated that PeopleSoft “does not believe” that Lawson sells any HRM or FMS software that has similar functionality to the same software sold by People-Soft. ' Tr. at 299:21-25 (Bergquist). Rather, Lawson has FMS and HRM that is “adequate for the basics of what an organization would need.” Tr. at 300:4-5 (Bergquist). If the organization has simple and repetitive tasks, then “the Lawson product does that very well.” . Tr. at 300:9-10, 304:1-4 (Bergquist). But if a customer starts going beyond those basic tasks,- then the customer needs features and functions that Lawson cannot supply. Tr. at 300:10-13 (Bergquist). Moreover, Lawson does not have to ability to support Unicode, a common character set for all languages of the world. Since Lawson cannot do that, it is “limite[d] to the US, Canada and UK.” Tr. at 301:19-25 (Bergquist).
The next topic was AMS and its role in the high function market. Bergquist stated that'AMS only has a “financial product that is meant for sale in the public sector.” Tr. at 309:14-15 (Bergquist). Further, the software was developed only for a minimum level of functionality and requires extensive customization before it can be implemented. Tr. at 309:12-17 (Bergquist). Further, AMS does not have an HRM product. Because AMS does not rise to the level of functionality required to be considered high function, AMS is not a high function vendor selling a product that *1138 competes in the proposed market. Tr. at 310:4-7 (Bergquist).
Next, Bergquist took aim at the best of breed solutions, stating that a customer “can’t assemble point solutions to get the full picture.” Tr. at 311:12-14 (Bergquist). These point solutions do not provide core functionality, requiring a customer to purchase core functionality from a different vendor, and then having “multiple solutions from point solutions,” creating extensive integration costs. Tr. at 311:12-25 (Bergquist). Accordingly, best of breed solutions are not a viable option for high function customers and therefore are not substitutes for high function software.
Next, Bergquist set out to prove that outsourcing is also not a viable option for high function customers stating that “we see it [outsourcing] as less capable software than that provided by PeopleSoft, SAP and Oracle.” Tr. at 314:11-12 (Bergquist).
Finally, Bergquist was questioned about potential localized competition between Oracle and PeopleSoft, thus establishing the likelihood of unilateral anticompetitive effects. Bergquist testified that there are some instances where Oracle is People-Soft’s closest competitor over SAP. Tr. at 319:6-8 (Bergquist). This type of situation arises in the service industries according to Bergquist because Oracle and People-Soft both “grew up in the same neighborhood,” the services industry neighborhood, thus making Oracle and PeopleSoft strong competitors in this vertical, especially among those who have a “buy-American tendency.” Tr. at 319:11-16 (Bergquist). Moreover, Bergquist testified that SAP has suffered from the “stereotype of German engineering” that leads many to view SAP software as less flexible and requiring more customization. Tr. at 320:11-15 (Bergquist). But Oracle and PeopleSoft are both seen as very flexible, again making them more likely competitors over SAP. Tr. at 320:16-18 (Bergquist).
On cross, Bergquist was first asked about the distinction between mid-market and high function customers and software. When asked if there were any PeopleSoft documents which describe this distinction between high function and mid-market customers or software, Bergquist said that he was not aware of any such documents. Tr. at 347:22-25 (Bergquist). Further, Bergquist admitted that there are no “clear-cut answers” or “firm dividing lines” that distinguish a mid-market customer from a high function customer. Tr. at 353:15-22 (Bergquist).
Next Bergquist was asked about his dismissal of Lawson from the high function market. When asked if PeopleSoft had lost any business from large and complex customers to Lawson, Bergquist replied: “I can’t think of any that we have * * * lost.” Tr. at 364:5 (Bergquist). Oracle then showed Bergquist a document, created by PeopleSoft, tabulating enterprise deals which PeopleSoft had competed for, and the name of the competitor on the deal. Ex. D6236A. The data read that Lawson was an enterprise competitor 27 times, with SAP competing 33 times and Oracle 38 times. Id at PS-C077332. But Bergquist stated: “I don’t know anything about this document * * * where it came from or how it was.” Tr. at 375:3-9 (Bergquist).
Bergquist was then asked about specific instances of competition with Lawson. When asked if he knew anything about PeopleSoft’s loss to Lawson on the Dean Foods account, Bergquist stated “no.” Tr. at 377:16-18 (Bergquist). When asked about PeopleSoft’s loss to Lawson on the Qwest [transcript misspelling] Communications account, again Bergquist stated that he knew nothing about that lost business. Tr. at 377:19-21 (Bergquist). *1139 Moreover, it appears Bergquist was not even aware of instances in which People-Soft won business when in competition ■with Lawson. When asked if he knew anything about PeopleSoft’s wins over Lawson on the Maricopa County account, Bergquist replied “no.” Tr. at 377:22-24 (Bergquist). Bergquist provided the same answer when asked about PeopleSoft’s win against Lawson on the San Diego Unified School District account. Tr. at 378:2 (Bergquist). Bergquist, like Ms Thomas before him, seemed to have been struck with a singular memory lapse. It appears both witness, while able to testify thoroughly about other vendors, drew a complete blank when asked about potential high function implementations of Lawson. The court began to wonder if this phenomenon, perhaps called “Lawson Amnesia,” would strike any more of plaintiffs’ witnesses.
The final part of Bergquist’s cross came when defense counsel began inquiring about the alleged localization of competition between PeopleSoft and Oracle in the services industry vertical:
Question (Oracle counsel): Can you identify for me any particular verticals in which you believe that SAP is not competitive with Oracle and PeopleSoft?
Answer (Bergquist): SAP may compete in almost all the verticals that are there. * * *. There is relative strength for PeopleSoft and Oracle in the services industries.
Question: I understand that you’ve said that sir, but my question is different. In any of those services industries, is it your testimony that SAP is not competitive with Oracle and PeopleSoft?
Answer: No.
Tr. at 388:1-11 (Bergquist) (emphasis added).
Notwithstanding any bias, Bergquist’s testimony served to hurt plaintiffs’ claims rather than bolster them. First, Bergquist conceded that no “clear-cut” dividing line exists in labeling a customer as “high function” rather than “mid-market.” Finding an articulable division between so-called high function and mid-market ERP is necessary to plaintiffs’ burden of establishing a product market. Second, Bergquist conceded that there is not one single services industry vertical in which SAP is not “competitive” with Oracle and PeopleSoft. The court must demarcate such a “node” or area of localized competition between Oracle and PeopleSoft as a prerequisite to finding any likelihood of unilateral anticompetitive effects. Bergquist’s testimony was also full of self-serving statements regarding the low functionality of AMS and Lawson, testimony that was shown to be wholly unreliable on cross-examination when Bergquist was rendered unable to remember key information regarding Lawson.
Philip Wilmington, Executive Vice President of PeopleSoft Americas, further testified in support of the plaintiffs’ proposed product market. Tr. at 1760:4 (Wilmington). Wilmington began by expounding how PeopleSoft characterizes the mid-market versus the “up-market” or high function market. Tr. at 1765-1766 (Wilmington)., Wilmington stated that the “up-market” is defined as customers that have revenues of $1 billion or above and have “complex requirements.” Tr. at 1765:16-22 (Wilmington). Prior to the PeopleSoft acquisition of J D Edwards, the demarcation line between mid-market and up-market had been $500 million. Tr. at 1847:7-17 (Wilmington). Predictably, Wilmington stated that PeopleSoft only competes in the up-market with Oracle and SAP. Tr. . at 1773:14 (Wilmington). Oracle and SAP are the “ones [PeopleSoft] runs into all the time.” Tr. at 1773:19-20 (Wilmington). Only these vendors have “the functionality” and the “references or customer successes” which allow them to *1140 be a competitive presence in the up-market. Tr. at 1773:21-1774:2 (Wilmington). Wilmington further testified that when these three competitors compete, it gets “very aggressive.” Tr. at 1797:20 (Wilmington). Moreover, Wilmington stated that oftentimes PeopleSoft knew the identity of its competitors on any given account, with that information driving higher discounts. Wilmington cited the example of the Oracle and PeopleSoft competition for Target, in which Target would communicate the other competitor’s discount offerings to PeopleSoft. Tr. at 1797:20-25 (Wilmington).
Wilmington testified that he did not believe that the “do nothing” option was a threat to PeopleSoft or other up-market vendors. Wilmington stated: “Almost never do I see a company that is invested in a [procurement] evaluation * * * just do nothing.” Tr. at 1792:2-3 (Wilmington). Wilmington stated that incumbent systems may simply “delay the decision” to buy ERP, but it is not a long-term solution for any customer. Tr. at 1792:21-22 (Wilmington).
Testimony turned to Lawson and its classification as a mid-market or up-market vendor. “Very, very infrequently do I see Lawson,” stated Wilmington in describing the competition for high function customers. Tr. at 1803:9 (Wilmington). “They are not a viable competitor for the up-market.” 1803:10-12 (Wilmington). Wilmington stated that Lawson competes, and competes well, in the mid-market sector, and perhaps it can be seen sporadically in the up-market healthcare and retail industry. Tr. at 1805:13-23 (Wilmington). When Wilmington was asked about the competition between PeopleSoft and Lawson on the Amerigroup account, Wilmington stated that Americgroup “was very much a mid-market opportunity.” Tr. at 1810:5-6 (Wilmington). Regarding Microsoft, Wilmington stated that PeopleSoft does not compete with Microsoft in the upmarket and only sees it from “time to time” in the mid-market. Tr. at 1811:14 (Wilmington).
Next, Wilmington was asked about outsourcing and its role in the up-market. Tr. at 1812:13-14 (Wilmington). Wilmington stated that he does not see outsourcing as a threat to PeopleSoft; rather, he sees outsourcing as an opportunity. Tr. at 1812-17-18 (Wilmington). Wilmington stated that outsourcers have to buy software to manage the client’s HR needs, and PeopleSoft tries to be the vendor to supply such software. Tr. at 1812:17-18 (Wilmington). Accordingly, outsourcing is a business opportunity, not a threat. When asked about outsourcers who use their own software to manage HR, Wilmington stated that he doesn’t feel threatened because that software lacks the “robust functionality that is going to be necessary to successfully meet the needs of [the] up-market.” Tr. at 1813:12-14 (Wilmington).
Finally, Wilmington testified regarding localized competition between Oracle and PeopleSoft. Wilmington stated that SAP software was “developed for a more rigid business model” and therefore lacks flexibility. Tr. at 1815:5-6 (Wilmington). Oracle and PeopleSoft possess such flexibility and therefore are better solutions for upmarket customers. Tr. at 1815:11-15 (Wilmington). Moreover, Wilmington testified that he believes SAP is more expensive, ranging anywhere from “20 to 50 percent, in terms of higher cost of ownership across the board.” Tr. at 1817:5-7 (Wilmington). In fact, Wilmington cited one example, the PNC Bank account, in which Oracle, SAP and PeopleSoft were all three competing. But SAP was eliminated because its software did not possess the flexibility that PNC required. Accordingly, PNC narrowed the competition to only Oracle and PeopleSoft. Finally, Wilmington stated *1141 that the PNC scenario was exemplary of the situation in the entire United States banking industry. Tr. at 1817-1818 (Wilmington).
On cross, Wilmington was questioned about PeopleSoft’s up-market versus mid-market demarcation. It was during this questioning that the court learned that the day prior to Oracle’s tender offer for Peo-pleSoft, the demarcation line was $500 million in revenues and/or 2,000 employees but, soon thereafter, the of demarcation line increased to $1 billion in revenue only. Tr. at 1848:10-16 (Wilmington). Wilmington stated that it was the J D Edwards acquisition, and not the tender offer, which caused the increase. But Oracle’s counsel then asked: “If you drew the line at 500 million and/or 2,000 employees for [the] mid-market [roof amount], then the upmarket would include players other than Oracle, PeopleSoft and SAP, isn’t that right?” Tr. at 1849:5-7 (Wilmington). Wilmington had trouble giving a direct answer to this question, choosing instead to argue that other factors were necessary, other than revenue, before being able to classify a customer in the mid-market or up-market. Tr. at 1849:8-15 (Wilmington). But Oracle’s counsel quickly brought Wilmington’s attention to a 2003 PeopleSoft Pricing Policy Document which stated that “[a] mid-market customer is defined as a customer with revenues of up to $500 million and/or 2,000 employees.” Ex. P4965 at 6-7. There is no mention of other factors such as scalability or functionality needs. Further, the document stated that “[t]he revenue-based metric is meant to be the single determinant of the Mid-Market Line * * Ex. P4965 at 6-7 (emphasis added).
Regarding Lawson, Wilmington stood by his deposition statement that not once in 25 years had Wilmington seen PeopleSoft compete with Lawson for a “enterprise customer.” Tr. at 1856:21-25 (Wilmington). Wilmington stated that he based this statement upon the new $1 billion demarcation line between mid-market and up-market (enterprise) customers. Tr. at 1857:5 (Wilmington). But Wilmington conceded that if the $500 million/2,000 employee line were used, then PeopleSoft had competed with Lawson for enterprise customers. Tr. at 1858:7-8 (Wilmington). Oracle then introduced a document created by PeopleSoft in July 2003, after the Oracle offer and the J D Edwards acquisition, which showed the number of times People-Soft had competed with certain vendors on enterprise deals. Tr. at 1858:10-17 (Wilmington);' Ex. D6236. Since the document was created after the J D Edwards acquisition, it would appear that the mid-market demarcation line used would be (or should have been) the $1 billion line. The document lists PeopleSoft as having competed with Lawson 27 times for an enterprise customer. Tr. at 1859:7-8 (Wilmington). When asked if Wilmington still stood by his testimony, Wilmington stated that he still believed that PeopleSoft “doesn’t see Lawson in enterprise deals.” Tr. at 1859:13-14, 1861:5-7 (Wilmington). Wilmington elected to “stand by his testimony.” Tr. at 1861:8 (Wilmington). Lawson Amnesia appeared to have claimed yet a third victim.
When asked about PeopleSoft’s competition with Lawson for HCA Columbia, Wilmington could not speak to that issue because “he had not been involved in the competition.” Tr. at 1868:4 (Wilmington). When shown a PeopleSoft document that listed Lawson as “the number one competitor in new market deals” in the western geographic region of the United States, Wilmington stated that he “did not know” if Lawson was really number one. Tr. at 1866:13-21 (Wilmington). When competing with Lawson for the Stanford University Medical Center, PeopleSoft documents written by Lynn Duffy, the sales team *1142 leader on the deal, stated that “Lawson is the competition,” but Wilmington stated that he “was not sure” if Duffy was right about that point. Tr. at 1870:5-6 (Wilmington). When asked about PeopleSoft’s loss to Lawson for the State of Michigan account, Wilmington stated that he was “not certain” if PeopleSoft had even competed against Lawson for that account. Tr. at 1878:15-16 (Wilmington). When asked about PeopleSoft’s loss to Lawson for ManuLife’s business, Wilmington stated that he “did not remember losing to them” on that account. Tr. at 1896:14 (Wilmington). When asked about the loss to Lawson for the Mayo Clinic account, Wilmington stated that he “was not familiar with the details of that competition.” Tr. at 1896:22-23 (Wilmington). Since Wilmington apparently was not aware of what PeopleSoft’s own documents reveal about Lawson as a competitor and is “not certain” whether PeopleSoft competed against Lawson for several large accounts, the court finds Wilmington’s testimony concerning Lawson’s absence from the upmarket largely incredible.
Regarding outsourcing, Wilmington was shown the same document created by Peo-pleSoft soon after the Oracle tender offer, which showed that PeopleSoft competed against ADP, an outsourcer, 15 times. Ex. D6236. Wilmington stated that he “did not know how this [sic] data was compiled and edited;” therefore, he could not state whether these data meant that PeopleSoft faced competition from ADP for up-market customers. Tr. at 1860:5-15 (Wilmington).
Finally, Oracle cross-examined Wilmington about any alleged localization between Oracle and PeopleSoft. Wilmington was shown his video deposition in which he was asked:
Question (Oracle counsel): Is there any vertical segment of the market in the United States, the ERP market, where you do not consider SAP to be a formidable competitor for large enterprise customers?
Answer (Wilmington): For large enterprise customers, no. I believe them to be a formidable competitor across the industry.
Tr. at 1957:10-21 (Wilmington) (emphasis added). When asked if he had given those answers, Wilmington replied “yes.” Tr. at 1957:20 (Wilmington).
For the same reasons the court mentioned above in discounting Berquist’s testimony, the court cannot accord much weight to Wilimgton’s testimony. First, Wilmington admitted that there is not a single vertical industry in which Wilmington does not believe SAP to be a “formidable competitor” undercutting plaintiffs’ unilateral effects claim. Further, in describing the way in which PeopleSoft characterizes mid-market customers, Wilmington was impeached by a document created by his own company. Likewise, the same document impeached his testimony about the absence of outsourcers from the upmarket. This impeachment, combined with Lawson Amnesia, leads the court to find that Wilmington did not offer reliable evidence establishing an articulable product market containing only Oracle, People-Soft and SAP.
Douglas Burgum, Senior Vice President of Microsoft Business Solutions (MBS), was another industry witness called by plaintiffs in order to support their theory of the high function product market and its three participants. Burgum began by describing how he literally “bet the farm” on a small software company called Great Plains in 1983. In 2001 Microsoft acquired Great Plains. Tr. at 2974:3-8 (Burgum). In 2002, Microsoft acquired Navison Software, a Danish company. The entire group was rebranded as “Microsoft Business Solutions” (MBS). Tr. at 2973:8-9 (Burgum). MBS sells four business appli *1143 cation products: Navison, Great Plains, Solomon and Axapta. Tr. at 2996:15-16 (Burgum). Burgum is responsible for the overall performance and market strategies of MBS, as well as ongoing developments of new products and enhancements to existing products. Tr. at 2974:18-20 (Bur-gum). Burgum began by stating that MBS is focused on selling its product to mid-market customers. Tr. at 2978:5-8 (Burgum).
To Microsoft, mid-market customers are customers who have employees ranging from 50 to 1000 employees and an average IT spend between $10,000 and $2 million. Ex. P2533R at 6. Further showing that Microsoft is only focused on mid-market customers, Burgum testified that MBS does not have a sales force. Rather, MBS is sold indirectly through reselling partners, companies whose sole purpose is to resell MBS products. Tr. at 2986-2988 (Burgum). Moreover, neither partners nor MBS itself offers implementation or consulting services for the products and do not intend to do so in the future. Tr. at 2995:3-18 (Burgum). When asked if MBS intended to expand its products’ ability to serve the large enterprise sector, Burgum responded “no, * * * that is not a segment we are targeting.” Tr. at 3001:20-3302:1 (Burgum). Moreover, MBS products do not have the functionality to meet large customers’ needs. Tr. at 3005:22-25 (Bur-gum). MBS products, Burg-um stated, cannot handle the “multi-multi-multi issues,” such as multiple languages and currencies that large organizations tend to need. Tr. at 3011:23-25 (Burgum). When asked what firms’ software could meet those needs, Burgum responded: Oracle, PeopleSoft and SAP. Tr. at 3006:8-9 (Bur-gum). Burgum stated that while Microsoft competes with these three from time to time, that competition only occurs for mid-market customers. Tr. at 3008:3-6 (Burgum).
Burgum cited the lost North Dakota account as an example of the limited functionality of the Great Plains product, both pre- and post-acquisition with Microsoft, and its inability to meet large functional needs. Tr. at 3022:3-7 (Burgum). Bur-gum was asked why Microsoft didn’t “just spend a bunch of money” to redevelop the code and the salesforce in order to compete for larger accounts. Tr. at 3024:3-10 (Burgum). Burgum stated that undertaking would “be a formidable task” and would “take more money than I would be willing to recommend that Microsoft spend.” Tr. at 3024:12-18 (Burgum). Plaintiffs asked Burgum about the Microsoft/SAP acquisition proposal. Burgum stated the “leading” reason that Microsoft wanted to acquire SAP was not to enter the high function market for ERP and thereby start competing with Oracle or PeopleSoft. Rather the acquisition was to create “a better value for the customers who would use Microsoft Office to work with and make decisions around the data that would come out of the SAP system.” Tr. at 3040:9-15 (Burgum). Microsoft simply wanted to purchase SAP in order to help “front-end users” be “better able to communicate with back-end data.” Tr. at 3039:25-3040:2 (Burgum). See Ex. P841R at 1. Apparently, the acquisition was not motivated by any ill-will towards Oracle or any desire to enter the market and begin undercutting Oracle. The discussions between Microsoft and SAP were concluded in early spring 2004, about the time this suit went to trial, with a decision not to move forward with the acquisition. Tr. at 3028:9-10 (Burgum).
When asked about Microsoft’s alliance with BearingPoint, Burgum testified that Microsoft had only the humblest of intentions in entering into this alliance. Under this agreement, Microsoft was to “provide funding for hiring, recruiting and training of people who would get skilled up on *1144 Axapta.” Tr. at 3055:15-17 (Burgum). In return, BearingPoint agreed to recommend, install and maintain MBS software, specifically Axapta, to BearingPoint consulting clients. Tr. at 3055:15-17 (Bur-gum); Ex. 3249R at 4, 15. MBS had no plan or expectation for BearingPoint to recommend Axapta software to high function customers. Tr. at 3053:1-6 (Burgum). To the contrary, this agreement was only entered into for BearingPoint to sell MBS to mid-market customers. Tr. at 3054:10-18 (Burgum).
Subsequent to trial, BearingPoint announced that the new Microsoft Business Solutions Axapta was “a compelling ERP solution” which “provides functionality across all key areas of the business * * * including financial management, CRM [and] HR management * * *.” See BearingPoint homepage at http:// www. bearingpoint. com/'solutions/enter-prisesolutions/microsoftJjus-solMml. The BearingPoint webpage claims the “key” functionalities of Microsoft Axapta include “multiple companies, multiple languages, and multiple currencies.” Id. Although these statements do not appear in the trial record, they are consistent with the substantial evidence in the record and afford additional reason to discount Bur-gum’s testimony that MBS is not at least a potential entrant in what plaintiffs characterize as the high function market.
The court accords little weight to Bur-gum’s testimony attempting to prove Microsoft’s absence from the so-called high function ERP product market. Burgum’s Uriah Heep like humility about Microsoft’s intentions regarding the failed SAP alliance and the successful BearingPoint alliance was unconvincing. It strains credulity to believe that Microsoft would offer billions of dollars to acquire SAP merely to make data processing easier for customers who use both Microsoft Office and SAP ERP. Further, this proposition is impeached by Microsoft’s actions with Bear-ingPoint concurrently, or soon after, the SAP alliance was discontinued. Finally, the court wholly discounts Burgum’s testimony that MBS software, especially Axap-ta, lacks the functionality to be considered high function ERP. Burgum stated that MBS products cannot provide the “multi, multi, multi” functionality, but Bearing-Point is selling Axapta on the basis that it can handle “multiple languages, currencies and businesses.” Accordingly, the court discounts Burgum’s testimony portraying MBS solely as a mere humble mid-market vendor.
Finally, in attempting to show the high barriers to entry into the high function market, plaintiffs called Richard Allen, former Executive Vice President of Finance Administration and CFO of J. D. Edwards. Tr. at 747:20-25 (Allen). Allen testified that J. D. Edwards, prior to being acquired by PeopleSoft, had been a company “focused on mid-market customers” that did not need “high levels of configurability,” “deep functionality” or high scalability. Tr. at 746:20-21, 757:25-758:20 (Allen). But Allen testified that in the early 1990s J. D. Edwards attempted to reposition itself in order to sell to “up-market customers.” Tr. at 770:22-771:10 (Allen). J. D. Edwards had to “create a software architecture to allow [its] products to run on multiple hardware platforms, with multiple databases and multiple operating systems.” Tr. at 771:16-21 (Allen). But J. D. Edwards ultimately abandoned this attempted repositioning in 2001. Tr. at 777:6 (Allen). Allen stated that “[J. D. Edwards] came to the conclusion that after about a decade involved in the effort, hundred of millions of dollars of investment, we didn’t have the products, services, and ultimately the reputation necessary to satisfy the requirements that up-market customers have.” Tr. at 777:8-13 (Allen). *1145 Accordingly, plaintiffs argue that no firm could enter the high function FMS and HRM markets within the required two year threshold set by the Guidelines. Pis. Post Brief (Doc. # 366) at 44 (citing Guidelines § 3.3).
Plaintiffs’ Expert: Elzinga
By far the most important of plaintiffs’ witnesses was Professor Kenneth Elzinga of the University of Virginia. Elzinga is a well known and highly regarded economist. Tr. 2142-2145 (Elzinga); Ex. 4014A. The court finds Elzinga to be highly qualified to offer testimony on market definition. Elzinga was the only one of plaintiffs’ witnesses who offered testimony from which the court could attempt to calculate market shares and apply the Philadelphia Nat Bank presumptions or perform the HHI calculations of the Guidelines.
In reaching his proposed market definition, Elzinga purported to follow the Guidelines approach. Tr. at 2163:18-19 (Elzinga). Elzinga concluded that the relevant market was limited to high function FMS and HRM software. Elzinga testified that a hypothetical monopolist could profitably impose a SSNIP in high function FMS and HRM. Tr. at 2149:16-22 (Elzinga). Elzinga posited that if a merged Oracle/PeopleSoft decided to increase the price of its high function FMS and HRM products, consumers would not substitute (1) mid-market solutions (such as those produced by ERP vendor Lawson), (2) best-of-breed solutions (such as those produced by vendor Kronos), (3) incumbent or legacy solutions or (4) the services of outsourcing firms (such as Fidelity and ADP). Tr. at 2178:10, 2179:8-14 (Elzinga).
Elzinga reached his conclusions by analyzing several “strains” of evidence: (1) Oracle discount approval forms; (2) reports from independent research firms; (3) information from high function FMS and HRM customers and consulting firms; and (4) internal documents from firms in the enterprise • software sector. Tr. • at 2168:9-11, 2180:4, 2184:10-12, 2188:23-25 (Elzinga). This being established, Elzinga then presented his conclusions on market shares. Tr. at 2209-2220 (Elzinga).
Discount approval forms. Elzinga’s first strain of evidence, and the one on which he appeared to place the greatest emphasis, was the analysis and tabulation of Oracle discount approval forms (DAF). See Exs. P1000-P1944.
Oracle salespersons have the discretion to offer a 20. to 25 percent discount off the fist price of HRM or FMS. Pr. at 2168:25-2169:2 (Elzinga). If a situation arises in which Oracle is competing with another ERP vendor arid this requires the Oracle salesperson to offer a larger discount on the Oracle ERP software, a DAF must be executed and approved by an Oracle official. Tr. at 2169:3-9 (Elzinga). In executing a' DAF, the salesperson lists the “justification” for pursuing an increased discount. In the justification column of the DAF, “sometimes the particular competitor or alternative [solution] that is justifying * * * or provoking the discount that is being requested” can be found. Tr. at 2169:17-19 (Elzinga).
Elzinga analyzed 222 DAFs that Oracle provided to the DOJ. Elzinga only analyzed the “forms that [1] had U.S. customers, [2] pertained to HRM or FMS software, * * * [3] [had a] net transaction price [of] over $500,000 and [4] where the justification section listed the competitor [or alternative solution] that was driving the request for the discount.” Tr. at 2174:14-18 (Elzinga). These criteria decreased the number of DAFs available for analysis to just over 200. Tr. at 2175:25 (Elzinga). After analyzing all the justifications that were proffered by Oracle salespersons, Elzinga created a graph that showed the number .of times each competitor or alternative solution forced an Oracle *1146 salesperson to request a discount (i.e., the number of times each competitor or alternative'constrained Oracle’s pricing of FMS or HRM). See Ex. P3175.
Based upon the graph, Oracle salespersons cited as primary justification, competition from: Peoplesoft, 122 times; SAP, 81 times; Lawson, 16 times; and Microsoft and AMS, each less than 10 times. Pis Post Brief (Doc. # 866) at 12; Tr. at 2177:10-11 (Elzinga). Elzinga concluded that this discount tabulation is “very powerful, robust evidence” that the relevant product market is high function FMS and HRM. Tr. at 2179:7-8 (Elzinga). “I think that [high function FMS and HRM] is [the relevant market] because I don’t see alternatives outside * * * of that market, such as mid-market, or [incumbent] or outsourcing, disciplining the Oracle pricing the way the [other] two manufacturers of high function FMS software and HRM software do, and that’s SAP and PeopleSoft.” Tr. at 2179:9-14 (Elzinga).
Market research studies. Independent market research organizations study certain product markets and summarize findings about any number of relevant aspects of that market. Most of these research organizations conduct research and issue reports for “people who buy [the product] and want to implement it, * * * but [these firms are] not writing to an antitrust economist or antitrust lawyer audience.” Tr. at 2182:12-16 (Elzinga). Elzinga found one such market research report, conducted by the Gartner Research firm, which analyzed the HRM pillar of the software industry. Tr. at 2181: 17-18 (Elzinga). In the Gartner report, Gartner had enumerated two characteristics upon which it chose to analyze HRM vendors: (1) “completeness of vision” and (2) “ability to execute.” Completeness of vision apparently refers to a vendor’s level of desire to have software capable of either broad and complex transactions (deemed “visionaries”) or limited and ordinary transactions (deemed “niche players”). Tr. at 2182: 17-19 (El-zinga). Ability to execute apparently refers to whether Gartner believed each vendor had the ability to execute its HRM capability “vision” (e.g., high levels of functionality and scalability). Tr. at 2182:24-25 (El-zinga).
According to Elzinga, the Gartner research only identified three firms as “visionaries” with a high ability to execute— Oracle, PeopleSoft and SAP. Tr. at 2183:1- *1147 4 (Elzinga). Elzinga concluded that the Gartner report is again “consistent with the notion that there is something different about high function enterprise software from other alternatives * * * and when it comes to high function software, there is something different about Oracle, PeopleSoft and SAP.” Tr. at 2183: 16-20 (Elzinga) (emphasis added).
Customers and consulting firms. The next strain of evidence Elzinga relied upon were declarations of ERP customers and the “Big Five consulting firms” that plaintiffs furnished him. Tr. at 2184:8-15 (El-zinga). In particular, Elzinga pointed to declarations of Perry Keating of Bearing Point (who also testified) and Deloitte’s [David] Dortenzo. See Elzinga demo # 6. Tr. at 2185:1-2188:9 (Elzinga). In seeking cost-effective solutions and recommendations in choosing ERP vendors, many consumers employ consulting firms to advise them in their negotiations with the vendors. Accenture, IBM Global Services, BearingPoint, Deloitte and CGEY are the consulting firms collectively known as the “Big Five.” At trial, plaintiffs offered the statement of BearingPoint’s Senior Vice President, Perry Keating. Tr. at 912:15-916:7 (Keating). Keating stated that Oracle, SAP and PeopleSoft “are the only [vendors] that provide a product that will be acceptable to a large company in terms of product capabilities * * *.” Id.
Elzinga described the similar testimony of Keating and Dortenzo in declarations as indicating that these Big Five systems integrators most frequently recommend PeopleSoft, Oracle and SAP for ERP implementations. Tr. at 2186: 7-2188:9 (El-zinga).
The customers’ declarations, Elzinga concluded, “were consistent with the hypothesis that there’s a distinction between high function enterprise software and the mid-market * * *. [Mid-market solutions] are not substitutes that a hypothetical monopolist] * * * would be constrained [by] in its pricing discretion [of high function FMS and HRM].” Tr. at 2184:17-22 (Elzinga).
Internal documents from ERP vendors. Elzinga was also privy to internal company documents, some of which he claimed were informative. Tr. at 2189:23-25, 2190:15 (Elzinga). First, Elzinga was privy to customer surveys that had been completed by Oracle ERP customers. These surveys had been given to Oracle customers who were classified by Oracle as having over $2 billion in sales. Tr. at 2189:5-7 (Elzinga). These 28 surveys asked the Oracle customer to identify any other “vendors [other than Oracle] that were considered.” Id. at 3-4 (Elzinga). Elzinga summarized the surveys and concluded that PeopleSoft was considered 50 percent of the time by Oracle customers. Ex. P3176. SAP was considered 28 percent of the time and Lawson was considered 18 percent of the time. Id. Microsoft was considered only 4 percent of the time. Id.
Also of interest to Elzinga was an internal document produced by Microsoft in response to the government’s civil investigative demand (CID), MS-OPCID 1610. The document was labeled “Microsoft Business Solutions: Scorecard Review.” See Elzinga demo # 8. In the document, Microsoft is characterized as worried about “Oracle, Peoplesoft, [and] SAP aggressively moving down-market, increasing pricing pressure (discounting levels) and creating new channel programs.” Tr. at 2192:8-11 (Elzinga). Elzinga concluded that this document showed that Microsoft (1) recognizes a difference between mid-market and high function software and (2) does not consider itself to be in the market for high function ERP. Tr. at 2192:13-20 (Elzinga).
From the foregoing, Elzinga crafted a metric to measure the product market.
*1148 Elzinga’s data were calculated exclusively for use in this trial. In estimating the product market from the non-public sales data of Oracle and PeopleSoft and third party vendors obtained through the government’s compulsory process, Pis Fact (Doc. #356) 6.2, Elzinga applied a minimum threshold purchase “of $500,000 per customer” to identify high function FMS and HRM. Tr. at 2210:2-4 (Elzinga). El-zinga used this threshold amount to filter out mid-market and point solution sales. Tr. at 2210:4-6 (Elzinga). Accordingly, any sale of FMS or HRM that resulted in at least $500,000 in net license revenues to the vendor, Elzinga considered to be a sale of high function FMS or HRM and thus was in the relevant market. Because plaintiffs’ product market definition has no widely accepted meaning in the industry, there were no generally available data explicating the proposed market’s participants and their relevant shares to backup Elzinga’s estimates.
From his numbers, Elzinga calculated the following United States high function FMS market shares: SAP, 39 percent; PeopleSoft, 31 percent; and Oracle, 17 percent. Tr. at 2212:22-24 (Elzinga). A merged Oracle/PeopleSoft would, in Elzin-ga’s view, possess a 48 percent market share. Tr. at 2212:24-25 (Elzinga). Using the same data, Elzinga calculated the HHIi in the high function FMS market to be 2800. Tr. at 2214:17-18 (Elzinga). Based upon Elzinga’s calculations, a merger between Oracle and PeopleSoft would increase the high function FMS HHI2 to 3800. Tr. at 2214:20-21 (Elzinga).
For high function HRM, Elzinga calculated PeopleSoft’s market share at 50 pe-eent, SAP at 30 percent and Oracle at 18 percent; hence, in Elzinga’s view, a merged Oracle/PeopleSoft would have a market share approaching 70 percent. Tr. at 2218:18-23 (Elzinga). Elzinga calculated an HHIj of 2800 in the high function HRM market. Tr. at 2219:7-9 (Elzinga). Post-merger, the HHI2 would increase to 5700. Tr. at 2219:10-11 (Elzinga).
Plainly, the levels of concentration reflected in Elzinga’s testimony exceed the thresholds for “significant competitive concerns” under the Guidelines. Guidelines § 1.51(c). Both HHI2 amounts exceed 1800, and both □HHI amounts exceed 50 points. Likewise, of course, post-merger market shares of this magnitude would satisfy the conditions to raise the anticom-petitive presumption described by the Supreme Court in Philadelphia Nat Bank.
But for reasons explained more fully following the discussion of Oracle’s expert witnesses, the court finds that Elzinga failed to carry the plaintiffs’ burden of (1) establishing an articulable product market and (2) providing post-merger market share and HHI measurements, in a properly defined market, invoking an anticom-petitive presumption under Philadelphia Nat Bank or the Guidelines.
Oracle’s Critique of Plaintiffs’ Product Market Definition
Oracle painted a quite different picture. Oracle assailed plaintiffs’ high function software “label,” arguing that there is “not a sufficient break in the chain of FMS and HRM substitutes to warrant calling ‘high-function’ software — meaning SAP, Oracle and PeopleSoft [FMS and HRM] products — a market unto themselves.” Def. Post Brief (Doc. # 365) at 17. Oracle argued that the relevant product market is, at least, the entire continuum of FMS and HRM software, including those sold by so-called mid-market vendors. Id. In support of this position, Oracle presented several witnesses.
Systems Integrator Witness
Oracle called Christy Bass, Global Managing Partner of Global Business Solutions for Accenture, to rebut plaintiffs’ product *1149 market definition as well as rebut the notion of localized competition between Oracle and PeopleSoft. Accenture is the largest systems integrator in the world, with annual revenue exceeding $11.4 billion. Tr. at 1610:15 (Bass). Bass testified that “all” of Accenture’s clients have high function needs. Tr. at 1613:6-7 (Bass). Bass testified that several high function clients, such as Best Buy and BellSouth, had chosen to outsource their entire HR function. Tr. at 1648:14-19 (Bass). While some of these outsourcing clients were on a “one-to-one” outsourcing model, in which it took a license directly from an ERP vendor, such as Oracle, Bass stated that Accenture is planning to launch the “one-to-many” model. Tr. at 1649:14-1650:13 (Bass). Under this model, the license will be between Accenture and the ERP vendor, with no contractual arrangement between the customer and the vendor. Tr. at 1650:3-13 (Bass). Moreover, Bass testified Accenture plans to begin outsourcing FMS on a “one-to-many” model within the next two years. Tr. at 1655:6 (Bass).
Bass also testified about best of breed solutions and their potential to constrain high function ERP prices. Bass stated that it was “extremely common” for high function clients to pursue a best of breed approach. Tr. at 1668:17 (Bass). Bass stated that these best of breed solutions could possibly offer greater functionality than Oracle, SAP or PeopleSoft. Tr. at 1668:24-1669:3 (Bass). She also stated that best of breed solutions put competitive pressure on these ERP vendors. Tr. at 1669:19-22 (Bass).
Bass rebutted plaintiffs’ assertions that SAP was a “struggling” firm and also plaintiffs’ evidence regarding localized competition between Oracle and People-Soft. Bass characterized SAP’s position in the United States ERP marketplace as “strong.” Further, she testified that she considered SAP to have a “stronger” position than either Oracle or PeopleSoft in regards to Global 2000 clients. Tr. at 1621:18-23 (Bass). When asked about SAP’s complete exclusion from the United States banking industry, Bass conceded that such a situation existed, but opined that change was on the horizon. Bass disclosed that SAP and Accenture have entered into a “strategic alliance” to co-develop a banking solution for European and United States banking firms. Tr. at 1633:4-7 (Bass); Ex. D5001. Bass stated that Accenture has relationships with all twenty of the largest United States banks, and Accenture “leveraged] the experience that [Accenture] has had in the banking industry” in order to get some of these banks to discuss implementing the co-developed software. Tr. at 1634:15-16, 1635:3-10,1636:1-6 (Bass).
The court finds Bass’ testimony to be reliable and informative on the issues of outsourcing and localized competition. Regarding high function clients that have chosen outsourcing as an ERP alternative, Bass gave specific examples of companies, both of which would seem to meet plaintiffs’ high function definition, that had chosen to outsource their entire HRM needs. Bass’ testimony of a lack of localized competition between Oracle and PeopleSoft was likewise supported by her explanation of the SAP/Accenture co-development alliance, under which Bass explicitly stated that Accenture will use its leverage and experience with United States banking firms in order to help SAP gain a larger competitive share in that vertical.
Industry Witnesses: Lawson and SAP
Jay Coughlan, CEO and President of Lawson Software testified regarding his view of the plaintiffs’ proposed product market and its relation to Lawson. Tr. at 3586:1-13 (Coughlan). Oracle wasted no time in questioning Coughlan about plaintiffs’ characterization of Lawson as only a *1150 mid-market vendor. Tr. at 3596:5-9 (Coughlan). Coughlan stated that he disagreed with this view, testifying that Lawson has customers that exceed $1 billion in revenues, employ more than 10,000 people and are listed among the Fortune 1000. Tr. at 3596:6-19 (Coughlan). Coughlan stated that the plaintiffs’ perception of Lawson may have been appropriate before 1996, when Lawson made a conscious decision to focus on specific verticals and winning larger shares in those verticals. Tr. at 3597:1-8 (Coughlan). The first vertical that Lawson focused on was healthcare and today it is providing procurement and HRM for HCA, the largest health care provider in the world with annual revenues exceeding $20 billion. Hr. at 3600:1-4 (Coughlan). Coughlan also stated that Lawson provides FMS and procurement to the Mayo Clinic, an account for which Lawson beat both Oracle and PeopleSoft. Tr. at 3601:2-6 (Coughlan).
Coughlan stated that Lawson next focused upon the retail vertical and has met with much success. Today, Lawson provides FMS to Safeway, the third largest grocery chain in the United States with approximate revenues of $30 billion. Tr. at 3604:1-8 (Coughlan). Lawson provides FMS to Walgreens, a convenience store chain with more than $30 billion in revenues. Tr. at 3604:12-21 (Coughlan). Lawson provides FMS for Target, a department store chain with more than 300,-000 employees and $50 billion in revenues. Tr. at 3605:2-13 (Coughlan). The same is true for Williams-Sonoma. Tr. at 3606:16-19 (Coughlan). In the apparel area, Lawson provides HRM and FMS to Ralph Lauren and Gucci. Tr. at 3605:19-25 (Coughlan). Lawson provides HRM to McDonald’s, a food retailer with over 100,- ' 000 employees. Tr. at 3607:4-19 (Cough-lan).
In the public sector vertical, Lawson has won major accounts with school districts in Florida, Virginia and Maryland, all of which Lawson competed for, and won, against Oracle and PeopleSoft. Lawson provides- HRM to the States of Michigan and Arizona. Tr. at 3615:4-15 (Coughlan). Lawson provides HRM for the City of Dallas and the University of Wisconsin. Tr. at 3613:4-12' (Coughlan). Coughlan’s testimony continued to describe Lawson accounts in insurance and financial services verticals as well as individual customers including Johnson & Johnson (HRM) and Sara Lee and McGraw-Hill (HRM and FMS). Tr. at 3636-3640 (Coughlan). See also Ex. D7140.
Moreover, Coughlan testified that Lawson software systems run in English, French and Spanish. Tr. at 3645:13-17 (Coughlan). Coughlan testified that Lawson software is able to handle multiple currencies as well, citing one Lawson customer, Schlumberger, a major supplier to the oil industry with $10 billion in revenues, 10,000 employees and international operations. Tr. at 3641:23-3642:11 (Coughlan). Schlumberger is utilizing Lawson FMS in close to 100 countries, but not the United States, thus showing that the Lawson software can handle currencies beyond the United States dollar. Tr. at 3642:16-3643:9 (Coughlan).
Finally, Oracle asked Coughlan about Professor Elzinga’s data expounding the market shares for high function HRM and FMS. Tr. at 3648:16-3655:19 (Coughlan). Oracle offered into evidence the DOJ subpoena to which Lawson had responded by telling the DOJ of a large number of HRM and FMS shipments that had been made in late 2002 and throughout 2003. Ex. D7079R. This list included FMS sales to Dollar Tree Store, Louisiana Pacific Corporation and ManuLife, with each sale totaling more than $500,000. Ex. D7079R; Tr. at 3650:3-8 (Coughlan). Moreover, FMS suites were sent to Schlumberger, *1151 Sara Lee and Johnson & Johnson, with each spending more than $1 million on the suites. Id. Accordingly, Coughlan stated that he was perplexed when told that according to Professor Elzinga’s statistics, Lawson had no market share of the high function FMS market because Lawson had made no sales of FMS over $500,000. Tr. at 3653:10-13 (Coughlan). Further, Coughlan was told that Elzinga’s HRM data listed Lawson as having made only one sale above $500,000, a sale for $995,000, leading Elzinga to call Lawson a “fringe player” in the HRM high function market. Tr. at 2219:16 (Elzinga). In response, Coughlan stated that he disagreed with Elzinga’s calculations, citing that Lawson “had one deal alone in HRMS in [2003] that was more than one million dollars.” Tr. at 3654:1-2 (Coughlan). Coughlan stated that he disagreed with plaintiffs’ attempts to characterize Lawson as “not a serious player [in the high function market].” Tr. at 3655:15-19 (Cough-lan).
On cross-examination, the plaintiffs were able to delve more deeply into the customer relationships that Lawson has with several of the customers discussed on direct. The City of Dallas had extensive problems with Lawson’s HRM software, Coughlan admitted, and its ability to handle overtime payroll functionality, leading Dallas to withhold payments to Lawson. Tr. at 3699:15-19 (Coughlan). Coughlan claimed the problem had been corrected. Tr. at 3700:11 (Coughlan). Next, an internal Lawson memo showed that McGraw-Hill was exploring the option of replacing Lawson, as was another client, Cendant. Ex. P3297. Moreover, the document stated that Johnson & Johnson was “not purchasing much in the way of additional applications.” Id. In summation, the document seemed to call into question Lawson’s ability to meet the HR needs of global organizations. Id. Coughlan conceded that the Mayo Clinic has had problems with its Lawson FMS software. Tr. at 3715:10-23 (Coughlan). Plaintiffs went through a series of Lawson customers that have had some implementation or service problem with Lawson software. Tr. at 3699-3711 (Coughlan).
This evidence was elicited in an attempt to show that Lawson is not a player in the high function ERP market. The evidence did show the existence of implementation or service problems. But the customers all appeared to fit plaintiffs’ definition of high function customers. Hence, this line of inquiry did not appear to demonstrate Lawson’s absence from this or any such market, only that some Lawson customers have had problems with its software. The court, therefore, discounts Coughlan’s cross-examination testimony for the purpose for which it was apparently offered. Plaintiffs did not show that implementation or service problems were absent or less frequent in Oracle, PeopleSoft or SAP products. Accordingly, the court credits Coughlan’s testimony regarding large and complex customers that have chosen Lawson ERP to meet their FMS and HRM needs. Not only was this evidence uncon-tradicted, but the testimony was amply supported by many exhibits.
Richard Knowles, Vice President of Operations for SAP America, was called by Oracle to refute the plaintiffs’ product market definition as well as to poke holes in plaintiffs’ theory of unilateral anticom-petitive effects. Tr. at 2805:4-9 (Knowles). At the outset, Knowles clarified some of the terms used in this case, or at least as those terms are understood by SAP. “High function” has no meaning apparently. SAP looks to customer characteristics in determining whether a vendor is mid-market or high function. SAP considers a customer to be mid-market if it has revenues less than $1.5 billion, but more than $200 million. Tr. at 2818:9-19 (Knowles). *1152 A customer above $1.5 billion is considered a “large enterprise.” Tr. at 2819:14-15 (Knowles). But Knowles stated that characterizing a customer as one or the other was far from “an exact science.” Tr. at 2820:18-19 (Knowles).
Oracle then proceeded to “name drop” a large number of SAP clients: Deloitte & Touche, Accenture, Halliburton, MCI, SBC, T-Mobile, AOL, Starbucks, Nike, Home Depot and Barnes & Noble, all clearly up-market customers. Tr. at 2829:6-2831:19 (Knowles). This evidence tended to rebut the suggestion that SAP was a struggling firm with substantial disadvantages in the United States. Next, Oracle questioned Knowles regarding two specific examples in which SAP had competed head to head against Oracle and other ERP vendors. First, Oracle presented an SAP DAF regarding a proposed ERP license transaction with ExpressJet, a company with approximately $1.5 billion in revenues, thus making it a large enterprise. D5641R at 1; Tr. at 2839:23 (Knowles). As with the DAFs used by Oracle, the SAP DAEs had a column for denoting the competitor that was requiring or motivating the increased discount request. D5641R at 1. In the case with ExpressJet, SAP was originally competing against PeopleSoft, Lawson, Exact, Microsoft, Oracle and Ultimate. Ex. D5641R at 2. Knowles stated that he recognized the name Lawson and that SAP “of course” competes with Lawson. Tr. at 2841:8-12 (Knowles). Moreover, once ExpressJet had narrowed the six vendors down to three, it was a contest with Lawson, Oracle and SAP. Tr. at 2842:23-2843:5 (Knowles).
Knowles stated that SAP was “agnostic” about which competitor makes it to the final round, because SAP is going to give the same level of discount regardless of the competitor. Tr. at 2848:7-10 (Knowles). Oracle then introduced another SAP DAF, this time for Kellogg, Brown & Root, a subsidiary of Halliburton. Ex. D5649R at 1. This form listed the justification for the discount as the “extreme competition” between Oracle and SAP. Id. Knowles stated that this type of scenario was to be expected, as SAP views Oracle to be “highly aggressive” on pricing. Tr. at 2856:10-11 (Knowles).
Next, Oracle introduced an email from Bill McDonald, the CEO of SAP America. Ex. D5636. The email contained Microsoft’s second quarter earnings for 2004. Id. at 1. The document began by reading: “These guys are here!”. Id. Knowles stated that McDermott was referring to Microsoft’s 32 percent year-over-year increase in the EAS market. Tr. at 2892:4-23 (Knowles).
Finally, Oracle questioned Knowles about any apprehensions SAP felt regarding increased prices should the proposed merger of Oracle and PeopleSoft be consummated. Tr. at 2858:9-11 (Knowles). Knowles responded that SAP has a neutral opinion on the merger. Knowles stated his belief that the merger will actually make the ERP market more competitive. Tr. at 2858:20-21 (Knowles).
On cross, Knowles conceded that the reason SAP America exists is because customers in the United States “want to have somebody here present to deal with in buying the type of software that [SAP] sells.” Tr. at 2902:12-15 (Knowles). Next, Knowles stated that SAP views Lawson as a “mid-market company.” Tr. at 2924:24 (Knowles). This characterization appears to rest on SAP’s labeling as mid-market of customers with less than $1.5 billion in revenues, a substantially different demarcation from plaintiffs’ labeling of a mid-market customer as one that does not buy software packages exceeding $500,000. See Tr. at 2924:5-7 (Knowles). Accordingly, the court accords no weight to Knowles’ statement inasmuch as it was offered to *1153 show that Lawson does not compete in the high function market. Otherwise, the court finds Knowles’ testimony to be reliable and uncontradicted.
Outsourcing Witnesses
Michael Sternklar, Executive Vice President of Fidelity Human Resources Services Company, testified regarding Fidelity’s outsourcing solutions for HR needs. Tr. at 3124-3126 (Sternklar). Sternklar stated that Fidelity currently has a license with Oracle for HRM software. Tr. at 3130:2-7, 3135:3-8 (Sternklar). Sternklar listed several of Fidelity’s “large” customers: Bank of America, IBM, American Corporation and Asea Brown Boveri (ABB). Tr. at 3136:13-3137:8 (Sternklar). Sternklar described the procurement process by which Fidelity won the ABB account. ABB made its choice between buying an in-house system from PeopleSoft or SAP, or instead, outsourcing ABB’s HRM needs through Fidelity. Tr. at 3138:14-25, 3139:10-13 (Sternklar). ABB chose Fidelity over PeopleSoft and SAP. Tr. at 3140:17-18 (Sternklar). Sternklar stated several reasons why a customer would choose outsourcing over an in-house ERP system. One important reason, Sternklar stated, was the “continued investment” involved in buying an in-house ERP system based upon the need continuously to upgrade such a system. Tr. at 3139:23-25 (Sternklar).
Jay Rising, President of National Accounts at ADP also testified about what he called “upgrade treadmill” and “hidden costs” that are involved in package software. Tr. at 4094:15-22 (Rising). Both Sternklar and Rising testified that there are no such costs associated with outsourcing because the outsourcer itself, not the customer, handles all upgrades and maintenance. Tr. at 3140:2-5 (Sternklar); 4093: 12-16 (Rising). The client need not bother with such hassles.
Both witnesses also testified that outsourcing companies are able to handle the HR needs of companies with large numbers of employees. ADP has 1000 customers that have over 1000 employees. Tr. at 4097:21-25 (Rising). ADP’s client list includes Comcast, Sysco, Xerox and Tyco. Tr. at 4100:13-24 (Rising). Fidelity outsources for Bank of America which currently has between 170,000 and 180,000 employees. Tr. at 3145:18-25 (Sternklar).
Finally, Sternklar stated that Fidelity was currently in the process of creating its own software called HR Access. Tr. at 3152:3-3153:13 (Sternklar). Fidelity’s goal is to move all customers onto HR Access within the next two years and cease using Oracle software completely. Tr. at 3154:9-15 (Sternklar).
The evidence of both of these outsourcing witnesses was reliable and amply supported by specific examples of high function customers that had chosen to outsource with Fidelity or ADP as an ERP alternative. Accordingly, the court credits this testimony in determining whether outsourcing solutions have a price-constraining effect on ERP vendors.
Expert Witnesses: Hausman and Campbell
Oracle did not propose a product market definition. Instead, Oracle picked apart plaintiffs’ market definition piece by piece. Two expert witnesses, Professor Jerry Hausman, an industrial organization economist at MIT, and Tom Campbell, dean of the Haas Graduate School of Business at the University of California (Berkeley) testified for Oracle. Among other important positions in government, Campbell served as director, Bureau of Competition, at the FTC. Both Hausman and Campbell assailed plaintiffs’ product market definition, describing it as vague, unrealistic and un-derinclusive. As with Elzinga, the court finds both Hausman and Campbell to be *1154 well qualified to offer their opinion testimony.
Vague. Hausman characterized the “high function” label as vague and too “hard to get your arms around.” Tr. at 3807:14-15 (Hausman). He cited plaintiffs’ changing description of “high function” ERP as illustrating the unreality of plaintiffs’ proposed product market definition. Tr. at 3809:20-3810:9 (Hausman). At first, plaintiffs argued for a customer-based product definition. Campbell characterized this initial customer-based market definition as “unprecedented” and “unusual.” Tr. at 2704:6 (Campbell). Hausman asserted that plaintiffs, in reaching this strange product market, clearly worked backwards from their desired result: finding a group of customers all of which had purchased SAP, Oracle or PeopleSoft ERP, then claiming that those customers were “similarly situated” and defined the market. But, at trial, Haus-man noted, plaintiffs shifted ground and argued that the high function market was based upon “product characteristics” of the software, such as functionality and scalability, not the customers who buy it. Tr. at 3809:20-3810:3 (Hausman). See also Pis Post Brief (Doc. #366) at 3-4.
Even accepting the plaintiffs’ second version of high function software, both experts asserted that the term is too imprecise to define a market. Hausman contended that Elzinga himself admitted that the high function definition contained no “quantitative metrics” that could be used to distinguish a vendor of high function ERP from a vendor of mid-market software. Tr. at 3807:16-17 (Hausman). See Tr. 2151:18-2152:3 (Elzinga). Hausman illustrated his point by reference to Manu-Life Insurance Company, the fifth largest insurance company in the United States with offices throughout North America. ManuLife has complex needs and transactions and thus by any objective measure would fit in plaintiffs’ high function market. But plaintiffs, for a reason Hausman said plaintiffs left unexplained, considered ManuLife to be a mid-market purchaser and therefore excluded from the plaintiffs’ market definition. Tr. at 3840:17-3841:13 (Hausman). The same applies to Johnson & Johnson and Safeway, both considered by plaintiffs as mid-market customers because they bought ERP solutions from vendors that Elzinga and plaintiffs put in the mid-market. But plainly these firms fit plaintiffs’ description of enterprises having high functional needs. So, concluded Hausman, plaintiffs have provided no objective way to distinguish ERP licenses in the high function market from those in the mid-market.
Both Hausman and Campbell made the obvious point that if the market is not precisely defined, then the market participants and their relative shares will be “economically inaccurate.” Tr. at 2702:16-19 (Campbell); 3793:9-11 (Hausman). Referring to plaintiffs’ customer witnesses, Hausman asserted that surveys that ask customers what their preferences are or what their hypothetical actions “would be” are known to be unreliable and subjective. Id.
Oracle summarized Hausman’s vagueness argument by claiming “there must be a clear break in the chain of substitutes in order for separate markets to be found.” Def Post Brief (Doc. # 365) at 17. According to Oracle, “[Tjhere is clearly not a sufficient break in the chain of FMS and HRM substitutes to warrant calling * * * software [sold by] Oracle, PeopleSoft and SAP, a market unto themselves.” Id. If such a clear break exists, plaintiffs have not proven it by a preponderance of the evidence, Oracle argued in closing. Id.
Disconnected. Oracle also argued that plaintiffs’ product market definition “does not address the market reality” of the way *1155 software is sold, a point to which Hausman testified. Def. Post Brief (Doc. # 365) at 1. Hausman posited that FMS and HRM are not products in and of themselves. Rather, “90 percent” of companies “are buying more than just FMS, more than just HRM. * * * [they are] buying bundles of software.” Tr. at 3815:10-12, 3813:12-22 (Hausman).
Hausman gave as an example a consumer purchasing a single package of software from PeopleSoft that included FMS, HRM, EPM and CRM pillars. In such a bundle, PeopleSoft would not offer discounts based on the individual pillars. Rather, People-Soft would give a “blended discount” across all products in the bundle in order to ensure that the consumer buys all the pillars from PeopleSoft. Tr. at 3814:3-22 (Hausman). If the vendor does not offer an acceptable discount, then the consumer can threaten to buy one of the pillars, such as CRM, from a best of breed vendor such as Siebel. Tr. at 3815:1-6 (Hausman). Based upon this analysis, Hausman opined that the presence of best of breed vendors constrains the prices that the ERP vendors can charge for a bundle of software. Tr. at 3814:18-22 (Hausman).
Underinchisive. Finally, Oracle’s witnesses stated that even if one assumes that a “high function HRM and FMS” market does exist and the market can be demarcated from other solutions, there are viable substitutes for high function ERP that must be included in the product market. Specifically, Oracle argued that (1) mid-market vendors, (2) outsourcing, (3) incumbent systems, and (4) best of breed solutions, discussed above, must all be included in the product market, as all are potential substitutes constraining a post-merger SSNIP.
Incumbent systems, also called legacy systems, refer to the FMS and HRM software systems that the DOJ’s “enterprise customers” already have in operation. These are the systems that the new software from Oracle or PeopleSoft or SAP will replace, should a consumer choose to purchase an integrated suite from one of the high function vendors. Oracle argued that if a post-merger Oracle/PeopleSoft imposed a SSNIP, consumers could constrain that SSNIP by simply refusing to buy high function FMS and HRM and choosing to use already existing products. Tr. at 3821:1-9 (Hausman). Hausman stated that the cost of maintaining and upgrading incumbent systems has been decreasing recently so that these systems have become a “credible threat” to ERP vendors. Tr. at 3821:13 (Hausman). Accordingly, if a customer finds a post-merger price offer too high, it can almost always credibly claim it will not buy the product and instead continue to operate its incumbent system. Tr. at 3821:13-14 (Hausman).
Campbell stated that “20 to 30 percent of the time, even after negotiations have started, the purchaser will opt to drop out” and remain with the system it already has. Tr. at 2708:23-25 (Campbell). Campbell claimed that this factor must be taken into account when calculating market shares, otherwise “you’ve made a very serious mistake in calculating your market shares,” because 20 to 30 percent of the relevant customers’ actual behavior is being ignored. Tr. at 2709:1-6 (Campbell).
Regarding outsourcing, Hausman presented evidence of over twenty large enterprises, such as Bank of America and A T & T, who currently outsource all or some of their HRM needs. Tr. at 3825:19-25 (Hausman). And this phenomenon was occurring long before Oracle made its take-over offer to PeopleSoft. These large enterprise customers would not be outsourcing if they did not find this option to be equal to or better than the purchase of high function software from a vendor. Tr. *1156 at 3828:19-23 (Hausman). If this many corporations can currently have their HRM needs effectively met by outsourcing, it only follows that many more customers could follow suit should a post-merger SSNIP occur in the high function market. Tr. at 3829:1-3 (Hausman).
Hausman gave the example of MIT, his employer, outsourcing its HRM to Fidelity, who he claims do “a heck of a lot better” than MIT personnel. Tr. at 3825:4 (Hausman). Hausman presented evidence that many companies have chosen outsourcing; these include: Bank of America, Motorola, International Paper, McKesson, American Express and Sony. Tr. at 3829:21-23 (Hausman). These are “sophisticated” companies, with a lot of complex transactions, and they have clearly found outsourcing a satisfactory alternative. Id. Hausman’s demonstratives alone listed seven outsourcing firms capable of handling the HR for large companies; these include Fidelity, Accenture, ACS, Exult and Mellon, among others. Haus-man demo # 10.
Accordingly, both Campbell and Haus-man asserted that any product market must include outsourcing solutions as a viable substitute to which consumers can turn in the event that a merged Oracle/PeopleSoft imposes a SSNIP.
Finally, Oracle attempted to show that the products of so-called mid-market vendors, such as Lawson and AMS are reasonably interchangeable for those of the alleged high function vendors, Oracle, Peo-pleSoft and SAP. Accordingly, Hausman stated that any market definition that is devoid of these vendors is too narrow. Tr. at 3939-3940 (Hausman). Hausman presented evidence of over thirty consumers, all of which have large and complex needs, and all of which had chosen to use Lawson or AMS for their FMS and HRM needs. Lawson’s customers include: Johnson & Johnson, Walgreens, Target, Williams-So-noma, Jack in the Box, the Federal Reserve Bank and Safeway. Hausman demo # 11. AMS’ customers include: United States Environmental Protection Agency, United States Postal Service, Library of Congress, Internal Revenue Service and the DOJ. Id. Very telling to Hausman was the fact that the DOJ, two weeks after bringing this case, chose to buy AMS FMS for $24 million, ranking AMS better than Oracle or PeopleSoft in the DOJ’s view for the DOJ’s needs. Tr. at 3842:7-13 (Haus-man).
Hausman admitted that these vendors are “not PeopleSoft,” nor do they “aspire to be.” Tr. at 3839:4-6 (Hausman). He also admitted that these three “cannot currently satisfy the entire market as defined by the DOJ.” But “you do not have to beat PeopleSoft to constrain it” argued Haus-man. Tr. at 3839:20-21 (Hausman). The question is not whether the entire market would switch to these other vendors in the event of a SSNIP, the question is whether enough consumers could potentially turn to a product to meet their needs, thereby making a SSNIP unprofitable. Clearly, if the high function needs of Johnson & Johnson and the DOJ are met by these mid-market vendors, then many other companies could also do so in the wake of a SSNIP. Accordingly, these two mid-market vendors should be included in the product market.
Infrastructure layer. Two of defendant’s expert witnesses discussed the infrastructure layer and its impact on the product markets. Tr. at 4138-4145 (Kut-nik); Tr. at 4364-4369, 4397-4398 (Teece). Traditionally, ERP software contained both business logic and applications services. Business logic is the logical structure of the business process being automated. Applications services are tools that support business logic across different business applications. Applications ser *1157 vices include directory services, security features and content management tools. Web services are a type of applications service.
Recent innovations in software technology have led to a “decoupling” of business logic from applications services. These innovations have resulted in the creation of an “infrastructure layer” that standardizes many of the applications services that were once incorporated with the business logic in an EAS program. The infrastructure layer has also been referred to as the “integration layer,” the “applications services” layer and the “composite applications” layer. Tr. at 325-31 (Bergquist).
Infrastructure layer products and ERP software share some degree of substituta-bility in that both address integration. Developments in infrastructure layer technology allow greater interoperability and easier horizontal integration. Ex. D7143 (Mills 5/27/04 Dep) at Tr. 59-61; Tr. 2886-89 (Knowles); Tr. 4150:9-19 (Kutnick). Similarly, pre-integration in ERP software suites allows greater interoperability and easier horizontal integration. Because one can choose more robust infrastructure layer products instead of pre-integration, the infrastructure layer is a partial substitute for the pre-integration in ERP software suites.
Oracle’s experts Kutnick and Teece testified that the emergence of the infrastructure layer constitutes a paradigm shift in ERP software products and affects the proper product market definition.
The following facts suggest that infrastructure layer products should not be included in the same relevant market as ERP software. First, the overlap in sub-stitutability between infrastructure layer products and ERP software is limited. ERP software products perform a large number of functions that are not performed by infrastructure layer products, and vice versa. Accordingly, sellers of infrastructure layer products likely could not constrain market power of a hypothetical monopoly over ERP software.
Second, the integration offered by infrastructure layer products is a poor substitute for pre-integration in ERP software suites. Pre-integration allows tighter integration than the integration offered by infrastructure layer products. Certain functions previously performed within the ERP software layer are now performed in the infrastructure layer. Infrastructure layer products, however, do not contain business logic. Tr. at 4144:8-11, 4187 (Kutnik); Tr. at 1813-1814 (Wilmington); Tr. at 331-332 (Bergquist). Because infrastructure layer products do not contain business logic, a purchaser could not choose a more robust infrastructure layer product instead of ERP software. Accordingly, the decoupling of the infrastructure layer from the ERP software layer does not suggest that the infrastructure layer products are partially substitutable for ERP software.
Oracle’s experts Kutnick and Teece contend that the emergence of the infrastructure layer constitutes a paradigm shift in ERP software products. The age of infrastructure layer products calls into question this contention. See D7143 (Mills 5/27/04 Dep) Tr. at 30-31 (stating that IBM’s mid-dleware products have been in the market for nearly twenty years); Tr. at 420 (Kut-nik) (testifying that applications servers have been available for seven to eight years); Tr. at 3414:2-18 (Wohl) (noting that Oracle’s applications server has been through several versions); Tr. at 328 (Bergquist) (testifying about the evolution of web services protocols).
Even if the emergence of the infrastructure layer will have a substantial impact on the EAS software industry, more robust infrastructure layer products both enhance and diminish the likelihood of stack compe *1158 tition. On the one hand, decoupling applications services from the business logic provides the interoperability standard necessary to create multi-seller clustering. See Tr. at 4378^379 (Teece). On the other hand, enhanced infrastructure layer products increase interoperability with other stacks. See Tr. at 2885-2889 (Knowles); Tr. at 1637:7-22 (Bass); P3337; D7143 (Mills 5/27/04 Dep) Tr. 59-61; Tr. at 4150:9-19 (Kutnik).
Findings of Fact: Product Market Definition
In order to sustain plaintiffs’ product market definition the court must find, by a preponderance of the. evidence, that plaintiffs’ have shown an articulable and distinct product market for HRM and FMS sold by Oracle, PeopleSoft and SAP only that does not include mid-market software, outsourcing solutions, best of breed solutions, legacy systems or the infrastructure layer.
Based upon a. review of the law and the evidence, the court concludes that the plaintiffs have not met their burden of establishing that the relevant product market is limited to so-called high function FMS and HRM sold by Oracle, PeopleSoft and SAP. The equivocal and vague evidence presented by plaintiffs at trial does not permit the court to exclude mid-market vendors, outsourcing or best of breed solutions from any product market that includes ERP software sold by Oracle, PeopleSoft and SAP.
For reasons discussed above, the court cannot rely upon the testimony of the customer witnesses offered by plaintiffs in determining if plaintiffs have met their burden.' Likewise, the testimony of all three industry witnesses offered by plaintiffs affords no reliable or articulable basis to distinguish a high function product market. Ironically, much of plaintiffs’ testimony supports a finding of no clear or articu-lable distinction.
Accordingly, the full weight of the plaintiffs’ product market burden fell at trial upon Elzinga. In resolving the battle of the expert witnesses on product definition, the court must conclude that Oracle’s witnesses presented the better and more convincing case. Elzinga for all his indubitable credentials as an economist seemed mostly to apply the techniques of his avocational interest in mystery writing. See Ex. P4014A. The evidence Elzinga mar-shalled was circumstantial and highly qualitative.
Elzinga’s tabulations of concentration statistics from responses to the DOJ CIDs, Elzinga demo. ## 10-11, suffer from several shortcomings. Elzinga defined high function ERP as any sale in excess of $500,000. As the DAFs establish, ERP vendors sell a cluster of products. Sales exceeding a half-million dollars, therefore, are likely in many instances, if not most, to include pillars other than FMS and HRM. Elzinga’s chosen demonstrative, Ex. 4015A, will make the point. The sale in question, to Teradyne Corporation, met the $500,000 threshold. Ex. 4015A at OR-LITE0086650. Yet the discount Oracle offered on the HRM pillar license fee was 100 percent, and the bundle included modules in the SCM pillar along with modules in the HRM pillar. Despite this, Elzinga tabulated this entire transaction as an Oracle HRM sale, even though Oracle appeared to give away for free an HRM license in order to sell modules in the SCM pillar. Id. at ORLITE0086654. The court has not attempted to retabulate market shares to correct for these problems.
Elzinga’s other statistical tabulations are sketchy at best. The tabulation of Oracle customer surveys was a tiny sample of only twenty-eight sales opportunities. El-zinga demo # 7. The roster of Oracle DAFs was also short. Elzinga demo # 3. But even more troubling, as pointed out in *1159 connection with the Teradyne sale, is that these tabulations did not break out FMS and HRM sales from the bundles in which they were sold. Again, the DAFs register prices and discounts on a mixture of different pillars and modules. Metaphorically, Elzinga did not separate the wheat from the chaff.
Not only does the court find Elzinga’s data to be unreliable in establishing a distinct and articulable product market, but Elzinga himself admitted that plaintiffs’ product market has no “quantitative metric” that could be used to determine the distinction between a high function product and a mid-market product. Tr. at 2311:3-17 (Elzinga). Rather, Elzinga kept telling the court that there is “something different” about the products sold by Oracle, SAP and PeopleSoft. But the court cannot delineate product boundaries in multi-billion dollar merger suits based upon the mere notion that there is “something different” about the merging products and all others, especially when that “something different” cannot be expressed in terms to make a judgment of the court have meaning. More is required.
Accordingly, based upon the evidence presented at trial, the court concludes that the following products cannot be excluded from the relevant product market for purposes of analyzing the effects of this merger.
Outsourcing. Professor Hausman presented evidence of over twenty large enterprises that currently outsource all or some of their HRM needs. Furthermore, the testimony of Peters, Bass, Sternklar and Rising all support Hausman’s contention that large companies can, and do, have their HRM needs effectively met by outsourcing. Accordingly, outsourcing solutions cannot be excluded.
Plaintiffs argue that because several of the outsourcing firms themselves use Oracle, SAP or PeopleSoft, these outsourcing firms do not count as independent competitors. But the court finds the testimony of Bass and Sternklar regarding “blanket licenses” or “one-to-many” licenses to be the most reliable on how outsourcing works. Most outsourcers that handle HRM needs for large enterprises either have, or soon will have, a type of blanket contract with an ERP vendor. Under these contracts, the software vendors agree to provide software to the outsour-cer at a set price up to a certain number of employees, or “seats,” usually numbering well into the millions of employees. Fidelity’s contract with Oracle provides for a “seat” capacity of 2 million employees, with Fidelity having the option to increase the number of employees at a pre-set fee. Ex. D7158. So if Company X chooses to outsource through Fidelity, which may be operating on Oracle software, there is no direct connection between Oracle and Company X. There is no license between Oracle and Company X and no chance for Oracle to take advantage of Company X which has no “post-merger” choice in ERP software. Company X is merely more “seats” in Fidelity’s millions of empty seats under its blanket contract.
Moreover, several outsourcing firms currently use their own proprietary software, such as Hewitt and ADP. Fidelity has also begun the process of migrating clients from Oracle software to Fidelity’s own software. Tr. at 3154:3-15 (Sternklar).
Mid-market vendors. The court is perplexed about plaintiffs’ position that “mid-market solutions” are not part of the product market for high function ERP. Plaintiffs claim that mid-market vendors, such as Lawson and AMS could not constrain a post-merger SSNIP. Pis. Post Brief (Doc. #366) at 14. Such a statement clearly implies that plaintiffs do not view Lawson and AMS as high function vendors. But Elzinga’s high function market share cal- *1160 eulations showed Lawson and AMS each had market shares. See Elzinga demo. ## 10-11. Further, Elzinga stated that his calculations probably understated Lawson’s market share in the high function market. Lawson and AMS plainly cannot have market shares in the high function market if they are not a part of it. Accordingly, plaintiffs’ own evidence shows that either (1) Lawson and AMS are a substitute for high function vendors or (2) no substantive demarcation between these two types of vendors exists. This evidence glaringly shows that plaintiffs have failed to prove a distinct relevant product market for this court to analyze.
As surprising as plaintiffs’ evidence and statistics on the mid-market is one of the plaintiffs actual behavior. Plaintiffs characterize vendors that serve the mid-market as “hav[ing] limited capacity to support customers with diverse operations such as multiple geographic locations, distinct legal entities * * * or numerous lines of business.” FAC (Doc. # 125) ¶ 12 at 8. But, soon after filing its complaint, the United States Department of Justice itself — which surely meets at least two of these criteria — chose AMS, a so-called mid-market vendor, to meet its HRM and FMS needs. The DOJ chose AMS over Oracle and Peo-pleSoft.
Plaintiffs’ statistics, expert witness and behavior all treat mid-market vendors Lawson and AMS as part of the high function market. The court sees no reason why it should not follow suit.
Microsoft. As discussed above the court finds Burgum’s testimony regarding Microsoft’s entry into the up-market to be incredible. The testimony of Keating, as well as BearingPoint’s homepage, make it clear that Microsoft has every intention of using Axapta and BearingPoint to compete for so-called up-market business. Furthermore, Allen’s testimony about the struggle of J D Edwards in trying to enter the up-market does not apply to Microsoft. Microsoft has the money, the reputation and now, due to the BearingPoint alliance, it has the sales force necessary to become a major competitor for up-market business. Accordingly, the court finds that Microsoft will be a viable substitute for a significant number of consumers should a post-merger Oracle impose a SSNIP in its pricing of ERP software.
Best of breed solutions. The court does not dismiss defendant’s bundle argument as an “elaborate distraction” or “economical nonsense” as plaintiffs urge. Pis Post Brief (Doc. # 366) at 21-22. The reality of this industry is that 90 percent of consumers purchase software “bundles” containing several pillars; rarely does a consumer purchase a single pillar. Tr. at 3815:10-13 (Hausman). FMS and HRM pillars typically are sold in a bundle along with additional kinds of EAS, such as CRM or SCM. Further, the discounts that are offered to potential consumers are based on the value of the entire bundle, not simply based upon the presence of an HRM or FMS pillar. Tr. at 3813:23-3814:1 (Haus-man). Accordingly, when Oracle or Peo-pleSoft offer a discount on a bundle, they are doing so in order to ensure that the customer purchases all the pillars from Oracle or PeopleSoft, rather than turn to a best of breed vendor.
Incumbent solutions. The court, however, is not persuaded that incumbent solutions would be able to constrain a post-merger Oracle from imposing a SSNIP. Companies can, and apparently do, threaten to “do nothing,” in hopes of getting a better price on ERP software. See Campbell demo. ##20-21. But it is highly unlikely that any monopolist would see this threat as “credible,” thereby preventing a SSNIP. Given the ever-changing conditions of both the regulatory and technological aspects of human resources and finan *1161 cial management, it is hard to sustain the idea that large corporations would rather employ an antiquated software system than pay 10 percent more for modern and continuously maintained products. Such a choice in today’s business world would be extremely risky and unlikely.
Accordingly, without a relevant market having been established, the court cannot conduct a burden-shifting statistical analysis under Philadelphia Nat Bank, much less hold that plaintiffs are entitled to such a presumption. Nor, of course, can the court apply the concentration methodology of the Guidelines. See Guidelines § 1.51.
Plaintiffs’ Proposed Geographic Market
Assuming that high function FMS and HRM is the relevant product market, plaintiffs claimed that the relevant geographic market is the United States. Pis Post Brief (Doc. #366) at 22. Again, plaintiffs relied heavily on Elzinga’s testimony. In reaching this market definition, Elzinga ironically enough did not rely upon the oft-used Elzinga-Hogarty (E-H) test, which he admitted has been used in “dozens and dozens of merger cases” and which he himself co-developed. Tr. at 2154:22-23 (Elzinga).
In informal terms, the E-H test “measures the accuracy of a market delineation by determining the amount of either imports into or exports from a tentative market. The test is based on the assumption that if an area has significant exports or imports, then that area is not a relevant geographic market. Under the [test], exports or imports greater than 10% suggest that the market examined is not a relevant market.” United States v. Country Lake Foods, Inc., 754 F.Supp. 669, 672 n. 2 (D.Minn.1990).
Elzinga stated that he did not believe the E-H test “fit this particular antitrust case.” Tr. at 2154:25-25 (Elzinga). Instead, Elzinga relied solely upon the Guidelines “hypothetical monopolist” test in determining the geographic market. Tr. at 2204:1-11. (Elzinga). See Guidelines § 1.21. “I am persuaded that the United States [is the geographic market because] if [some] one were the sole supplier of high function FMS and HRM * * * in the US, and [he imposed a SNNIP], he would not be thwarted or undercut by economic * * * agents outside the United States.” Id.
Elzinga cited several relevant factors that led him to believe the Guidelines required a United States-only geographic market. Tr. at 2203:24-25 (Elzinga).
Where software code is written is not relevant to geographic market. “The [product] market here is high function FMS and HRM, and that is not just code. What you buy when you buy this product * * * is a relationship.” Tr. at 2154:10-14 (Elzinga) (emphasis added). Plaintiffs urged the court to exclude from the geographic market the site of manufacture. Hence, Elzinga urged the court to look beyond the location of manufacture for FMS and HRM. Since all of SAP’s software is manufactured in Germany and SAP indisputably produces high function ERP, inclusion of SAP’s site of manufacture would wholly undermine plaintiffs’ proposed geographic market.
Rather, Elzinga stated that the relevant factor in determining the geographic market is how the products are “marketed and supported” (i.e., the relationship) between the ERP vendor and the consumer. Tr. at 2202 (Elzinga). Elzinga argued that purchasing high function FMS and HRM entails installation, implementation, maintenance and upgrades — a relationship that has an inherently “local” aspect. Tr. at 2154:21-25 (Elzinga). Accordingly, since the relevant factor is the marketing and support of the software (which occurs in the United States) and not the “shipment” of the software from the manufacturing *1162 site (which could occur outside the United States), the E-H “shipments” test is not appropriate for this merger analysis. Tr. at 2205.T1-14 (Elzinga).
Under the Guidelines, because the relevant factor of “relationship” occurs only within the United States for United States customers, these customers could not seek substitutes abroad in the event of a SSNIP, thus making the United States the geographic market, according to Elzinga.
No arbitrage exists in this market. Arbitrage occurs when a consumer of a product buys the product from a vendor in one geographic location at a low price, but then sells the product to another consumer in a different geographic location for a higher price. Tr. at 2157:15-19 (Elzinga). Arbitrage is a factor that Elzinga stated can “stitch” together two geographic locations into “one geographic market” for merger analysis. Tr. at 2157:20-22 (Elzinga). El-zinga illustrated the phenomenon of arbitrage for the court via a precious stone hypothetical. “If the price of diamonds got relatively high in the United States, compared to * * * Europe, * * * arbitragers could buy diamonds where the price is low [Europe] and ship them to where the price is high * * * thereby eliminating the price difference [between] the two parts of the world.” Tr. at 2157:16-19 (Elzinga).
But, according to Elzinga, arbitrage is not a factor that can “stitch” the United States high function FMS and HRM markets to the same markets in other parts of the world. Tr. at 2205:21 (Elzinga). Arbitrage does not exist in the high function FMS and HRM markets for two reasons, he testified. First, the products that consumers buy from Oracle, PeopleSoft or SAP are licensed products, accordingly, the consumers “do not have the legal authority” to resell the software to other consumers. Tr. at 2158:6 (Elzinga). Second, high function FMS and HRM is tooled to “work * * * and meet the specific configurations and capabilities [of only one consumer], it won’t work [on another consumer’s computers].” Tr. at 2158:18-21 (Elzinga). Therefore, lack of the “arbitrage factor” reinforced Elzinga’s proposition that consumers cannot find substitute products outside of the United States, he testified.
Prices in the United States are not affected by prices in other parts of the world. Elzinga posited that United States consumers of high function FMS and HRM cannot expect to be charged the same price that a European consumer is paying. Tr. at 2206:7-11 (Elzinga). “The United States is not affected by prices or output of [high function FMS and HRM] outside the United States. And the flip side is also true. [P]rices charged outside of the United States aren’t affected by prices charged inside.” Tr. at 2206:10-12 (Elzin-ga).
Oracle’s Proposed Geographic Market
Oracle asks the court to reject the plaintiffs’ proposed geographic market. Oracle argues that the geographic market in this case is “so clear[ly] [a global market] that reasonable people ought not be debating it.” Def Post Brief (Doe. # 365) at 22. Further, Oracle noted that this is not the first time the DOJ has tried (unsuccessfully) to claim a United States-only market in the face of overwhelming evidence of a worldwide market. Def Post Brief (Doc. # 365) at 23 n. 19 (citing United States v. Eastman Kodak, 63 F.3d 95 (2d Cir.1995)).
Oracle assailed plaintiffs’ severance of SAP into two distinct companies. “The proposed United States-only market is a way of * * * making SAP appear ‘smaller’ than it really is and simultaneously making Oracle and PeopleSoft appear ‘bigger’ than they really are.” Def. Post Brief (Doc. # 365) at 23. While SAP America is responsible for all sales of SAP software in *1163 the United States and Canada, it sells software that was manufactured in Germany. SAP America has no North American manufacturing sites. Def. Fact (Doc. # 357) ¶ 100 at 50. Further, all large discount rates offered to United States customers by SAP America must be approved by SAP AG. Tr. at 2836:22-24 (Knowles). Accordingly, without SAP AG, SAP America would (1) have nothing to sell and (2) not be able to offer competitive discounts.
Moreover, simply because SAP has a larger market share in Europe does not mean that the geographic market should be limited to the United States. “Shares are not determinative of how you define the [geographic] market” Hausman testified. SAP — all of SAP — must be included he stated.
Once SAP is seen as a single entity, defendant claims that there are four different ways of analyzing the geographic market in this ease, all of which point to a worldwide market. Tr. at 3793:18-19 (Hausman).
First, Hausman analyzed the geographic market under the “hypothetical monopolist” test from the Guidelines. Tr. at 3794:9-10 (Hausman). See Guidelines § 1.21. Even assuming SAP America is distinct from SAP AG, Hausman stated that if a hypothetical monopolist in the United States imposed a SSNIP, SAP AG could “of course hire plenty of salespeople * * * and come in and compete.” Tr. at 3795:2-6 (Hausman). “[SAP AG’s] product would do just fine in the United States.” Id. Accordingly, “if [the court] looks at this market from a Merger Guidelines approach, you need to look at this on a worldwide basis.” Tr. at 3795:1-12 (Hausman).
Second, Hausman analyzed the geographic market under the plaintiffs’ description of the “high function needs” of the customers who buy high function software. Tr. at 3795:24-25 (Hausman). Hausman described the DOJ’s product definition as “multi, multi, multi,” referring to the functionality that the DOJ claims high function software possesses. Tr. at 3796:1-2 (Hausman). The software must be able to handle multiple currencies, from multiple jurisdictions, while understanding multiple languages. Different currencies and different languages are clearly “international or worldwide features,” and therefore “bring a worldwide aspect” to the analysis. Tr. at 3798:7-8, 18-20 (Haus-man).
Third, Hausman employed the E-H test that was rejected by its own creator. Tr. at 3800-3804 (Hausman). Hausman stated that this is a point that both he and Elzin-ga agree upon: the E-H test would only be satisfied if the geographic market were defined worldwide. Tr. at 3801:7-11 (Hausman). Hausman stated that Elzin-ga’s rejection of the E-H test was “inappropriate” for two reasons. First, there are several markets, other than the high function ERP market, where the client buys a “relationship” with the vendor (e.g., the purchase of a mainframe computer or server). But, it has never been argued that the computer market is not a worldwide market. Tr. at 3802:1-15 (Hausman). “We see [this kind of relationship] in all sorts of high-technology markets. Yet, people agree that those are all world markets.” Tr. at 3802:13-15 (Hausman). Second, the E-H test has “often” been applied to several cases involving services based upon a relationship with customers, such as hospital merger cases. Tr. at 3803:12-20 (Hausman). Accordingly, the E-H test is appropriate for this type of relationship-oriented scenario as well, and all agree that the E-H test mandates a worldwide market.
Finally, Hausman opined that there is empirical evidence showing that prices in Europe constrain prices in the United *1164 States, and vice versa. Tr. at 3805 (Haus-man). Hausman studied the PeopleSoft DAFs submitted to the DOJ. He found that the average discount rates for People-Soft in the United States was 45.2 percent. Tr. at 3805:19 (Hausman). In Europe, the average discount was 45.1 percent. Tr. at 3805:20 (Hausman). Hausman stated that these discount rates are “virtually identical.” Tr. at 3805:22. If the competitive conditions in Europe and the United States were wholly independent of each other, one would expect to see completely different discounts in both regions. But these facts demonstrate, in Hausman’s view, that the market needs to be analyzed on a global scale. Tr. at 3806 (Hausman).
Accordingly, Oracle urged the court to look at concentration figures based upon a global market of all FMS and HRM software. Def. Fact (Doc. # 357) at 56. Using these product and geographic market definitions, Oracle offered the following global FMS market shares: SAP, 19.2 percent; Oracle, 16.8 percent; and People-Soft, 12 percent. Ex. P0825 at 21. A merged Oraele/PeopleSoft would, in Oracle’s view, possess a 28.8 percent market share in the FMS market.
For global HRM, Oracle offers the following market shares: SAP, 11.9 percent; PeopleSoft, 11.3 percent; and Oracle, 3.04 percent. Ex. D5815 at 9. A merged Oracle/PeopleSoft would possess only a 14.3 percent market share in the HRM market.
Findings of Fact: Geographic Market
The court finds that the relevant geographic market (“the area of effective competition”) in this case is a worldwide market. Tampa Electric Co. v. Nashville Coal Co., 365 U.S. 320, 327-28, 81 S.Ct. 623, 5 L.Ed.2d 580 (1961).
At the outset, the court must address the plaintiffs’ attempt to sever SAP into two companies — SAP America and SAP AG. The court finds this argument wholly unpersuasive. SAP America, while critical to SAP’s success in North America, is not an independent company. This fact was clearly shown by the testimony of Knowles who stated that any large discount (usually above 70 percent) that SAP America offers, clearly in the face of competition, must get that discount approved by SAP AG in Germany. Further, while the source of the code is not determinative of this severance inquiry, it is important to note that all of SAP America’s software is manufactured and shipped from SAP AG. So without SAP AG, SAP America would have nothing to sell, and even if it did have its own manufacturing, SAP America would still have to get competitive discount rates approved by SAP AG. To view these two dependent branches of SAP as separate entities would be asking the court to ignore the reality of how the industry presently operates. Accordingly, the court finds that SAP must be viewed as one single entity.
Next the court must decide the geographic boundaries within which the market participants effectively compete. This court (per Judge Chesney) has used the E-H test in defining the relevant geographic market for merger analysis. See State v. Sutter Health System, 84 F.Supp.2d 1057, 1069 (N.D.Cal.2000) (“The analytical process [of defining the geographic market] generally begins with an application of the Elzinga-Hogarty test * * Furthermore, the results of employing the E-H test are undisputed. See Tr. at 2155:9-10 (Elzinga) (admitting that the E-H test would dictate the court to view the market as a global market). El-zinga’s basis for rejecting the E-H test is unpersuasive. The court, while agreeing that “relationships” are important in ERP sales, does not find that such relationships render the E-H test inapplicable. First, the court can think of a number of sales transactions that involve marketing and negotiation as well as installation and *1165 maintenance “relationships” between seller and vendor: computer sales, copier sales, motor vehicles to name a few. But to argue that these markets, all involving major foreign vendors, are limited to the United States would be untenable.
Second, the E-H test has been used when important vendor-customer relationships are involved. A clear example is Slitter Health, where Judge Chesney used the E-H test in a hospital merger case to determine whether patients seeking medical treatment outside of the Bay area required a geographic market expansion. No one can argue that medical treatment is not a “relationship” between the patient and the doctor. But the E-H test controlled the analysis, not the location of the “relationship” between the doctor and the patient. Further, Judge Chesney is not alone in applying the E-H test to hospital merger cases. See United States v. Mercy Health Services, 902 F.Supp. 968, 980 (N.D.Iowa 1995).
Finally, the court cannot allow this “relationship” factor to solely dictate the geographic boundaries for this case, as the court has already found that “non-relationship” solutions (i.e., outsourcing) cannot be excluded from the product market.
Accordingly, the court holds that the EH test is an appropriate method of determining the “area of effective competition” between vendors in this relevant market. Tampa Electric, 365 U.S. at 327, 81 S.Ct. 623. Elzinga, creator of the test, admitted that applying the E-H test would mandate a global market. The court therefore finds that the geographic market in this case is global.
Findings of Fact: Market Shares and Concentration
In addition to failing to meet their burden of proving a distinct product market, plaintiffs have failed to prove that the relevant product market in this ease is geographically bound to the United States. Accordingly, the market share and concentration statistics presented by Elzinga are wholly inapplicable to the court’s analysis. The court is left with a new product market definition which includes, at least: (1) ERP sold by Oracle, SAP, PeopleSoft, Lawson, AMS and Microsoft; (2) outsourcing solutions; and (3) best of breed solutions. Further, this product market must be analyzed as a global one.
. Not surprisingly, plaintiffs did not offer any market share data other than those of Elzinga. Oracle, while successfully picking apart plaintiffs’ market definition did not provide a definitive alternative of its own. The only statistical data Oracle offered showed the 2002 global HRM shares of Oracle, PeopleSoft and SAP, but did not include HRM data on AMS or Microsoft’s share since the BearingPoint alliance. Moreover, Oracle offered even less in the way of FMS shares or concentration.
But it is plaintiffs, not defendant, who carry the burden of proving market shares and concentration in order to invoke the presumptions of the case law or to sustain a showing in accordance with the Guidelines. The court cannot furnish its own statistics.
Without the benefit of presumptions, the burden remains upon plaintiffs to come forward with evidence of actual anticom-petitive effects.
Plaintiffs’ Evidence of Coordinated Effects
Plaintiffs presented no evidence at trial on coordinated effects. This was a wise decision, as proving the probability of such collusion would definitely be an uphill battle for two reasons. First, the products in the relevant market are not homogeneous. Plaintiffs themselves even argue against homogeneity. Pis. Post Brief (Doc. # 366) at 30 (stating that the prod *1166 ucts in the high function HRM and FMS market are “highly differentiated”). Second, there is no price transparency in this market. Prices and discount rates for software are known only to the vendor and the customer, both of whom take great pains to keep such information confidential. Without homogeneity or transparency, the market conditions are not conducive to coordinated effects, either tacit or express. Plaintiffs recognized this unlikelihood. Id. at 38 (“The fact that high function software is a differentiated product and that pricing is not transparent make price coordination between Oracle and SAP unlikely.”).
But in plaintiffs’ post-trial brief they unexpectedly included an entire section arguing that a post-merger Oracle and SAP could tacitly collude in allocating customers or markets. Id. at 38^40. Plaintiffs argue that “Oracle is strong in the high technology and telecommunications” area while “SAP dominates the oil and gas industry.” Id. at 39. Accordingly, Oracle and SAP could reach a tacit understanding based upon “mutual trust and forbearance” and stop competing against each other in those relevant areas. Pis. Post Brief at 38 (quoting Hospital Corp. of Am., 807 F.2d at 1391). But the court has searched in vain for any testimony or exhibits regarding tacit territorial or market divisions by Oracle and SAP. With no evidence in the record regarding such a speculative coordinated effects argument, the court finds this new theory to be without merit.
Plaintiffs’ Evidence of Unilateral Effects
Plaintiffs rest their theory of anticom-petitive effects on an attempt to prove that Oracle and PeopleSoft are in a “localized” competition sphere (a “node”) within the high function FMS and HRM market. This sphere does not include SAP or any other vendors, and a merger of Oracle and PeopleSoft would, therefore, adversely affect competition in this localized market. Pis. Post Brief (Doc. # 366) at 31-36; Tr. at 2448-2450 (McAfee). Plaintiffs also offered evidence to show that SAP could not reposition itself to replace the localized competition that would allegedly be lost if Oracle and PeopleSoft merge. Pis. Post Brief (Doc. # 366) at 32-33.
In attempting to prove localized competition between Oracle and PeopleSoft, plaintiffs relied on virtually the same kind of evidence used to prove the product market, including internal corporate documents, SAP executive testimony, customer and consultant firm testimony and expert testimony.
Internal documents. Plaintiffs rely upon several quarterly “win/loss analysis” documents that were compiled by Oracle during 2003 to show that Oracle and Peo-pleSoft are each other’s “closest competitors.” Pis. Post Brief (Doc. # 366) at 31. In Quarter 1 of 2003, plaintiffs offered evidence that Oracle lost to PeopleSoft 37 percent of the time when the two were in competition, while Oracle lost to SAP only 15 percent of the time the two competed. Ex. P2090. Plaintiffs then offered evidence from Quarter 3 in which Oracle “explicitly states” that “PeopleSoft is our Number # 1 competitor” and “SAP is our Number # 2 competitor.” Ex. P2093.
But what plaintiffs failed to mention regarding the Quarter 3 findings is that Oracle lost to PeopleSoft 54 percent of the time, while they lost to SAP 53 percent of the time. Accordingly, what separates the “# 1 competitor” and “# 2 competitor” of Oracle is merely one percent. Ex. P2093. Moreover, these roughly equal loss ratios continued into Quarter 4 when Oracle lost to PeopleSoft 59 percent of the time, while losing to SAP 50 percent of the time. Ex. P2095. Accordingly, the court can draw no conclusions from the conflicting data within the win/loss reports upon which *1167 plaintiffs focus. In fact, these documents arguably negate a showing of localization between Oracle and PeopleSoft more than they support such a finding.
SAP executive testimony. Plaintiffs attempt to localize PeopleSoft and Oracle by showing that many customers have a negative “perception” of SAP and that SAP is at a “substantial disadvantage” when it comes to competing for customers in the United States (the geographic market that the court has already rejected). Pis. Post Brief (Doc. # 366) at 31-32. In proving these negative perceptions, plaintiffs pointed to the testimony of SAP America’s Knowles. At trial, Knowles agreed that SAP has had to deal with “perceptions” that SAP is “too costly and difficult to implement.” Tr. at 2950:8-12 (Knowles). Further, plaintiffs cited evidence from consulting firms and Knowles stating that SAP has had “trouble” breaking into certain verticals in the United States. See Ex. P3037 (Knowles dep 5/3/04) at Tr. 67:21-68:7 (difficulty breaking into services sector); Tr. at 1698:1-8 (Bass) (difficulty in entering banking industry).
In deciding the merits of this argument, the court is again perplexed by the inconsistency within plaintiffs’ own evidence. In trying to prove Oracle and PeopleSoft are in localized competition, plaintiffs tried to downplay SAP’s presence in the United States and characterize SAP has being “disadvantaged” and unable to enter several markets. But plaintiffs’ own evidence on market shares negates such a finding. Even assuming the relevant geographic market in this case was the United States, Elzinga’s calculations of market shares in so-called high function FMS has SAP ranked highest (above Oracle and People-Soft) with a 39 percent market share. El-zinga demo # 10. Moreover, in the HRM high function market, plaintiffs’ expert ranked SAP second with a 29 percent market share (beating Oracle). Elzinga demo # 11. SAP is not a “disadvantaged” and “troubled” competitor in the United States. If it were, SAP should not be beating Oracle in both markets and beating PeopleSoft in the FMS market. Accordingly, the court cannot credit plaintiffs’ argument that SAP is suffering from negative customer perceptions or is disadvantaged in competing against Oracle and PeopleSoft.
Customer and consulting firm testimony. In furtherance of this localization theory, plaintiffs argued that customer testimony shows that “Oracle and PeopleSoft present better alternatives in the United States than SAP.” Pis. Post Brief (Doc. # 366) at 32. Plaintiffs support this assertion by citing the testimony of five customers who eliminated SAP from the final round of negotiations and instead chose to deal with Oracle and PeopleSoft. Id. (citing testimony of North Dakota, Neiman Marcus, Greyhound, AIMCO and Cox).
The court finds this evidence unpersuasive for two reasons. First, the court cannot take the self-interested testimony of five companies which chose to eliminate SAP from consideration, and from that sample draw the general conclusion that SAP does not present a competitive alternative to Oracle, and PeopleSoft. , Drawing generalized conclusions about an extremely heterogeneous customer market based upon testimony from a small sample is not only unreliable, it is nearly impossible. See Sungard Data Sys., 172 F Supp 2d at 182-83. Second, the most persuasive testimony from customers is not what they say in court, but what they do in the market. And as Elzinga’s statistics showed, customers are buying SAP FMS more than Oracle and PeopleSoft FMS. Elzinga demo # 10. Customers are buying SAP HRM more than that of Oracle. Elzinga demo #11.
*1168 Plaintiffs rely upon two of the Big Five consulting firms’ testimony stating “they believe SAP is often the third choice of many U.S. customers.” Pis Post Brief (Doc. # 366) at 32. According to Bearing-Point’s Keating, SAP has long been the least flexible of the three vendors in the way it has sold its HRM and FMS software. Tr. at 901:6-20, 946:18-20 (Keat-ing). Also, Accenture’s Bass testified that SAP was “less likely to discount than Oracle and PeopleSoft.” Pis Post Brief (Doc. #366) at 32; P3198 (Bass Dep) at Tr. 132:08-133:07. But the plaintiffs’ own evidence discounts this argument. While it may be true that SAP has been the least flexible and least likely to discount, the evidence introduced by Elzinga shows that customers apparently are not deterred by SAP’s inflexibility or higher pricing. Customers still buy SAP software over Oracle and PeopleSoft. See Elzinga demo. ## 10-11.
Taken as a whole, the customer and consulting firm testimony falls short of proving that Oracle and PeopleSoft engage in competition to which SAP is simply not a party. Moreover, both PeopleSoft industry witnesses conceded there is no vertical in which SAP is not competitive with Oracle and PeopleSoft. Tr. at 388:1-11 (Bergquist); 1957:10-21 (Wilmington).
Expert testimony. Finally, plaintiffs offered the testimony of Professor McAfee to show that PeopleSoft and Oracle are engaged in localized competition to which SAP is not a party. McAfee conducted three independent analyses to reach his conclusions. Pis Post Brief (Doc. #366) at 34.
First, McAfee examined, in detail, twenty-five of Oracle’s DAFs in which Oracle salespersons had listed PeopleSoft as their justification for seeking a higher discount. Second, McAfee, using charts of discount trends provided by Oracle, ran a regression analysis to assess the effect of People-Soft’s presence on Oracle’s discount levels. Third, using the market statistics calculated by Elzinga, McAfee conducted a “merger simulation” to assess the theoretical effects of an Oracle/PeopleSoft merger. Tr. at 2447-2449 (McAfee). Based upon these three independent studies, McAfee concluded that in many instances People-Soft and Oracle are each other’s closest competitor and a merger between the two would cause significant anticompetitive effects. Tr. at 2466:8-13, 2449:22-24 (McAf-ee).
Twenty-five case studies. At trial, McAfee showed the court several DAFs in which the presence of PeopleSoft had justified an Oracle salesperson seeking a steep discount. McAfee then picked out explicit language from the justification column to prove that when Oracle and Peo-pleSoft compete, they do so vigorously. For example, when seeking a discount on the Hallmark account, a salesperson’s justification for a discount was an “EXTREMELY competitive situation against” PeopleSoft. Because of this competition, a “higher discount was warranted.” Tr. at 2464:15-21 (McAfee). Likewise, in trying to win the Greyhound account Oracle wanted to cause a “third straight loss” for PeopleSoft and “only aggressive proposals” would win Greyhound over. Tr. at 2466:14-20 (McAfee).
These two examples are representative of the many that McAfee showed the court — clear examples of how vigorously PeopleSoft and Oracle compete when they go “head to head” against each other, he asserted. McAfee concluded that such head to head competition between Oracle and PeopleSoft would be lost if this merger were consummated. Tr. at 2488:13-25 (McAfee).
Regressions. Next, in trying to show localized competition, McAfee used a regression technique to calculate what effect, *1169 if any, the presence of PeopleSoft or another competitor has on the discounts offered by Oracle. Tr. at 2495:22-25 (McAf-ee). McAfee ran two regression analyses. In the first, McAfee was privy to sales representative surveys identifying the discount percentages given to Oracle customers that had purchased the E-Business Suite. The surveys also identified the competitor that Oracle had beaten to get the account. Tr. at 2497:10-14 (McAfee). McAfee narrowed the sample to all sales that were over $500,000, in order to equate the sample with Elzinga’s market definition. McAfee used these variables (competitor, net revenue, discount percentage) and ran the regressions. Tr. at 2498:3-20 (McAfee). The data led McAfee to conclude that “PeopleSoft has a .097 (9.7 percent) effect” on the discount Oracle offers. Tr. at 2499:22-25 (McAfee). In other words, when Oracle competes against Peo-pleSoft for the sale of Oracle’s E-Business Suite, the consumer obtains a 9.7 percent greater discount than when Oracle competes against no one in selling the suite.
Wanting to look at more than just the sale of the E-Business suite, McAfee then analyzed all of the DAFs that Elzinga had used in defining the product market and matched those with the data from the sales representative forms to create a larger sample with more variables. Tr. at 2504:22-25 (McAfee). The DAFs listed the percentage requested along with the competitor justifying such a discount. Once McAfee ran this second regression, he concluded that PeopleSoft had a .136 effect on Oracle’s discount rates (i.e., 13.6 percent greater discount). Tr. at 2507:6-11 (McAfee). Accordingly, McAfee concluded that when PeopleSoft is competing against Oracle, Oracle’s discounts are 9 to 14 percentage points greater. Tr. at 2508 (McAfee).
Based upon these DAF studies and regression analyses, it is safe for the court to conclude that Oracle and PeopleSoft do compete frequently for ERP customers and when they do compete, that competition can be vigorous. But these two contentions are not disputed by anyone in the case. Oracle concedes that PeopleSoft is a frequent rival. Def Post Brief (Doc. # 365) at 34. The court fails to understand what this undisputed fact is supposed to show about whether Oracle and PeopleSoft are competing head to head in a product space in which SAP is not a party. McAfee himself stated that from these twenty-five DAFs, he drew the “broad conclusion that in many instances PeopleSoft and Oracle are each others’ closest competitors.” Tr. at 2466:10-12 (McAfee). But these DAFs tell the court nothing about how often SAP competes against PeopleSoft or Oracle (a key factual issue if trying to exclude SAP) or whether that competition is equally fierce. What would have been more helpful to the court would have been the DAFs of PeopleSoft and SAP as well. Defendants introduced several SAP DAFs during trial, one showing a very aggressive competition against Oracle, so it is clear that such forms exist. Ex D5649R A more complete DAF record would perhaps have evidenced localized competition. between Oracle and People-Soft. But plaintiffs did not provide such DAFs to McAfee, nor is it clear whether they even sought to obtain such documents during discovery.
. Simply because Oracle and PeopleSoft often meet on the battlefield and fight aggressively does not lead to the conclusion that they do so in the absence of SAP.
Merger simulation. Finally, McAfee conducted a merger simulation analysis. There are several merger simulation models that can be used depending on the characteristics of the industry. Tr. at 2511:12-19 (McAfee). McAfee chose the “English auction” model (also called the *1170 oral ascending auction) because one of the features of this model is its allowance of multiple bidders and multiple rounds of bidding. Tr. at 2511:19-22 (McAfee). The simulation works by putting in necessary variables and assumptions, such as market shares and percentage of wins in head to head competition. Once these variables were accounted for, McAfee still had to set a variable for “how competitive the market [was] pre-merger.” Tr. at 2526:17-22 (McAfee). One way of creating such a measurement is by estimating the “total value of the product that accrues to the buyer” (i.e., how “much of the value of the software to the buyer actually accrues to the buyer and how much accrues to the vendors in the form of price”). Tr. at 2517:1-4 (McAfee). McAfee ran the simulation based upon five different “buyer accrual” estimates: .5 (only 50 percent accrual) to .9 (90 percent accrual). McAfee used the market shares calculated by El-zinga as his market shares variable. Once all the data are compiled and the variables accounted for, the merger is simulated by merging the shares of the two merging firms. Once this is done, data can be calculated showing how much the price of the relevant product is expected to increase. McAfee asserted that in the high function FMS market, after the Oracle/PeopleSoft merger, he expects price to increase anywhere from 5 percent (.50 accrual variable) to 11 percent (.90 accrual variable). In the high function HRM market, McAfee concluded that the price would increase by 13 percent (.50 accrual variable) to 30 percent (.90 accrual variable). Ex. P4024.
McAfee asserted that this merger simulation, using Elzinga’s market share statistics, shows that a merger between Oracle and PeopleSoft will lead to a unilateral price increase in both markets.
But the court has already found that Elzinga’s market share statistics are not a reliable indicator of Oracle, SAP and Peo-pleSoft’s positions in the ERP market. Accordingly, because this merger simulation is based upon these unreliable data, the court concludes that the simulation results are likewise unreliable.
Oracle’s Competitive Effects Rebuttal
Oracle takes issue with all of the plaintiffs’ evidence regarding the likelihood of anticompetitive unilateral effects.
First, Oracle claims that the present case is not the type of case for which the doctrine of unilateral effects was created. Oracle offered Campbell’s expert testimony that a fundamental assumption of the unilateral effects theory is not present in this case. Tr. at 2721:3-5 (Campbell). Campbell testified that the “unilateral effects doctrine is posited on the notion of a localized market powered by a seller and a group of purchasers located in product space or geographic space around that particular seller.” Tr. at 2721:5-9 (Campbell). This “product space” is defined by characteristics of the product or products within the space. Campbell offered a homey example of product space using breakfast cereal. Tr. at 2721:15-18 (Campbell). A number of customers have characteristic preferences for their breakfast cereal that could create a product space within the entire breakfast cereal market. For example, some customers prefer cereal to be crunchy, sugar-free and high in fiber. These characteristics of the product will narrow the entire market down to a “space” in which only crunchy, sugarless, high fiber cereals occupy the space and only those companies that produce such cereal are competitors. Campbell called this space a “node,” with the buyers being centered around this node. The unilateral effects theory is concerned about there being only one vendor operating inside the node, thereby being able to increase the price unilaterally. Tr. at *1171 2721:19-23 (Campbell). Plaintiffs attempted to carve out a “node” for high function FMS and HRM software in the United States in which only Oracle and PeopleSoft compete. Accordingly, if a merger takes place, there will be only one vendor in this node with the ability unilaterally to reduce output and raise price within the node.
Campbell asserted that the unilateral effects theory is predicated on the fundamental assumption that the consumers in the node have no “buyer power.” He testified that the theory assumes that customers are unsophisticated buyers who will not be able to rebuff a price increase. Tr. at 2721:23 (Campbell). This fundamental assumption does not hold in the case of the products in suit. Campbell asserted that the buyers of high function FMS and HRM are extremely sophisticated and knowledgeable and engage in extensive and intensive one-on-one negotiations with vendors. These customers clearly have a lot of power during these negotiations, Campbell claimed, and they are aware of this power. Tr. at 2722:1-4 (Campbell). Campbell gave examples of high function consumers such as Emerson Electric and Daimler whose representatives testified that their companies have “leverage” and “power over people they deal with,” and use their “size” and “the size of the deal” to gain better deals on software. Campbell demo #25 (citing Tr. at 1287:22-1288:2 (Peters); Tr. at 1407:20-1408:1 (Gorriz)). Campbell concluded that the unilateral effects theory is “dogma developed for a totally different context” from the present case. Tr. at 2728:6-7 (Campbell).
Even assuming arguendo that a unilateral effects theory is appropriate for this case, Oracle attacks each piece of evidence that plaintiffs put forward attempting to prove localization between Oracle and Peo-pleSoft.
Oracle objected to plaintiffs’ characterization of SAP as a struggling firm with a substantial disadvantage which prevents it from being in a localized space with Oracle and PeopleSoft. Def. Post Brief (Doc. #365) at 33. Oracle claims that these SAP “struggling” assertions are “not remotely true” and are belied by the fact that SAP has over 22,000 professional service customers. Id. While Oracle admits that SAP does not “dominate” the United States in the manner that it may “dominate elsewhere,” non-domination does not equate with “struggling.” Id.
Finally, Oracle takes aim at McAfee’s expert testimony on anticompetitive effects. First, Oracle claims that McAfee’s “case studies” based upon the Oracle DAFs do nothing more than “show Oracle and PeopleSoft are frequent rivals.” Id. at 34. This evidence reveals nothing about localization between Oracle and PeopleSoft in a product space in which SAP is not encompassed. McAfee offered no insights regarding the characteristics of high function FMS and HRM that create the alleged product space between Oracle and PeopleSoft. Further, these case studies are devoid of any information about whether head to head competition between Oracle and SAP, or PeopleSoft and SAP, is equally vigorous.
With regards to McAfee’s regression analysis, Oracle argued the analysis was flawed from the outset. The data upon which McAfee based his regression were “not based on any set of data identifying * * * high function HRM and FMS software, but only on data involving broader suites of EAS.” Id. at 36. Accordingly, it is impossible to know if these alleged increased discount rates were the product of high function FMS and HRM, other ERP pillars or the bundling of all. Without this crucial information, the regression analysis shows nothing in regards to localization *1172 between Oracle and PeopleSoft in a high function FMS and HRM product space. Id. at 37.
Furthermore, Oracle assails McAfee’s merger simulation as “simplistic” and “spurious.” Id. Oracle cites two major flaws in the merger simulation. First, the “auction” model is wholly inappropriate for the present market because (1) the customers in this market are extremely powerful at bargaining and (2) vendors of ERP do not simply “bid” for business; rather these negotiations are extensive and prolonged, with the purchaser having complete control over information disclosure. Id. Second, the “market shares — the only input having any connection to real-world data — were those produced by Elzinga using the plaintiffs’ market definition.” Id. at 38. Because Oracle wholly rejects plaintiffs’ “gerrymandered” market definition, market statistics based upon this definition are equally flawed. Accordingly, the merger simulation’s prediction of price increases after the merger are inaccurate and unreliable, based as it is on an inappropriate model using inaccurate data.
Finally, and perhaps most importantly, Oracle contends that plaintiffs have offered no “econometric calculations in trying to prove localization.” Id. at 31. Oracle argues that proving localization requires “extensive econometric analysis,” such as diversion ratios, price-cost margins and the like, of which plaintiffs have offered none. When Oracle cross-examined plaintiffs’ expert witnesses, both admitted that they “did not even attempt to calculate diversion ratios, or cross-elasticities, or any other economically meaningful measurement of whether the products of Oracle and PeopleSoft are uniquely close substitutes for each other.” Id. See Tr. at 2293:23-25 (Elzinga); Tr. at 2599:3-8 (McAfee).
Findings of Fact: Unilateral Effects
The court finds that the plaintiffs have wholly failed to prove the fundamental aspect of a unilateral effects case — they have failed to show a “node” or an area of localized competition between Oracle and PeopleSoft. In other words, plaintiffs have failed to prove that there are a significant number of customers (the “node”) who regard Oracle and PeopleSoft as their first and second choices. If plaintiffs had made such a showing, then the court could analyze the potential for exercise of monopoly power over this “node” by a post-merger Oracle or the ability of SAP or Lawson to reposition itself within the node in order to constrain such an exercise of monopoly power.
Plaintiffs’ attempt to show localized competition based upon customer and expert testimony was flawed and unreliable. Moreover, plaintiffs’ evidence was devoid of any thorough econometric analysis such as diversion ratios showing recapture ef-' fects. Both the Kraft Gen Foods and Swedish Match courts, the only other courts explicitly to address unilateral effects, based their rulings in part upon econometric evidence submitted by the parties. Kraft Gen. Foods, 926 F.Supp. at 356 (relying on econometric evidence of the cross-price elasticity of demand between Post cereal brands and Nabisco brands); Swedish Match, 131 F Supp 2d at 169 (relying upon the diversion ratio between two brands of loose leaf tobacco).
Plaintiffs claim they were unable to present the court with such econometric data because “this [the high function HRM and FMS market] is a market that’s shot through with price discrimination,” and therefore such data would be unreliable. Tr. at 2291:15-16 (Elzinga). But the court finds plaintiffs’ price discrimination argument unpersuasive. First, “this” market which Elzinga claims is plagued by price discrimination, is the so-called high function FMS and HRM market that the court has already rejected as being the relevant *1173 product market in which to examine the effects of the proposed merger. Second, assuming that the high function FMS and HRM market were the relevant market, which it clearly is not, plaintiffs only evidence regarding price discrimination came from Elzinga’s analysis of the Oracle DAFs. Elzinga stated that there was a wide range of discounts offered by Oracle to these 222 customers. Tr. at 2222:13-19 (Elzinga). Elzinga stated that because Oracle charged different discounts to these customers, Oracle must be able to determine what price it can charge a customer before the customer eliminates Oracle as a potential vendor (i.e., Oracle price discriminates). And since Oracle price discriminates, then SAP and PeopleSoft must price discriminate as well.
But Elzinga admitted he conducted no formal studies of price discrimination in “this” market. Tr. at 2343:14-20 (Elzin-ga). Nor did he examine the discounts given by PeopleSoft or SAP to their respective customers. Tr. at 2351:10-14 (Elzinga). Elzinga’s assertion that this market is “shot through” with price discrimination because “somehow” Oracle was able to determine what level of discount it could offer to different customers uncannily resembles his argument that there is “something different” about Oracle, PeopleSoft and SAP. Again, the court refuses to sustain plaintiffs’ inarticulable contentions.
In sum, the court finds that plaintiffs have failed to show an area of localized competition between Oracle and People-Soft.
Oracle’s Efficiency Defense
Oracle offers an efficiency defense to rebut plaintiffs’ claim of anticompetitive effects. Def. Post Brief (Doc. # 365) at 39-40. Oracle claims that the merger will result in two overall efficiencies: (1) significant cost-savings for Oracle in many areas of business, and (2) an increase in Oracle’s scale (i.e., customer base), thereby fueling more competition with SAP, Siebel and Microsoft resulting in higher innovation and lower costs. Def. Fact (Doc. #357) ¶¶ 234-237 at 113, ¶¶ 247-251 at 118-21.
Oracle’s cost-savings evidence came from a spreadsheet originally compiled in May 2003 when Oracle wanted to acquire J. D. Edwards. The spreadsheet was revamped in June 2003 when Oracle sought to acquire PeopleSoft. It was finalized in July 2003 when Oracle looked at acquiring both. Tr. at 3469:5-12, 3470:19-20 (Catz); Ex. D7132. (Acquisition Efficiencies Analysis) (AEA). The AEA lists, as of July 2003, PeopleSoft’s total costs for the areas of sales and marketing (S & M), research and development (R & D) and general and administrative (G & A). Id. For 2003, PeopleSoft’s total cost of S & M was $769.3 million, R & D was $466.9 million and G & A costs were $214 million. Id. The AEA projects that one year after Oracle has acquired PeopleSoft, the cost of S & M will decrease to $34 million ($735.3 million in savings), R & D will decrease to $201.3 million ($265.6 million in savings) and G & A will decrease to $37.4 million ($176.6 million in savings). Id. Accordingly, Oracle argues that post merger, it will achieve cost-savings of over $1 billion. Def Fact (Doc. # 357) ¶ 234 at 113. Moreover, the cost savings are annual. So Oracle would save $1 billion in 2005, $1 billion in 2006, and so forth. Tr. at 3493:2-5 (Catz).
Catz further testified to the efficiencies that would result if Oracle’s scale were expanded to include PeopleSoft’s customers. Tr. at 3438-3439 (Catz). Catz stated that one of the main reasons, aside from cost savings, that led Oracle to make a tender offer for PeopleSoft was the potential acquisition of PeopleSoft’s “customer base.” Tr. at 3438:20 (Catz). The scale of a company is a source of annuity revenue, revenue which allows a company to invest *1174 more in research and development of its products. Id. By acquiring PeopleSoft, Oracle would capture the extra revenue of PeopleSoft’s customer base as well as the potential for revenue from sales of add-on products. Tr. at 3439:6-12 (Catz). This additional revenue and customer base would allow Oracle to expand its R & D, thereby fueling more innovation of Oracle software. Specifically, Catz testified about a new “superset product line” that would have the “best features from Oracle” and the “best features and modules from Peo-pleSoft.” Tr. at 3451:2-7 (Catz).
Further, the larger customer base and increased innovation would allow Oracle to compete with larger competitors, such as Microsoft, and compete better in other ERP markets, such as SCM and CMS. Tr. at 3440:3-7 (Catz); Def Fact (Doc. # 357) ¶251. Reduced costs, increased innovation and more competition are efficiencies Oracle claims outweigh, and thus rebut, any showing of anticompetitive effects plaintiffs have put forward.
Plaintiffs’ Efficiency Rebuttal
Plaintiffs rebutted the efficiency defense by calling Professor Zmijewski, a professor of business from the University of Chicago. Zmijweski was asked to verify the arithmetic in the AEA spreadsheet that Oracle claims explicate its large cost-saving efficiencies. To verify the spreadsheet, Zmijewski was required to “tease out” all of the inputs (i.e., the pre-merger costs and the post-merger costs of all departments) that had been plugged in by Oracle, verify that those inputs were true (based in fact) and then recalculate the numbers to verify that the final efficiency amounts were the same as the amounts represented on the AEA. Tr. at 4509:16, 4517-4518 (Zmijewski).
Zmijewski teased out the inputs successfully then began looking at information provided by Oracle and the SEC for some “factual foundation” for these inputs and post-merger assumptions Oracle had used in calculating the AEA. Tr. at 4520:5 (Zmi-jewski). But Zmijewski hit a “dead end” every time he tried to find some factual basis for any of the inputs in the spreadsheet. Id. A four month search through the documents left Zmijewski with “essentially none” of the information he needed to verify the AEA inputs. Tr. at 4520:11 (Zmijewski). Zmijewski’s uneasiness about his fruitless search was relieved when he found that there was no factual basis for the inputs. Catz had “used her personal judgment” based upon consultation with Larry Ellison and others in determining the inputs that went into the AEA. Tr. at 4520:14-23 (Zmijewski); 3558:1-8 (Catz). Further, there were no documents that could explain how Catz and others had reached these personal judgments on the inputs. Tr. at 3558:21 (Catz). This led Zmijewski to conclude that the AEA is “not verifiable” and therefore not reliable under the verification standards used by many professionals, including the SEC. Tr. at 4519:24, 4516:5-12 (Zmijewski). Plaintiffs claim that cost-saving efficiencies require defendant to “ ‘explain the methods used to calculate’ ” the cost-saving numbers. Pis Post Brief (Doc. # 366) at 47 (quoting Staples, 970 F.Supp. at 1089). According to plaintiffs, Oracle has provided no explanation of the methods used to calculate the AEA other than the judgment of Catz and her colleagues.
Finally, plaintiffs urge the court to put no stock in Oracle’s innovation claims, as they are unverified and not merger-specific. Pis. Post Brief (Doc. # 366) at 49-50. When Catz was cross-examined about the superset product line, the innovative hybrid of Oracle and PeopleSoft, she did not have any documents discussing this proposed innovation, nor did she know any details about when the product would be available. Pis. Post Brief (Doc. # 366) at 50; Tr. at 3533:8-16 (Catz). Plaintiffs *1175 claim this “vague” assertion of a superset product line is not a cognizable innovation claim under case law or the Guidelines. Pis. Post Brief (Doc. #365) at 49 (citing Heinz, 246 F.3d at 723 (requiring “reliable and significant evidence that the merger will permit innovation that otherwise could not be accomplished * * *.”)).
Findings of Fact: Efficiencies
In order for a claimed efficiency to be “cognizable,” it must be “substantiate[d]” and “verfi[able].” Guidelines § 4.0. The court finds Oracle’s evidence on the claimed cost-savings efficiency to be flawed and unverifiable. Catz and Ellison’s personal estimations regarding the potential cost-savings to Oracle are much too speculative to be afforded credibility. Oracle’s efficiency defense based upon future innovations (e.g., the superset product) was not verified by internal documents. Oracle presented no evidence regarding the functionality or characteristics the innovative product will contain, nor any evidence regarding its date of availability.
Accordingly, both claimed efficiencies are much too vague and unreliable to rebut a showing of anticompetitive effects.
Conclusions Of Law
This court has jurisdiction over this action pursuant to 28 USC §§ 1331, 1337(a) and 1345 and Section 15 of the Clayton Act, 15 USC § 25. Venue is proper in this district pursuant to 15 USC § 22 and 28 USC § 1391(c).
In order to succeed on their claim, plaintiffs must prove by a preponderance of the evidence (1) the relevant product and geographic market, and within this market (2) the effect of Oracle’s acquisition of People-Soft may be substantially to diminish competition. See Penn-Olin, 378 U.S. at 171, 84 S.Ct. 1710.
Plaintiffs alleged a product market limited to HRM and FMS software licensed by Oracle, PeopleSoft and SAP. Plaintiffs also alleged a geographic market limited to the United States.
Plaintiffs have proven that the relevant product market does not include incumbent systems or the integration layer. But plaintiffs failed to prove that outsourcing solutions, best of breed solutions and so-called mid-market vendors should be excluded from the relevant product market. Furthermore, plaintiffs have failed to establish that the area of effective competition is limited to the United States. Accordingly, plaintiffs have failed to meet their burden of proving the relevant market for section 7 analysis.
Because plaintiffs have failed to meet this predicative burden, plaintiffs are not entitled to a presumption of illegality under Philadelphia Nat Bank or the Guidelines.
Plaintiffs have failed to prove the likelihood that a post-merger Oracle and SAP would tacitly coordinate by allocating customers or markets. Accordingly, the plaintiffs have not met their burden of establishing anticompetitive coordinated effects.
Plaintiffs have failed to prove an area of localized competition between Oracle and PeopleSoft in which a post-merger Oracle could profitably impose a SSNIP. Accordingly, plaintiffs have not met their burden of establishing the likelihood of anticom-petitive unilateral effects.
Notwithstanding that plaintiffs have failed to carry their burden to be entitled to relief, Oracle has not proved by a preponderance of the evidence cognizable efficiencies sufficient to rebut any anticom-petitive effects of Oracle’s acquisition of PeopleSoft.
Because plaintiffs have not shown by a preponderance of the evidence that the merger of Oracle and PeopleSoft is likely substantially to lessen competition in a *1176 relevant product and geographic market in violation of 15 USC § 7, the court directs the entry of judgment against plaintiffs and in favor of defendant Oracle Corporation.
This order is stayed 10 days to permit plaintiffs to apply for appellate remedies.
IT IS SO ORDERED.
Related
Cite This Page — Counsel Stack
331 F. Supp. 2d 1098, 2004 U.S. Dist. LEXIS 18063, 2004 WL 2006847, Counsel Stack Legal Research, https://law.counselstack.com/opinion/united-states-v-oracle-corp-cand-2004.