CLEVENGER, Circuit Judge.
In this appeal, NavCom Defense Electronics, Inc. (“NavCom”) challenges the denial by the Armed Services Board of Contract Appeals (“Board”) of its claims for equitable adjustment arising from a contract with the Navy. We affirm the Board’s decision on claims 1, 2, 3, 4, 6, 7, 8, 10, 11 and 13. On claim 9 regarding the Level of Repair Analysis, the Board’s findings are not supported by substantial evidence, so we reverse the denial of entitlement on that claim and remand for a determination of quantum.
I
This appeal concerns a contract between the government and NavCom for the production of IFF Radar Test Sets (“RTS”). “IFF,” an acronym for “Identification of Friend or Foe,” involves a radar system mainly used today in civil aviation traffic control. In IFF systems, a ground or airborne interrogator sends out a signal which is recognized by a transponder at the receiving end and followed by an identification signal response by the transponder.
By the early 1980’s, the nation’s RTS systems had become outdated. Thus, in 1982, the Navy contracted with NavCom’s predecessor in interest to design, develop, and test a functional engineering model for a new generation of IFF RTS. Due to their substantial experience in designing and producing military electronics hardware equipment, NavCom- and its predecessor in interest knew about the requirements of military research and development (“R&D”) environments and the needs of mass production for actual use.
As part of the R&D contract specifications, the Navy required NavCom to pre[899]*899pare and submit extensive test procedures for the development model. Although the Navy had reviewed and approved NavCom’s testing procedures, NavCom did not perform its tests as required by the specifications. When the Navy informed NavCom of the model’s failures to meet the specifications, the contractor asserted that the Navy had to accept the RTS models because the prototypes still met the test procedures. Since NavCom was a year late in delivering the models and the government’s budget for the project was depleted, the Navy decided to terminate the project. In closing the project, the Navy deleted certain tests and deemed that the models proved the feasibility of the RTS.
Although the R&D contract was terminated, the Navy decided to re-test those prototypes to learn from the models and refine the existing specifications before undertaking the procurement of production units. NavCom understood that its tests were not comprehensive and that the Navy would conduct further operational tests after the delivery of the prototypes. Based on those tests, the Navy concluded that NavCom’s prototypes were unacceptable as production units.
The Navy then prepared a new set of specifications for the production contract. NavCom was aware that the specifications for the R&D and production contracts would be different. Indeed, the Navy was constantly updating the specifications for the RTS, even during the course of the R&D contract. The government also shared those revisions with NavCom personnel both during the course of the R&D contract and after the Navy terminated the contract. In 1987, Navy personnel visited NavCom to acquaint it with the new production specifications because the government planned an ambitious schedule for the upcoming production contract.
The government approved the procurement of production RTS units in March 1987, and issued a request for proposal the following December. The request for proposal included the new production specifications and required that each bidder address in its proposal the specific changes to the R&D units. In particular, the new production specifications required that the contractor perform a battery of tests on the first RTS and the first six accompanying Analog Controller Multiplexers (“ACM”) that were manufactured.
NavCom viewed the new specifications simply as revisions to the R&D specifications and based its proposal on that assumption. Hence, NavCom believed that it could use essentially the same test procedures used for the R&D contract and underestimated the technical risks in the production contract.
Problems and delays occurred during the performance of the production contract, and NavCom submitted a request for equitable adjustment in 1995. NavCom asked for $11,338,676 under thirteen claims. After negotiation broke down, NavCom submitted a certified claim for $13,435,058 in 1997, but the contracting officer declined to respond. On appeal, the Board denied all disputed claims. In re NavCom Defense Elecs. Inc., 01-2BCA ¶ 31,546, 2001 WL 865388 (ASCBA July 25, 2001). In its appeal to this court, NavCom only challenges ten of the entitlement decisions and one quantum issue. We have jurisdiction under 28 U.S.C. § 1295(a)(10).
We review the Board’s legal determinations de novo, Ingalls Shipbuilding, Inc. v. Dalton, 119 F.3d 972, 975 (Fed.Cir.1997), but will not disturb the Board’s factual findings unless the appellant can show that the findings are arbitrary, capricious, so erroneous as to necessarily imply bad faith, or unsupported by substantial evi[900]*900dence. Cessna Aircraft Co. v. Dalton, 126 F.3d 1442, 1446 (Fed.Cir.1997).
II
A. The Constructive Change Doctrine
NavCom bases all of its claims for equitable adjustment on the doctrine of constructive change. A constructive change occurs where a contractor performs work beyond the contract requirements, without a formal order under the Changes Clause, either due to an informal order from, or through the fault of, the government. See Ets-Hokin Corp. v. United States, 190 Ct.Cl. 668, 420 F.2d 716, 720 (1970); Len Co. & Assocs. v. United States, 181 Ct.Cl. 29, 385 F.2d 438, 443 (1967). As the Court of Claims explained:
where a contract contains the standard “changes” provision and the contracting officer, without issuing a formal change order, requires the contractor to perform work or to utilize materials which the contractor regards as being beyond the requirements of the pertinent specifications or drawings, the contractor may elect to treat the contracting officer’s directive as a constructive change order and prosecute a claim for an equitable adjustment under the “changes” provision of the contract.
Ets-Hokin, 420 F.2d at 720.
“Where it requires a constructive change in a contract, the Government must fairly compensate the contractor for the costs of the change.” Aydin Corp. (West) v. Widnall, 61 F.3d 1571, 1577 (Fed.Cir. 1995) (citing J.B. Williams Co. v. United States, 196 Ct.Cl. 491, 450 F.2d 1379, 1394 (1971)). Like our predecessor court, we have permitted equitable adjustment recovery based on the doctrine of constructive change. E.g., Aydin, 61 F.3d at 1577-78; John C. Grimberg Co., Inc. v. United States, 869 F.2d 1475, 1478 (Fed.Cir.1989).
Before it can recover, the contractor must show that the government ordered it to perform the additional work. Len Co., 385 F.2d at 443. The contractor cannot merely show that the government disapproved a mode of performance. Singer Co., Librascope Div. v. United States, 215 Ct.Cl. 281, 568 F.2d 695, 701 (1977). Rather, the contractor must show that the government actually compelled the additional work. Len Co. 385 F.2d at 443. The government’s order need not be formal or in writing. Wm. A. Smith Contracting Co. v. United States, 188 Ct.Cl. 1062, 412 F.2d 1325, 1340 (1969).
Of course, not all the costs for additional work are recoverable under this doctrine. The government generally has the right to insist on performance in strict compliance with the contract specifications and may require a contractor to correct nonconforming work. S.S. Silberblatt, Inc. v. United States, 193 Ct.Cl. 269, 433 F.2d 1314, 1323 (1970). The additional work or use of material must be “beyond the requirements of the pertinent specifications or drawings.” Ets-Hokin, 420 F.2d at 720. At the same time, the additional work performed by the contractor cannot be beyond the general scope of the contract. Embassy Moving & Storage Co. v. United States, 191 Ct.Cl. 537, 424 F.2d 602, 606-07 (1970). Drastic modifications or fundamental alterations ordered by the government beyond the scope of the contract would constitute a breach of contract. Id. The additional work must therefore be beyond the requirements of the contract, albeit still within the general scope of the contract.
In this case, NavCom bases its claims for recovery on allegations that it had to perform certain Navy-ordered activities that were beyond what the contract required. Normally, to determine whether a constructive change occurred, this court [901]*901consults the contract language. Aydin, 61 F.3d at 1577. NavCom, however, contends that we need not examine the contract, arguing that its interpretations of the contract were reasonable and thus binding on the government. Although we have previously applied that rule, see C. Sanchez & Son, Inc. v. United States, 6 F.3d 1539, 1543-44 (Fed.Cir.1993), we did so in the context of a de novo review of a summary judgment, where we were required to draw all factual inferences in favor of the contractor. Id. at 1541. In this case, except for the claim related to the level of repair analysis, the Board’s factual findings, which are binding upon us, indicate that NavCom’s interpretations of the contract were not reasonable in light of the circumstances. We cannot disturb those findings unless the appellant can show that the findings are arbitrary, capricious, so erroneous as to necessarily imply bad faith, or unsupported by substantial evidence. Cessna Aircraft, 126 F.3d at 1446. There is no evidence of bad faith or arbitrariness in the very comprehensive and carefully written opinion by the Board. Rather, there is ample evidence to support that decision and its findings that NavCom’s interpretations were not reasonable. Hence, where the resolution of the claim turns on whether the work was within the contract’s scope, we independently construe the agreement in accordance with its express terms and its plain meaning. Hol-Gar Mfg. Corp. v. United States, 169 Ct.Cl. 384, 351 F.2d 972, 974 (1965).
B. Claims Nos. 1, 3 & 7: Additional Testing of ACMs
NavCom claims that the Navy constructively changed the production contract by requiring additional testing of the ACMs. There is no dispute that the testing occurred at the Navy’s direction. The parties only disagree on whether the work was beyond what the contract requires.
“[Tjhis dispute, turning as it does on an interpretation of contract provisions, involves a question of law which the court is free to answer independently of the Board’s decision.” HRH Constr. Corp. v. United States, 192 Ct.Cl. 912, 428 F.2d 1267, 1271 (1970). Nonetheless, in light of the Board’s specialized expertise with government contracts, we give its interpretation “careful consideration.” Interstate Gen. Gov’t Contractors, Inc. v. Stone, 980 F.2d 1433, 1434 (Fed.Cir.1992). In resolving questions of contract interpretations, we begin with the plain language of the agreement. Gould, Inc. v. United States, 935 F.2d 1271, 1274 (Fed.Cir.1991).
Section C of the production contract contains the contract’s Schedule and provides the production contract’s work statement. Item 0001 of that Schedule provides that:
The First Article Contractor Testing called for hereunder shall be performed in accordance with Attachment (1) equipment specification, “Military Specification Test Set, Radar, AN/UPM-O, MIL-T-24664(E C) dated 28 May 1986” and Attachment (2) “Changes to MILT-24664(EC) dated 29 April 1988” and the Section H special contract requirement A First Article Unit for this contract is defined as consisting of:
one (1) each AN/UPM() Test Set
one (1) each Interface Cable Assembly for AN/UPX-23 and AN/UPX-27
one (1) each Interface Cable Assembly for AN/APX-76 A & B(V)
one (1) each Interface Cable Assembly for RT-8590/APX-72 and RT-727, RT-728 and RT-731/APX64
one (1) each Interface Cable Assembly for RT-1157/APX-100
one (1) each Interface Cable Assembly for RT-1284, RT-1285, RT-128, RT-1296
[902]*902and RT-1426/APX-100
one (1) each Interface Cable Assembly for RT-1063B/APX-101
one (1) each Interface Cable Assembly for KY-532/ASQ and KY533A/ASQ
The Interface Cable Assembly units listed in this contract are synonymous with the disputed ACMs. The Schedule thus provides for certain tests to be performed in accordance with attached specification MIL-T-24664 and its 1988 changes. In the attached MIL-T-24664 specification, ¶ 4.3 lists “all examinations and testing necessary to determine compliance with the requirements of this specification” and refers to Table III for the specific tests applicable to each article.
Table III, in turn, refers to different paragraphs in the specification that states how the tests were to be run. Thus, Table III instructs that performance of the electromagnetic interference (“EMI”) tests should occur in accordance with ¶ 4.5.8, which states that “equipment shall be subjected to an EMI compatibility test in accordance with the EMI test paragraph of MIL-T-28800.” MIL-T-28800 refers to a military specification dated December 1981 that states the general requirements for testing of electrical and electronic equipments. Similarly, Table III refers to high impact shock tests that should take place consistent with ¶ 4.5.5 “in accordance with the examination and test methods specified in MIL-T-28800.” Likewise, maintainability testing to determine the operator’s ease of troubleshooting is included in Table III and subject to 114.7, which provides for equipment demonstrations and simulated malfunction diagnostics.
Giving effect to the plain language of the contract, we determine that the agreement requires that the ACMs be subjected to EMI, shock, and maintainability tests. The Schedule expressly defines “First Article Units” as encompassing six ACMs and explicitly provides that those defined First Articles be tested in accordance with the attached specification MIL-T-24664 and its 1988 changes. The specification’s Table III unambiguously provides that those first articles be subjected to specific tests, including EMI, shock, and maintainability examinations. In other words, the contract contemplated the testing of ACMs for EMI, shock, and maintainability. In thus construing the contract as a whole to give reasonable meaning to all of its parts, we give effect to the plain language of the agreement as intended by the parties. See Granite Constr. Co. v. United States, 962 F.2d 998, 1003 (Fed.Cir.1992). Accordingly, because the contract required those tests, the Navy was within its right to demand performance of the agreement and did not effect a constructive change of the contract. Id. at 1006-07.
Despite the contract’s language, NavCom contends that the ACM is an “accessory” rather than “equipment.” This difference is crucial, according to NavCom, because the specification only requires that “equipment” should undergo those environmental tests, while “accessories” only need a superficial preoperational inspection. We are, however, unpersuaded. First, the plain language of the specification does not list ACMs, or the synonymous Interface Cable Assemblies (“ICA”), as an accessory. Although the MIL-T-24664 specification mentions ICAs in ¶ 3.4.1.5.1 entitled “[ajccessory stowage,” the context indicates that the paragraph merely requires that there be sufficient storage space in the front panel of the RTS to store a few items, including the ACMs. Second, the statement in MIL-T-24664 ¶ 1.1 that the RTS is “hereinafter referred to as the equipment” does not necessarily define ACMs as “accessories.” According to NavCom’s reading of ¶ 1.1,
[903]*903the term “equipment” only denotes the RTS. But, the specification does not say so much. That paragraph only says that references or allusions to the RTS in the specification would occur in the form of the term “equipment.” Thus, “equipment” in MIL-T-24664 necessarily includes RTS, but does not exclude other items incorporated by other sections of the contract. Since the Schedule section clearly contemplates that the ACMs be tested under the First Article provisions, we decline to adopt NavCom’s narrow reading of ¶ 1.1 that would necessarily render Item 0001 in the Schedule superfluous and meaningless. Blake Constr. Co. v. United States, 220 Ct.Cl. 56, 597 F.2d 1357, 1359 (1981) (refusing to adopt an interpretation that renders part of the contract superfluous). Third, NavCom’s interpretation of the contract would contravene the ACM testing provisions in the Schedule, see Santa Fe Eng’rs, Inc. v. United States, 801 F.2d 379, 382-83 (Fed.Cir.1986) (stating that the court “will construe the agreement, to the extent it is fairly possible to do, so as not to eliminate the standard article or deprive it of most of its ordinary coverage” (citations omitted and emphasis removed)), or at best create a conflict where none needs to exist, Hol-Gar Mfg., 351 F.2d at 979. Fourth, the predispute course of conduct between the parties indicates an understanding that the ACMs were not “accessories” free of the required environmental tests. Indeed, the Board found that NavCom never submitted any ACMs as accessories, while tendering for accessory preoperational inspection all the items listed from ¶ 3.4.1.5.2 to ¶ 3.4.1.5.9. This course of conduct indicates that neither NavCom nor the Navy viewed the production contract as defining ACMs as “accessories.” See Julius Goldman’s Egg City v. United States, 697 F.2d 1051, 1058 (Fed. Cir.1983) (interpreting contract “in accordance with the parties’ understanding as shown by their conduct before the controversy”).
Consequently, the contract unambiguously required testing of the ACMs as part of the First Article Units examination, and the Navy properly demanded performance of the contract. The Board correctly denied equitable adjustment for claims 1, 3, and 7.
C. Claim No. A Frequency Measurement Accuracy
NavCom accuses the Navy of constructively changing the contract by requiring that the RTS be able to accurately measure pulsed frequency over the entire range from 12 to 1200 MHz over a wide range of temperatures. In NavCom’s view, the contract only requires the RTS to accurately measure frequency only at two to six frequency points between 12 and 1200 MHz at ambient temperature. Because this is a contract interpretation issue, we review this issue de novo, albeit with due respect to the Board’s reasoning. Grumman Data Sys. Corp. v. Dalton, 88 F.3d 990, 997 (Fed.Cir.1996).
The language at issue is ¶ 3.6.16.5.1 of the MIL-T-24664 specification, which governs the testing of RTS. That paragraph provides in relevant part that: “[t]he counter shall measure frequencies on the low power input jack from 12 MHz to 1200 MHz at power levels from 10 milliwatts (mW) CW and 10 mW pulsed to the lower limits of the MAIN and AUX jacks.” The parties simply dispute the meaning of this provision.
As always, contract interpretation begins with the plain language of the written agreement. McAbee Constr., Inc. v. United States, 97 F.3d 1431, 1435 (Fed.Cir. 1996). Here, the specification requires that the module be able to “measure frequencies ... from 12 MHz to 1200 MHz.” [904]*904The provision requires measurement capabilities at any frequency within the given range. The specification does not limit the measurements to only two to six fixed, predetermined frequencies as NavCom argues. “If we accepted [NavCom’s] argument, we would have to rewrite the contract, and insert words the parties never agreed to, which we do not have the authority to do.” George Hyman Constr. Co. v. United States, 832 F.2d 574, 581 (Fed. Cir.1987). We cannot do so. Therefore, we conclude that the Board correctly ruled on this claim.
D. Claim No. 2: RTS Power Module
On the claim related to the RTS power module, NavCom asserts that the Navy constructively changed the contract by requiring the contractor to revise and separate the combined Satisfactory Operating Test/Satisfactory Operating Check (“SOT/SOC”) test procedures. Those procedures ensure that the equipment is functional before commencing a test and that the RTS is accurate. We are not convinced by NavCom’s arguments.
When the Navy received the actual software deliverables during the first article testing of the production contract, it realized that the combined SOT/SOC test was merely a functional check of outputs, rather than the more complicated procedures NavCom promised. While NavCom assured the Navy that there was a signal, the Navy could not test whether the RTS had a signal that met the specification requirements. In fact, the SOT test could not meet the specification requirements even under the R&D contract, precluding NavCom from using that approach for the production contract. There is no evidence that the parties viewed the SOT/SOC provisions of the contracts any differently from their plain meaning, despite NavCom’s assertions to the contrary. The Navy was entitled to strict performance of its contract. S.S. Silberblatt, 433 F.2d at 1323.
As to NavCom’s argument regarding the meaning of “satisfactory operation” and the contractor’s broad discretion, that contention is based on an out of context reading of that term in ¶ 4.5.4 of MIL-T-24664. That paragraph actually states that the “parameters to be verified and limits to acceptability for the satisfactory operating test, as required by MIL-T-28800, shall be selected by the equipment contractor and be approved by the procuring activity” (emphasis added). The discretion to select the parameters is thus constrained by the MIL-T-28800 specification, negating NavCom’s alleged conflict. In this case, NavCom could not prove that its SOT test could even meet the requirements of MILT-28800 specification. There was no constructive change on this claim.
E. Claim No. 8: Channel-to-Channel Isolation
NavCom further contends that the Navy improperly insisted on the isolation accuracy of ±1.0 db in the “worst case scenario,” where one output channel is set at wide open maximum and the other channel set at absolute minimum. According to NavCom, that demand forced a major and costly redesign.
The parties do not dispute that the provision in question involved a performance requirement. Thus, NavCom had discretion to select the means to achieve that goal. See P.R. Burke Corp. v. United States, 277 F.3d 1346, 1357-58 (Fed.Cir. 2002). Discretion to select the means does not translate into unfettered discretion to change the goals. As part of the production specification, ¶ 3.6.17.2.5 requires that “the specified accuracy tolerance (±1 db) ... shall be maintained regardless of the modulation or attenuator setting of the [905]*905other channel.” The plain language of the contract requires that the accuracy level of ± 1 db must remain constant, regardless of the channel setting. In other words, the accuracy level must remain at ± 1 db, even when the channel outputs are at their extreme differentials.
The method that NavCom preferred would not have achieved the specified goal. NavCom proposed a software prompt on the RTS screen telling the operator to adjust his dial, thus avoiding the “worst case scenario.” By proposing that option, NavCom essentially conceded that, as designed, its equipment could not maintain the specified accuracy tolerance of ±1 db “regardless of the modulation or attenuator setting of the other channel,” especially when the channel outputs are set at then-extreme differentials. That failure falls short of satisfying the provision’s goal. The software fix would not have solved NavCom’s failure to comply with the specification; it would have only masked it.
The correct method required substantial isolation. In its proposal, NavCom determined that only 102 db of isolation was required. But, those calculations were erroneous because NavCom used a formula for incoherent signals rather than coherent signals. Had it used the correct formula, it would have recognized that it needed 120 db. Acknowledging that 100 db of isolation was not enough and that 120 db was essential, it nonetheless resisted the need to add the extra isolation because it was “outrageously difficult.” As the Board determined, the Navy had no reason to double-check NavCom’s calculations since it did not know what specific design NavCom would use. The Navy also had no reason to expect that NavCom intended to limit its proposal of “100 db of isolation” to no more than 100 db. Since NavCom’s interpretation was unreasonable and the Navy did not know about NavCom’s errors, we decline to hold the government hable for a contractor’s mistake. The redesign was within the scope of the contract.
F. Claim No. 6: 1030 CW Frequency Isolation
During first article testing with the 1030 CW output channel turned on (which was NavCom’s chosen default position), the two main output channels sometimes gave incorrect readings because of distortions from the 1030 CW oscillator. To avoid a technician running the tests with those distortions, the Navy requested that NavCom switch the 1030 CW signal on and off during the testing process. By testing the RTS both when the 1030 CW channel was on and off, the Navy could be sure that any interference would remain within the accuracy tolerance of ± 1 db.
NavCom contends that it was denied discretion in this performance requirement and that it reasonably interpreted the specification. We are not convinced. Although the specification is silent on whether the 1030 CW channel was supposed to be on or off, ¶ 3.6.17.2.5 nonetheless requires that “the specified accuracy tolerance (± 1 db) ... shall be maintained regardless of the modulation or attenuator setting of the other channel.” It is undisputed here that the RTS could not maintain the specified accuracy tolerance when the 1030 CW channel was turned on. The software prompt proposed by the contractor would only mask NavCom’s inability to satisfy the ± 1 db tolerance whenever the 1030 CW channel was on. Since the specification requires the maintenance of ± 1 db tolerance, the Navy was within its contractual rights to demand that NavCom comply with the specification regardless of whether the 1030 CW channel was on or off during the testing process.
[906]*906Similarly, we cannot accept as reasonable NavCom’s interpretation that the specification permitted it to overlook the ±l'db tolerance requirement if the 1030 CW channel is on. The specification did not say “the specified accuracy tolerance (±1 db) shall be maintained only when the 10S0 CW channel is on.” NavCom invites us to rewrite the contract to its advantage, but we cannot do so. See George Hyman Constr., 832 F.2d at 581. Therefore, the Navy’s demand for redesign fell within the scope of the contract; there was no constructive change on this claim.
G. Claim No. 9: Level of Repair Analysis
NavCom also contends that the Navy compelled it to perform a Level of Repair Analysis (“LORA”) in violation of the contract provision. We agree.
LORA is an economic model that recommends whether the Navy should repair or discard an item when it fails. Under normal contracting procedures, the government may decide to either run the LORA itself or delegate it to the contractor through the agreement. If the contractor performs the LORA, it must submit a LORA Report once the analysis is completed. If the government chooses to perform the LORA itself, the contractor must provide a LOR Input Report to enable the government to determine what data it should analyze. A LORA Report comes from the contractor after the analysis. Regardless of who performs the LORA, the contractor must submit a LOR plan in its proposal to describe “how the contractor will conduct the LOR program to fulfill the requirements of this” specification.
In this case, NavCom contends that the production agreement obligated the Navy to perform the LORA, and that NavCom’s performance of the LORA was beyond the scope of the contract and thus a constructive change. As an issue of contract interpretation, we decide that issue without deference to the Board’s construction of the contract. Barseback Kraft AB v. United States, 121 F.3d 1475, 1479 (Fed.Cir.1997).
The production contract contains on the same page two different items related to LORA. First, item C001 calls for a “LOR Analysis Report” in accordance with DIL-2085A. Provision DI-L-2085 provides that:
This [LORA] report is to advise the procuring activity of the results of the LOR analysis conducted. This report will document and support the contractor’s recommendations for most economical repair levels, repair versus discard at operational site, spare part provisioning, etc.
That item also states: “Submit 45 days following LOR analysis. The Government will respond within 30 days. Contractor to resubmit 30 days after receipt of Government comments.” Second, the production contract also includes item C002 which calls for a “Plan Program Level of Repair (LOR) Government Analysis” in accordance with DI-L-22332C (emphasis added). DI-L-22332C requires a plan to provide the necessary input data for the government to run a LORA. Thus, C001 provides for a report by the contractor once the Navy completes the LORA and provides it to NavCom. C002 unambiguously indicates that the Navy, not NavCom, would perform the LORA. Given the clear language of the agreement, we rule that the contract requires the Navy to perform the LORA. We thus agree with the Board’s ruling that “NavCom could reasonably interpret the contract to require the Government, not the contractor to perform LORA.”
We, however, disagree with the Board’s determination that NavCom cannot recover because its bid indicates that the con[907]*907tractor did not rely on that language. There is no evidence to support that determination. The Board took one phrase out of context in NavCom’s bid proposal to determine that NavCom agreed to perform the LORA. When considered in the proper context of the bid, the proposal indicates that NavCom expected the Navy to run the LORA and that it would simply have to prepare a report using the results of the LORA. The lack of hours estimated for C002 does not support the Board’s ruling either. Indeed, since C002 requires the government to perform the analysis, there is no reason for NavCom to request any funding for the item dealing with the LORA. Consequently, we conclude that there is no substantial evidence to support the Board’s denial of this claim. Rather, we hold that a constructive change occurred on this claim.
H. Claim No. 10: Technical Manuals
NavCom claims compensation for alleged constructive changes related to additional efforts to prepare technical manuals due to the LORA-related delays. The production contract required NavCom to produce technical manuals that the Navy would use to repair the RTS. Because the preparation of those manuals is dependent on the completion of the LORA, NavCom could not complete those manuals while it disputed the assignment of LORA responsibilities. But, even after the LORA dispute was resolved, NavCom continued to add to the delay by questioning the Navy’s decisions related to levels of repairs. Given the extra work and the existing delays, the parties entered into a bilateral modification of the contract where the Navy agreed to delete certain unrelated requirements in exchange for an expanded scope of work on the technical manuals. The bilateral modification compensated NavCom for the additional work it had to perform on the technical manuals.
The bilateral modification barred the contractor’s subsequent claim for compensation on the technical manuals. Fraass Surgical Mfg. Co. v. United States, 205 Ct.Cl. 585, 505 F.2d 707, 712 (1974) (“[T]his court has also held that a supplemental agreement entered into between a contractor and the government operates as a bar to all claims not specifically reserved.”). Under the law, NavCom cannot claim further equitable adjustment for a claim for which it received full compensation.
Instead, NavCom challenges the Board’s factual finding that it was compensated through the modification. We will not disturb the Board’s factual findings unless the appellant can show that the finding is arbitrary, capricious, so erroneous as to necessarily imply bad faith, or unsupported by substantial evidence. Cessna Aircraft Co., 126 F.3d at 1446. There is no showing of arbitrariness, capriciousness or fraud in this case. Rather, NavCom attempts to relitigate this claim on appeal by asserting that a contracting officer testified that the bilateral modification was “[ojnly talking about the depot level right in here.” But, the portion cited by NavCom does not support its claim. The witness did not negotiate the modification, and he was not sure that the modification pertained to the depot level in this contract. He only guessed. The Board, who had the opportunity to hear the entire testimony and to consider the credibility of the witness, determined that the modification did compensate NavCom for the additional work. NavCom has not shown that the Board’s determination was unsupported by substantial evidence. Accordingly, NavCom is not entitled to compensation for this claim.
I. Claim No. IS: Delay in Government Furnished Equipment
NavCom further claims that constructive changes arose from the Navy’s [908]*908delay in delivering government furnished equipment (“GFE”). According to NavCom, had the Navy delivered all the GFE on time, NavCom would have realized long before August 1989 that the universal ACM system it was designing was not feasible.
The production contract obligated the Navy to deliver twelve pieces of GFE not later than 45 days after the contract award. The Board found that the Navy had to deliver the GFE no later than March 20, 1989. Of the twelve pieces of GFE, three were delivered ahead of schedule while the remaining nine were late.
Instead of waiting to receive all the GFEs, NavCom started designing its ACM interface board. In designing that interface board, NavCom believed a single, all-purpose ACM design would suffice. However, it did not investigate whether that universal ACM design was technically feasible, despite having received three of the twelve GFEs early and having full access to any specialists or technical documents in the Navy. Ultimately, NavCom had to design five-instead of one-different interface boards for the ACMs. The additional designs, however, did not delay the first article testing provided by the contract.
NavCom has not shown that the Navy’s delivery delays caused the additional costs it incurred for the ACM design. See Wieman v. United States, 230 Ct.Cl. 563, 678 F.2d 207, 214-17 (1982). The Board found “NavCom’s universal ACM not to be a viable design” and that “NavCom did not have to have all of the GFE before it could determine that its universal ACM was not a viable design.” In making those findings, the Board indicated that NavCom failed to address the Navy’s contention that the universal board design was infeasible. The Board also gave substantial credence to a technical expert on IFF test sets who testified that the three GFEs along with manuals and the Navy’s technical support was all NavCom needed to realize the infeasibility of that concept. NavCom has not shown that the Navy’s delivery delays caused the additional design costs it incurred, and it failed to rebut the Navy’s evidence that the actual cause of the additional costs was NavCom’s mistaken assumption that a single, universal ACM was technically feasible. Consequently, we agree with the Board’s denial of this claim.
J. Claim No. 11: Denial of Damages
On NavCom’s claim of equitable adjustment based on specification defects, the Board determined that NavCom was entitled to recover on four out of thirty sub-claims. However, because NavCom could not provide adequate support for the claim’s quantum under a “jury verdict” approach, the Board denied recovery on those four sub-claims. In its appeal, NavCom assigns error to that denial based on an allegation that there was sufficient evidence to justify a “jury verdict” award and that the Board erred in permitting the Navy’s damages expert, Professor Kauffmann, to testify. We disagree.
Under a “jury verdict” method of recovery, a court may adopt a reasonable equitable adjustment after receiving evidence from the parties if: (1) clear proof of injury appears; (2) there is no more reliable method for computing damages; and (3) the evidence is sufficient for the court to make a fair and reasonable approximation of the damages. Bluebonnet Sav. Bank, F.S.B. v. United States, 266 F.3d 1348, 1357 (Fed.Cir.2001). Under this “jury verdict” approach, the claimant bears the burden of proving the fact and amount of loss with sufficient certainty. See Willems Indus., Inc. v. United States, 155 Ct.Cl. 360, 295 F.2d 822, 831 (1961).
[909]*909In this case, we agree with the Board that, even if NavCom could satisfy the first two prongs of the test, it could not provide sufficient evidence for the Board to make a fair and reasonable approximation of the damages. NavCom did not have any proper system to track the expenditures and costs incurred in this project. In its bid, NavCom promised to provide a management data system to control the project and provide supporting cost information. However, it never implemented that system. Nor did it issue charge numbers to keep track of out of scope work. As the Board found, “NavCom has not demonstrated a justifiable inability to substantiate the amount of its damages by direct and specific proof.”
Without any way to identify in and out of scope work and without the project management system it promised to implement, NavCom had to devise ways to estimate cost allocation to support its claims. The Board, however, rejected as too speculative the three methods NavCom used for this claim. Under one method, NavCom reviewed its engineering managers’ reports to find references to an issue determined to be out of scope work, matched them with time sheets and charged those hours to the out of scope work. The Board found that method inadequate because it overstated the out of scope claims. Under the second method, NavCom used a ratio of alleged in and out of scope rework for redesigns. The Board rejected that method because NavCom did not explain its methodology or prove adequate reliability. There was no assurance that the alleged out of scope work did not result from the contractor’s own cost-cutting measures, continuing in-scope design or design error corrections. Third, NavCom also tried to capture the support costs of its different departments for all the asserted claims as a “representative measurement of the amount of in-scope versus out of scope support effort.” The Board determined that NavCom did not establish that all support functions were necessarily required in all claims and did not prove that it was entitled to all of the alleged out of scope nonrecurring engineering costs of all the claims for equitable relief. Moreover, that third method indicates that “there was no original scope defined for support, and as a result, it is not possible to differentiate in and out of scope support activities.” The Board determined that NavCom had not substantiated any of the three methods it used for this claim, did not provide expert testimony to support their use, and had not rebutted or addressed the problems found by the Navy’s damages expert. In light of the record, we agree that NavCom cannot recover under the “jury verdict” method.
Furthermore, the Board did not err in admitting and considering the testimony of Professor Kauffmann, the Navy’s expert witness. We review the trial court’s decision to admit or exclude expert testimony for an abuse of discretion. See Gen. Elec. Co. v. Joiner, 522 U.S. 136, 141-42, 118 S.Ct. 512, 139 L.Ed.2d 508 (1997). NavCom considers Professor Kauffmann to be unqualified because he is allegedly an expert in cigarette manufacturing. However, as the Board found and as the record supports, Professor Kauffmann has an extensive background in engineering, engineering management, and cost estimating. He also has broad experience with these issues, especially in government contracts. Given that background and experience, the Board was well within its discretion in admitting Professor Kauffmann’s testimony.
Consequently, the Board did not err in denying recovery on these four sub-claims. NavCom did not prove that it was entitled to a “jury verdict” award or that the Board [910]*910erred in admitting Professor Kauffmann’s testimony.
rv
In sum, we affirm the Board’s determination on claims 1, 2, 3, 4, 6, 7, 8, 10, 11 and 13. We, however, determine that the Board’s factual determination on claim 9 regarding the Level of Repair Analysis is unsupported by substantial evidence. Its conclusion on that claim is also legally erroneous. Therefore, we remand claim 9 to the Board for a determination of that claim’s quantum.
COSTS
No costs.