GODBOLD, Senior Circuit Judge:
This is an appeal from a permanent injunction enjoining defendants from infringing a copyright of Warren Publishing. As a predicate for the injunction the district court denied a motion by the defendants for partial summary judgment on the infringement issue and granted a cross motion for partial summary judgment on that issue by Warren Publishing.1 We affirm.
[951]*951I. Background and Copyrightability
Warren compiles and publishes annually a printed directory known as the “Television and Cable Factbook,” which provides information on cable television systems throughout the country. The subject matter of this case is the 1988 issue of the Factbook, which contains a “Directory of Cable Systems” section and a “Group Ownership of .Cable Systems” section, together containing approximately 1,350 pages of information concerning 8,413 cable systems throughout the country and their owners.
The district court explained the format of the Factbook:
The directory of cable systems section arranges entries alphabetically by state, and, within each state, alphabetically by the name of the “lead” or “principal” community served by the particular cable system. The group ownership section contains a listing of selected information about entities owning and/or operating more than one cable system. These entities are called multiple system operators, or MSOs. The 1988 Factbook contained information on 8,413 cable systems. Information on each cable system and MSO entry is broken down into a uniform set of “data fields” which provide a specific piece of information about the system. The Fact-book uses the same pattern for each cable system entry. The Factbook’s format is based on identifying cable systems, then providing information on the cable systems through the use of specific groups of data fields.
Order, p. 2.
Warren has been publishing cable television information since 1948, constantly adding to its work systems and data fields of systems. It is not disputed that the Fact-book is a copyrighted work and is appropriately registered. The Factbook is, however, a compilation.
A ‘compilation’ is a work formed by the collection and assembling of preexisting matérials or of data that are selected, coordinated; or arranged in such a way that the resulting work as a whole constitutes an original work of authorship.
17 U.S.C. § 101 (1990) (Emphasis added.).
A ‘compilation’ results from a process of selecting, bringing together, organizing and arranging previously existing material of all kinds, regardless of whether the individual items in the material have been or ever could have been subject to copyright.
H.Rep. No. 1476, 94th Cong., 2d Sess. 57, reprinted in 1976 U.S.C.C.A.N. 5659, 5670.2
The preexisting materials in this case consist of information about cable television systems. A copyright of a compilation does not lead automatically to a conclusion that all materials therein are copyrighted. The owner must prove that the alleged infringer appropriated a protectable element of the compilation consisting of the owner’s original selection, coordination or arrangement. Feist Publications, Inc. v. Rural Tel. Serv. Co,, 499 U.S. 340, 111 S.Ct. 1282, 113 L.Ed.2d 358 (1991). The district court found, on the summary judgment record, that the Factbook had sufficient originality in its selection, coordination and arrangement of the data contained therein to be copyrighted, a conclusion that Microdos does not seriously contest. Warren contended that the elements of the compilation that were copyrighted and infringed were three: (1) the communities covered; (2) the selection, sequencing and arrangement of data fields; and (3) the content of the data fields. With respect to (2), the court found that the selection of data fields to be incorporated into the Factbook was obvious to persons in the industry and did not require the creativity and originality necessary to permit copyright protection. The arrangement of data fields was held to be creative and copyrighted but not infringed because not substantially appropriated by Microdos. With respect to (3), the content of data fields was found to be [952]*952merely facts thus not copyrightable. Warren has not cross appealed, so the findings with respect to data fields are not before us for review.
The district court described Warren’s system for selecting and presenting information on cable systems:
How one defines a ‘cable system’ will dictate the communities selected to represent those systems.
Warren Publishing uses a ‘principle [sic] community’ system in selecting and presenting its information on cable systems contained in its Factbook. A ‘cable system’ is defined as an entity composed as one or more communities that are offered the same, service by the same cable system owner at the same price. The principle [sic] community, used to represent the entire cable system, is then selected by contacting the cable operator to determine which community is considered the lead community within the cable system. Other communities within the same cable system are then listed under the principle [sic] community, not independently.
Order, p. 10. The court then went on to hold that only Warren’s selection of communities was sufficiently creatively original to be copyrightable.
In effect, Warren had admitted that the coordination and arrangement of the communities selected for coverage in the Fact-book was an obvious, mechanical, or routine task which required no creativity. Therefore, the coordination and arrangement of the communities selected is not copyrightable. However, Warren argues that the selection of those communities was creative and protectable because Warren uses a unique system in selecting the communities that will be represented in the Factbook. The court finds that Warren Publishing’s system of selecting communities is sufficiently creative and original to be copyrightable.3
3 ' This is not to say that the selection of cable systems would be copyrightable in all cases. Had Warren selected every cable system listed by the F.C.C., then there would not be sufficient originality in the "selection” to warrant copy-rightability.
* ‡ * # #
In this ease a choice was made as to which communities were to be listed.
Order, p. 11.
Cable system information compiled and arranged by various compilers in the industry is commonly organized alphabetically by state and then, for each state, alphabetically by the names of communities having cable service. Cable systems offering service may serve one geographical community (single-community service) or more than one geographical community (multiple-community service). A system will serve only a community, or communities, for which it has been granted a franchise(s) by one or more local governments. A list of geographical communities served will not, therefore, coincide with a listing of systems — one lists apples and the other oranges. Warren chose to present information on cable systems, each an entity served by the same owner offering the same service at the same price. The limits of the entity are not geographic but owner/price/serviee bounded. For the purpose of listing, Warren identified systems by the names of communities served. Theoretically, it might have listed systems by the names of operators listed alphabetically, but this would be of doubtful use. Rather Warren chose to list systems geographically by “lead” or “principal” communities served. More than half of the systems were single-community systems, and, for each of them, the lead or principal (hereinafter “principal”) community was the sole community served. For a multiple-community system it was necessary to determine a principal community to be listed.3
According to Warren’s evidence it made the determination of principal community by [953]*953questionnaires (45% of the cases) and contact with the operators (55% of the cases). It acknowledges that it used Federal Communication Commission data to update its files. Once a principal community was determined for a multiple-community system, all other communities served by the system were listed under the principal name, not independently.
The parties chose Illinois as a representative state, and 1988 as a representative year, for purposes of this case. The district court set out differences brought about by different listing methods. The FCC attempts to list alphabetically every geographical Illinois community having cable service. For 1988 it listed 724 communities served. Warren’s 1988 Factbook listed 406 as suitable for listing using its principal community concept.4 The Broadcast Yearbook, another industry source (its criteria were not spelled out by the court), listed 243 communities. The district court found that each source listed some communities not included in other works. The court cumulated the names of communities listed by all sources and found 808.
In Feist Publications, Inc., 499 U.S. 340, 111 S.Ct. 1282, the Supreme Court considered copyrightability of Rural’s white page telephone directory for several communities, the listings of which were copied by Feist in an area-wide directory. The Court held that Rural’s material did not meet minimum standards for copyright protection. Rural simply took raw data — name, town and telephone number — of éach person who had applied for, and was receiving, telephone service from it and listed it alphabetically by surname. The court recognized that this was “selection” of a sort but held that it lacked “the modicum of creativity necessary to transform mere assertion into copyrightable expression.” Id. at 361, 111 S.Ct. at 1296. The Court set out principles for copyrightability of factual compilations of preexisting factual materials. The material contained need not be protecta-ble. Id. at 347, 111 S.Ct. at 1289. The constitutional touchstone is originality in selection, coordination or arrangement of the preexisting materials. The level of creativity required is low, “some creative spark, no matter how crude, humble or obvious.” Id. at 344, 111 S.Ct. at 1287. The copyright is “thin.” Id, at 347, 111 S.Ct. at 1289. The “sweat-of-the-brow” doctrine was rejected. Id. at 351, 111 S.Ct. at 1291.
Feist has been accorded very narrow scope.
Most applications of Feist have recognized the circumscribed sphere to which its holding applies, ruling that it invalidates the copyright only in the most banal of works,64 such as the white pages of a phone book.64-1
Nimmer on Copyright, ¶ 3.04(B), p. 3-31 (footnote 63 omitted).
Bellsouth Adver. & Pub. Corp. v. Donnelley Info. Pub., Inc., 999 F.2d 1436 (11th Cir.1993) (en banc), cert. denied, — U.S. -, 114 S.Ct. 943, 127 L.Ed.2d 323 (1994), is this court’s post-Feist directory case. BAPCO, a subsidiary of BellSouth, published a “yellow pages” advertising directory for the Miami area, organized into an alphabetical list of business classifications. Donnelley, [954]*954the alleged infringer, obtained the name, address, phone number, and other data from each listing in the BAPCO directory. It entered this into its computer, published it in its directory, and prepared from it advertising leads. This court en banc reversed a summary judgment for BellSouth.
BellSouth is like Feist. BAPCO took its own data — telephone listings of its subscribers — and arranged it in an alphabetical listing of business types, with individual businesses listed in alphabetical order under the applicable headings. These acts of coordination and arrangement, like those of Warren, were — for a business telephone directory— “not only unoriginal but practically inevitable.” The court considered BAPCO’s asserted áets of selection — i.e., determining the geographical scope of its directory and setting the closing date after which no listing would be accepted, and various marketing techniques to generate listings. The court pointed out that any expression of facts fixed in a tangible medium of expression requires a closing date and, where appropriate, a geographical limit, and that marketing techniques were techniques for collection of facts. No creative originality in selection was present.
In contrast, Warren utilized not raw data from its files but an external universe of existing material drawn from the industry and not itself precisely contoured, and presented and listed in various forms by various compilers. It chose to present information on cable systems by listing cable systems. For that purpose it defined “system.” It selected a method for placing each system within the listing structure, i.e., it chose to identify each system by the geographical name of the principal community served, determined in the manner we have described for single-community systems and for multi-community systems.
It is obvious that Warren’s listings offer advantages to persons using, serving, or selling to cable operators, who can deal with a single-community system, or a multi-commu-nity system connected by service and operation, by identifying and dealing with the heart of the system.
Microdos makes a number of contentions directed to whether Warren’s selection process has originality. It says it lacks originality because publicly available information is contained in FCC records, reports and filings.
The FCC reports provide a solid basis for determining cable systems and communities served. The FCC reports are a snapshot of cable systems nation-wide at a fixed time. Cable operations are constantly changing. They are not static and the changes are reported by cable operators to the FCC.
Reply Brief, p. 3. Microdos also says that originality is lacking'because Warren contacts operators by questionnaires and telephone, to determine from each what respective community it considers to be its principal community. Microdos’s argument overlooks the nature of a compilation, which is copyrightable because it compiles preexisting material in an original manner. It overlooks that the underlying data in this case does not in itself reveal that listing will be by principal communities, and, except for single-community operations, it does not reveal the name of the principal community, which is central to Warren’s selection process.
Warren does not, as Microdos suggests, assert copyrightability of a mere list of historically established geographical names. No such claim is made. Except for a single-community system, a geographical designation of “X” is not used by Warren in its local lore/mapbook sense but as the designation of a community that may include A, B and C as well as geographical area X, whose name has been selected as a label and useful as an entry point for the researcher.
We discuss below under “Infringement” Microdos’s independent source defense in which it asserts that FCC data can be used to produce the information that Warren lists. That argument is also relevant to the claim that Warren’s work lacks originality. For the reasons we give below, it does not demonstrate lack of originality.
The merger doctrine is not a bar to copy-rightability in this case. It precludes copy-rightability where an idea can only be expressed in a limited number of ways. Here [955]*955the concern is copyrightability of a compilation embracing selection of items that might, and often are, selected and listed in many forms by authors of existing compilations. The “sweat-of-the-brow” doctrine is not utilized.
The district court did not err in holding that Warren’s system of selecting communities is sufficiently creative and original to be copyrightable.
II. Infringement
Microdos markets a computer software package called Cable Access, which, like the Faetbook, provides detailed information on the cable television industry. The district court described it:
The Cable Access software package is broken into three databases. The first database provides information on the individual cable systems. This database is referred to as ‘the systems database.’ The second database provides information on multiple system operators and is simply referred to as ‘the MSO database.’ The third database is a historical database which provides selected information on the cable industry from 1965 to the present. Only the systems database and the MSO database are alleged to be infringing upon plaintiffs copyright.
Defendant’s Cable Access software package comes presorted by state and city. The customer may rearrange the data in a format of its choosing. The customer may • construct searches of the database’s information on cable systems as required to fit its particular needs, as well as output the data to a hard copy in various formats, again to fit the specific needs of the customer.
Order, p. 3. Cable Access has been marketed in four versions which follow essentially the same format. Warren contends that all four versions infringe upon its copyright.
The test for infringement is whether there is substantial similarity between Microdos and the Warren selection of principal communities, the element of copyrightability in Warren’s compilation. There is strong evidence of copying by Microdos. Warren included in its Faetbook nine or ten fictitious cable system entries, created as decoys. The phony operator(s) appeared in the initial version of Cable Access. Microdos’s only explanation to this court is the statement that it has no explanation for how this happened.5
The district court found that Cable Access’s versions one to four are virtually identical to the 1988 Faetbook selection. The Faetbook listed 406 communities under its principal community concept. The first version of Cable Access listed 405 communities served, the second 394, the third 393, and the fourth 398. These were almost 1:1 matches with the Faetbook — 99.9% by version one, 97% by version two, 96.8% by version three, and 98% by version four.6
Faced with this commanding evidence of almost verbatim copying, Microdos asserted an independent creation defense, i.e., that the high correlation of listings was the result of its use of similar but independent community selection processes that produced similar results. The district court held that, in view of the high degree of correlation, Microdos was required to produce detailed evidence showing how it arrived at its listings and how the system worked to produce substantially similar results.
In its independent creation argument Mi-crodos sets out that FCC lists all cable operators and that operator that is part of a [956]*956multi-community operation is given a “system-iden” number that identifies it as part of a multi-community operation. Thus, it says that it grouped together all systems having the same system-iden code, thereby establishing a geographical area for the multiple system. It then selected a name for the area, “in most eases” the name of the “major community,” which was the community with the largest number of subscribers, or in some cases where the headend7 was located. This process of selecting a name produced what Microdos calls an “inherent selection.” It tells us in its brief (p. 27) that the name of the geographical area “may, however, differ from this initial selection.” Microdos made telephone calls to cable systems'which might reveal changes such as headend location and changes in franchise areas by adding or deleting communities served.
The district court found this description of independent source not satisfactory. Micro-dos did not clearly state how, in utilizing it, it determined the major community in each multi-community area, a determination that had to be made to convert FCC’s listing of communities served into listings that matched Warren’s principal community listings. We have the same difficulty as the district court had. At one point Microdos says that the “major community” is the community with the most subscribers, elsewhere that it is the most populous, elsewhere that in “most cases” it is where the headend is located. It asserts that FCC requires that cable systems report communities served by headends and, that as a matter of common sense, the headend will be the principal community. The district court voiced its dissatisfaction with the changeable, and changing, explanations of listing by headend community, most populous community, common sense community, and “most cases” as against “other cases” not clearly explained. It is self-evident that a single-community operation will be listed in both the Factbook and Cable Access. Additionally, some communities listed by FCC were assigned multiple system-iden codes, which made it unlikely that Mi-erodos’s asserted use of FCC data would produce consistent and reliable community listings.8
In its reply brief Microdos seems to describe another method that it says produced a list of roughly 390 communities. This method embraced using an FCC report “among other public documents and FCC reports,” which produced 458 “potential” systems, and this figure was then corrected for 57 errors and irregularities resolved by telephone calls to the FCC and to operators, and also by removal of ten systems listed under other states but serving Illinois communities, producing a final figure of roughly 390 systems “having an Illinois community as the name of the geographical area served by the cable systems.”
Like the district court, we are simply not informed by Microdos how its listings of communities served is almost a 1:1 match of Warren’s list of principal communities. The district court did not err in holding that Microdos did not establish a plausible and coherent method for arriving at its selection of communities and submitted insufficient evidence that relied on any principled criterion for community selection.
.The judgment granting an injunction is AFFIRMED.