No. 86-1033.United States Court of Appeals, Third Circuit.Argued August 19, 1986.
Decided December 31, 1986.
Richard H. Compere (argued), Pamela J. Johnson, William Brinks Olds Hofer Gilson Lione, Ltd., Chicago, Ill., Jerome R. Richter, James R. Kahn, Blank, Rome, Comisky McCauley, Philadelphia, Pa., for appellant.
Arthur H. Seidel (argued), Kathleen J. Gallagher, Seidel, Gonda, Goldhammer Abbott, P.C., Hal A. Barrow, Garfinkel
Volpicelli, Philadelphia, Pa., for appellees.
Appeal from the United States District Court for the Eastern District of Pennsylvania.
Before BECKER, MANSMANN, Circuit Judges and TEITELBAUM, District Judge[*]
 OPINION OF THE COURT
BECKER, Circuit Judge.
 This appeal concerns trademark rights in the phrase “Diet Chocolate Fudge Soda,” the name of a drink with which Plaintiff-Appellant, A.J. Canfield Co. (“Canfield”), has recently achieved meteoric success. In the district court, Canfield sought a preliminary injunction barring Defendant Concord Beverage Co. (“Concord”) from using this name in connection with a soft drink Concord sells in Pennsylvania, New Jersey, and New York. The district court refused to issue the injunction, A.J. Canfield Co. v. Concord Beverage Co., 629 F. Supp. 200 (E.D.Pa. 1985), and Canfield appeals. 28 U.S.C. § 1292(a)(1).
 We must decide whether the designation “chocolate fudge” as applied to diet soda deserves protection as a trademark under the federal law of unfair competition. 15 U.S.C. § 1125(a) (Section 43(a) of the Lanham Act). The term is protectable if it is suggestive or if it is descriptive and is buttressed by proof of secondary meaning. The designation is unprotectable if it is generic or if it is descriptive but lacks sufficient secondary meaning. If the mark is protectable, we must also decide its geographic reach.
 At the present stage of the case, we hold the designation “chocolate fudge” for diet soda generic, hence unprotectable, on the basis of two essential considerations. First, as the district court justifiably found, the term “chocolate fudge” denotes a particular full and rich chocolate flavor, that distinguishes Canfield’s product from plain chocolate sodas. Second, on the basis of the present record, it appears difficult if not impossible for a competing producer to convey to the public that its product shares this functional flavor characteristic without using the words “chocolate fudge.”
 Unlike Athena, however, this conclusion does not emerge full grown from the existing jurisprudence. Indicative of this legal lacuna is the fact that three distinguished district judges have considered the level of inherent distinctiveness of chocolate fudge soda and have arrived at disparate conclusions. See A.J. Canfield Co. v. Vess Beverages, Inc., 612 F. Supp. 1081 (N.D.Ill. 1985) (Shadur, J.), aff’d, 796 F.2d 903 (7th Cir. 1986) (finding “a reasonable basis for considering term suggestive although even greater basis for holding it merely descriptive”); Canfield v. Concord Beverages Co. (Scirica, J.) (holding term descriptive) Yoo-Hoo Chocolate Beverages Corp. v. A.J. Canfield Co., 229 U.S.P.Q. 653 (D.N.J. 1986) (Sarokin, J.) (finding term generic) appeal stayed per stipulation, App. No. 86-5584 (3d Cir. Oct. 17, 1986).
 The jurisprudence of genericness revolves around the primary significance test, which inquires whether the primary significance of a term in the minds of the
consuming public is the product or the producer. The primary significance test, however, cannot be applied until after we have decided the question that lies at the core of this case: whether the relevant product category or genus for purposes of evaluating genericness is chocolate soda or chocolate fudge soda. In order to resolve this initial question, we must develop a test that is congruent with general principles of trademark law and the primary significance test. We conclude that when a producer introduces a product that differs from an established product class in a significant, functional characteristic, and uses the common descriptive term of that characteristic as its name, that new product becomes its own genus, and the term denoting the genus becomes generic if there is no commonly used alternative that effectively communicates the same functional information.
 Because, on the present record, Chocolate Fudge Soda appears to fit our test for a generic term, and because generic designations are not protectable under trademark law, we will affirm the order of the district court denying the preliminary injunction.
 I. FACTS AND PROCEDURAL HISTORY
 Canfield, a medium-sized bottler located in the Chicago area, has produced “Canfield’s Diet Chocolate Fudge Soda” since the early 1970’s and until recently has made sales almost exclusively in the Midwest. On January 13, 1985, however, Bob Greene, a nationally syndicated columnist of the Chicago Tribune, wrote an article about the product entitled “Heaven For Dieters: Two Calorie Fudge Soda” that dramatically expanded the market for Canfield’s diet chocolate fudge soda. “To say it tastes like chocolate is to insult it,” wrote Greene. “Taking a sip is like biting into a hot fudge sundae. . . .” The column went on to explain that Canfield had combined “the proper combination of chemicals to precisely duplicate the taste of chocolate fudge without the use of chocolate,” so that the soda helped Greene satisfy his craving for chocolate and yet maintain his diet.
 The column set off a splurge of national publicity, including reports on television, by CNN News Broadcast, CBS Morning News, Good Morning America, and PM Magazine, and report in print media by The New York Times, Time Magazine, The Philadelphia Daily News, The Washington Post, and People Magazine. Exploiting this publicity, Canfield increased its advertising and, by mid-February, had begun shipments to various East Coast markets. The publicity encouraged bottlers around the country to seek permission to produce Canfield’s soda, and Canfield eventually granted eleven licenses, including a few licenses to bottlers in Concord’s marketing areas. One of those licensees, located in the Philadelphia area, began to distribute Canfield’s soda sometime in late May 1985, selling 50,000 cases in the first week. Exactly how much diet chocolate fudge soda Canfield sent to the central east coast before this time is unclear, but the amount is certainly small.
 Around this date and a decade after first marketing the soda, Canfield applied for the first time to the U.S. Patent and Trademark Office for protection of the designation “chocolate fudge” under the Lanham Act. That application was denied on May 23, 1985, on the decision of an examining attorney who found that “chocolate fudge” designates a flavor and as such is incapable of distinguishing the source of the goods. The examiner apparently considered chocolate fudge as applied to diet soda generic.
 Concord is a medium-sized bottler on the East Coast, producing both national “name” brands under license and a line of beverages under its housemark “Vintage.” Shortly after the Greene column appeared, and with the encouragement of its customers, Concord asked Canfield for a license to market diet chocolate fudge soda. The parties dispute whether Concord sought to license the entire name “Canfield’s Diet Chocolate Fudge Soda” or only the words “Chocolate Fudge Soda.” Failing to receive
a positive response, Concord quickly produced its own diet chocolate fudge soda, which it successfully marketed under its “Vintage” mark. The can design of the resulting soda strongly resembled Canfield’s design; on both cans the names Canfield’s or Vintage appear in large, diagonal, red letters and the term chocolate fudge soda appears in smaller, diagonal, brown letters. Concord began distributing Vintage Diet Chocolate Fudge Soda sometime late April or May, 1985, at approximately the same time that Canfield’s Philadelphia licensee began introducing Canfield’s.
 On August 30, 1985, Canfield commenced this suit in the United States District Court for the Eastern District of Pennsylvania claiming that Concord’s use of the phrase “diet chocolate fudge soda” constituted trademark infringement and unfair competition, in violation of state common law and § 43(a) of the Lanham Act, 15 U.S.C. § 1125(a), which governs unregistered trademarks. On September 3, 1985, Canfield moved for a preliminary injunction. After receiving evidence and hearing arguments, the district court filed a comprehensive and thoughtful opinion, denying the request for an injunction.
 The district court’s denial of the preliminary injunction flowed from its determination that Canfield had not demonstrated a likelihood of success on the merits. The court relied exclusively on federal law because Canfield had failed to pursue its state law claims in its briefs or at oral argument and because, the court noted, federal courts presume an identity of standards under federal and state law in the absence of any suggestion by the parties to the contrary.
 Analyzing the merits, the court first considered the designation’s level of inherent distinctiveness. It found the term “chocolate fudge soda” descriptive of the flavor of the soda but concluded that the term was not the name of the class of product, which the court considered to be diet chocolate sodas. Noting that a descriptive mark can acquire protected trademark status through the acquisition of “secondary meaning,” the court then considered whether “chocolate fudge soda” had acquired secondary meaning in Concord’s geographic market area. Relying on our decision in Natural Footwear, Ltd. v. Hart, Schaffner Marx, 760 F.2d 1383 (3d Cir.), cert denied, ___ U.S. ___, 106 S.Ct. 249, 88 L.Ed.2d 257 (1985), the court held that secondary meaning cannot exist in a market that plaintiff has not penetrated. Finding insufficient evidence of actual sales, growth trends, percentage of sales customers, or amount of product advertising to demonstrate market penetration, the court found that Canfield had not established secondary meaning in Concord’s area.
 The court then considered claims that Canfield had established secondary meaning in the term chocolate fudge soda in Chicago and thus could “prevent defendants from intentionally adopting the same or similar mark.” In its action for trademark infringement in Chicago against Vess Beverages, Inc., Canfield had gained a preliminary injunction based in part on its advertising and sales in the Midwest and in part on a study that purported to demonstrate identification of the term “chocolate fudge soda” with Canfield. The court assumed, without finding, that such secondary meaning existed in the Chicago area but still rejected Canfield’s claim. First,
the court rejected Canfield’s claim based on the “secondary meaning in the making” doctrine, holding that doctrine inimical to the concept of secondary meaning and rejected by all but a few New York courts whose only endorsement was in dicta.
Next, the court rejected the idea that Concord’s knowledge of Canfield’s designation could itself support an inference that secondary meaning had spread to Concord’s area. Finally, the court rejected claims based on the “Tea Rose-Rectanus” doctrine, holding that intentional copying in this context did not constitute a “bad faith” intention to trade on Canfield’s goodwill.
 In a final section of its opinion, the district court held that even if Canfield had proven a valid trademark, it would still have had to demonstrate a likelihood of confusion. The court dismissed any suggestion that the similar trade dress contributed to confusion, finding insufficient proof of market penetration to demonstrate confusion and finding defendant’s intentional copying “mitigated by the uniformity of packaging in the soft drink industry.” 629 F. Supp. at 213 n. 7. The court also found a dearth of evidence showing actual confusion. Accordingly, the court found that Canfield had not shown a likelihood of confusion.
 On appeal, Canfield renews its contention that “chocolate fudge soda” is suggestive or, if it is descriptive, that it at least has acquired trademark status through secondary meaning in the Chicago area. Canfield claims that the district court erred in requiring it to establish secondary meaning in Concord’s area because of a claimed rule of trademark law that a junior user’s knowledge of a senior user’s designation, even if descriptive, automatically entitles the senior use to enjoin the junior user from using the mark. Canfield also claims a more limited rule that Canfield need not establish secondary meaning in Concord’s market areas because (1) Canfield has a valid mark in Chicago, (2) Concord knew of the mark, and (3) Concord sells in areas of Canfield’s natural expansion. In support of these points, Canfield asserts that the district court erred in focusing on “bad faith” because [it contends] the relevant question is only whether the junior user knew of the senior user’s use of the mark.
 Canfield also claims that the district court applied the wrong tests to determine whether chocolate fudge soda had acquired secondary meaning in Concord’s area by focusing on evidence of market penetration. Canfield claims that the “[f]ame of a mark can extend beyond its actual selling market” and may constitute secondary meaning, and Canfield argues that we should infer secondary meaning from Concord’s intentional copying, from product requests by East Coast chain stores, from East Coast advertising, and from the phenomenal sales of 50,000 cases of chocolate fudge soda in the first week of sales in the Philadelphia area. Finally, Canfield disputes the district court’s suggestion that a finding of secondary meaning in the term chocolate fudge in Concord’s area would not itself demonstrate likelihood of confusion,
citing our decision in Interpace Corp. v. Lapp, Inc., 721 F.2d 460 (3d Cir. 1983).
 In response, Concord renews its claim that chocolate fudge soda is generic and therefore unprotectable. Concord otherwise supports the rulings and findings of the district court.
 Because we agree with Concord that the designation “chocolate fudge” is generic, we do not reach the claims that predominated in the district court and in the briefs of the parties.
 II. LEGAL BACKGROUND: ARBITRARY, SUGGESTIVE, DESCRIPTIVE,AND GENERIC MARKS AND THEIR LEGAL CONSEQUENCES
 Federal courts have long held that § 43(a) of the Lanham Act extends protection to unregistered trademarks on the principle that unlicensed use of a designation serving the function of a registered mark constitutes a false designation of origin and a false description or representation. A designation may only receive protection, however, if the public recognizes it as identifying the claimant’s “goods or services and distinguishing them from those of others.” 1 J. McCarthy, Trademarks and Unfair Competition § 15:1 at 657 (2d ed. 1984). Such identification depends, in the first instance, on a designation’s level of inherent distinctiveness, and for this purpose, courts have divided designations into four categories: arbitrary (or fanciful) terms, which bear “no logical or suggestive relation to the actual characteristics of the goods;” suggestive terms, which suggest rather than describe the characteristics of the goods; descriptive terms, which describe a characteristic or ingredient of the article to which it refers, and generic terms, which function as the common descriptive name of a product class Keebler Co. v. Rovira Biscuit Corp., 624 F.2d 366, 374 n. 8 (1st Cir. 1980).
 Courts and commentators have recognized the difficulties of distinguishing between suggestive, descriptive, and generic marks. One commentary has even called
the distinction between descriptive and generic “artificial and unworkable.” Folsom Teply, Trademarked Generic Words, 89 Yale L.J. 1323, 1351 (1980). Despite this difficulty, these distinctions are crucial. If we hold a term arbitrary or suggestive, we treat it as distinctive, and it automatically qualifies for trademark protection at least in those geographic and product areas in which the senior user applies it to its goods. Rovira Biscuit, 624 F.2d at 374 n. 8. If we hold a mark descriptive, a claimant can still establish trademark rights, but only if it proves that consumers identify the term with the claimant, for that identification proves secondary meaning Scott Paper Co. v. Scott’s Liquid Gold, Inc., 589 F.2d 1225, 1228 (3d Cir. 1978). The distinction between suggestive and descriptive may also dictate different standards for determining its scope of geographic protection. See 2 McCarthy, supra, §§ 26:11, at 307-08.
 Finally, if we hold a designation generic, it is never protectable because even complete “success . . . in securing public identification . . . cannot deprive competing manufacturers of the product of the right to call an article by its name.”Abercrombie Fitch Co. v. Hunting World, Inc., 537 F.2d 4, 9 (2d Cir. 1976). Therefore, evidence that a generic term is identified with one producer, indicative of a secondary meaning for a descriptive term, proves only what courts call “de facto” secondary meaning. Such evidence indicates only that the producer has dominated the market for a product (often because it has had a patent). See Kellogg Co. v. National Biscuit Co., 305 U.S. 111, 118, 59 S.Ct. 109, 113, 83 L.Ed. 73 (1938); Schulmerich Electronics, Inc. v. J.C. Deagan, Inc., 202 F.2d 772, 776-78 (C.C.P.A. 1953); J. Kohnstam Ltd. v. Louis Marx Co. Inc., 280 F.2d 437, 440 (C.C.P.A. 1960); Folsom Tepley, supra, at 1351 (same evidence can establish secondary meaning or de facto secondary meaning); 1 McCarthy, supra, § 12:15, at 432. Accordingly, although some courts have established a sliding scale, demanding stronger proof of secondary meaning as a designation becomes more descriptive, see American Heritage Life Ins. Co. v. Heritage Life Ins. Co., 494 F.2d 3, 11 (5th Cir. 1974), Aloe Creme Laboratories, Inc. v. Milsan, Inc., 423 F.2d 845, 849-50 (5th Cir. 1970), see generally 1 McCarthy supra, § 15:12, at 687-85, evidence of secondary meaning is not even relevant if the term is arbitrary, suggestive or generic See, e.g., Liquid Controls Corp. v. Liquid Control Corp., 802 F.2d 934, 938 n. 6 (7th Cir. 1986) (discounting study that indicated secondary understanding of “liquid control” as not relevant to whether term is generic).
 Because a nonregistered mark has no presumption of validity, the burden of proving chocolate fudge soda suggestive or descriptive, as Canfield claims, lies on Canfield. See Technical Publishing Co. v. Lebhar-Friedman, Inc., 729 F.2d 1136 (7th Cir. 1984); Anheuser Busch Inc. v. Stroh Brewery Co., 750 F.2d 631, 638 (8th Cir. 1984); Reese Publishing Co. v. Hampton Int’l Commodities, 620 F.2d 7, 11 (2d Cir. 1980).
 III. IS “CHOCOLATE FUDGE” SODA SUGGESTIVE?
 We first consider whether the appellation “chocolate fudge” for diet soda is suggestive. The district court followed the standard set out in Stix Products, Inc. v. United Merchants Manufacturers, Inc., 295 F. Supp. 479, 488 (S.D.N.Y. 1968), which correctly stated the distinction between suggestive and descriptive terms as follows:
A term is suggestive if it requires imagination, thought or perception to reach a conclusion as to the nature of goods. A term is descriptive if it forthwith conveys an immediate idea of the ingredients, qualities or characteristics of the goods.
 Accord 1 McCarthy, Trademarks and Unfair Competition § 11:21, at 491-92 (1984). Canfield’s argument that the term “diet chocolate fudge soda” is suggestive focuses
on the word fudge. Canfield claims that the addition of the word “fudge” to chocolate does not describe a difference in flavor but merely conveys a concept of texture. Fudge, Canfield reminds us, is made of sugar, butter, chocolate and cocoa, none of which are contained in diet soda. Because diet soda obviously has neither the ingredients nor the texture of chocolate fudge, Canfield argues that “chocolate fudge” communicates only that the sensation of drinking Canfield’s soda is similar in some way to the sensation of eating chocolate fudge. Because a fudge-like sensation is said to be difficult to imagine in a diet soda, Canfield claims the link requires imagination and is thus suggestive.
 As the district court found, this contention is unpersuasive in the first instance because chocolate fudge communicates a particular kind of chocolate taste, “a deep chocolate taste.” 629 F. Supp. at 208, 228 U.S.P.Q. at 484. The Greene article itself, which Canfield concedes sparked the demand for this soda, praised the soda because it “precisely duplicate[s] the taste of chocolate fudge.” There is evidence in the record from which the district court might conclude that Canfield developed the soda by explicitly trying to duplicate the “taste” of chocolate fudge, as the Greene article stated. This focus on the chocolate fudge taste was then echoed in many of the articles that followed the initial Greene column.
 The district court also found that many food products are designated by the terms “chocolate fudge.” Canfield argues that all but one of the examples adduced by Concord have the consistency and ingredients found in chocolate fudge, suggesting that we can infer that “fudge” refers to consistency not to flavor. But we note that many of these products also have plain chocolate versions. To demonstrate that the addition of the word fudge refers to consistency not flavor, Canfield must demonstrate that the chocolate fudge versions are more like chocolate fudge in consistency than their plain chocolate equivalents. Canfield offers no evidence on this point. Significantly, many of the products specifically indicate that their name refers only to taste, calling themselves “chocolate fudge flavor.” Furthermore, at least one of Concord’s examples, a diet chocolate fudge drink by Carnation, is not claimed to have anything like the consistency of chocolate fudge. The district court justifiably inferred that the use of the term fudge communicates a variation of chocolate flavor not consistency.
 Because chocolate fudge denotes a flavor, no imagination is required for a potential consumer to reach a conclusion about the nature of Canfield’s soda. In accordance with the accepted definitions, “chocolate fudge” as applied to diet soda cannot be suggestive.
 IV. IS CHOCOLATE FUDGE AS APPLIED TO DIET SODA DESCRIPTIVE OR GENERIC?
 The question remains whether the term “chocolate fudge” as applied to diet soda is descriptive because it only describes characteristics or functions of the product or whether it is so commonly descriptive of the name of the product that we should consider it generic. See Salton Inc. v. Cornwall Corp., 477 F. Supp. 975, 984-87 (1979). That in turn depends on how we define the relevant product category, or “genus.” See Park ‘N Fly, 105 S.Ct. at 661-62 (“A generic term is one that refers to the genus of which the particular product is a species.”). The district court found that chocolate fudge as applied to diet soda is not generic because “[t]here is no proof that the term has come to be understood as referring to the genus of all chocolate-flavored diet sodas.” 629 F. Supp. at 208 (citation omitted). This statement assumes, without any analysis, that the relevant genus is chocolate diet sodas, so that the term chocolate fudge provides a description of this particular chocolate soda. We do not consider this assumption obvious, for the relevant genus could be chocolate fudge soda, just as the relevant genus of similarly flavored ice creams could be chocolate fudge ice creams rather than chocolate ice
creams (with chocolate fudge a species). If that is the case, then Canfield has chosen the generic name of the product as the name of its brand and cannot gain trademark protection.
 Thus, a fundamental question in this case is whether chocolate soda or chocolate fudge soda is the relevant product genus for evaluating genericness. As we have explained above, this question precedes any examinations of secondary meaning. If chocolate fudge soda is the relevant product class, identification by the public of the term chocolate fudge soda with Canfield would prove only the obvious point that Canfield has been the sole producer of the product, and would not establish trademark status.
 We cannot, however, determine the relevant genus on the basis of abstract analysis, intuition or common sense alone. Rather, we first examine the principles and tests that Congress and the courts have developed for judging genericness to determine if any established method is available for distinguishing product brand from product class. We conclude, however, that neither of the two dominant principles of genericness, the primary significance test and its related test of consumer understanding, directly provide an answer to our question, for both tests become applicable only after we have determined the relevant genus. We therefore must develop our own rule out of basic principles of trademark law that is consistent with these tests.
 A. The Primary Significance Test
 The dominant principle for determining whether a term is generic is the primary significance test. This test has its origins in the Supreme Court case of Kellogg Co. v. National Biscuit Co., 305 U.S. 111, 59 S.Ct. 109, 83 L.Ed. 73 (1938), which concerned rights to the term “shredded wheat” after a patent held by defendant Nabisco had expired. Nabisco claimed that it remained entitled to exclusive use of the term “shredded wheat” because evidence established that the public identified the term with Nabisco and thus had established secondary meaning. The Supreme Court held to the contrary:
There is no basis here for applying the doctrine of secondary meaning. The evidence shows only that due to the long period in which the plaintiff or its predecessor was the only manufacturer of the product, many people have come to associate the product, and as a consequence the name by which the product is generally known, with the plaintiff’s factory at Niagara Falls. But to establish a trade name in the term `shredded wheat’ the plaintiff must show more than a subordinate meaning which applies to it. It must show that the primary significance of the term in the minds of the consuming public is not the product but the producer.
 Id. at 118 (emphasis supplied). Congress codified this holding in the Trademark Clarification Act of 1984, Pub.L. No. 98-620, Title I, § 102, 98 Stat. 3335 (1984) (codified at 15 U.S.C. § 1064), mandating that “the primary significance” test “shall be the test for determining whether the registered mark” has become generic. 15 U.S.C. § 1064.
 Although the primary significance test instructs us to inquire whether a term primarily signifies the product or the producer, a trademark need not identify source directly or explicitly. The Clarification Act endorsed the long-recognized anonymous source rule, Pub.L. 98-620, Title I, § 103(1) (codified at 15 U.S.C. § 1127), which recognizes that a term may function as an indicator of source and therefore as a valid trademark, even though consumers may not know the name of the manufacturer or producer of the product. The rule recognizes that “the primary significance” test is met even if the primary significance of a mark is not directly the producer but rather if “the primary significance of the mark to consumers . . . is to identify a product or service which emanates from a particular source, known or unknown,” for it still provides the assurance to the public “that the product is of uniform quality and performance.” S.Rep. No. 98-627, 98th Cong., 2d Sess., 5 (1984), reprinted in U.S. Code Cong. Ad.News, 5718, 5722 (hereinafter Senate Report; accord In re DC Comics, 689 F.2d 1042, 1054 (C.C.P.A. 1982) (Nies, J., concurring) (“a trademark is functioning to indicate `source’ when it identifies goods of a particular source”) (emphasis omitted); see generally 2 McCarthy, supra, at § 3:4, at 111-12 (trademark may indicate source or quality).
 In codifying the primary significance test, Congress has also recognized that trademarks need not indicate source exclusively, but may have a “dual function — that of identifying a product while at the same time indicating its source.” Senate Report at 5, U.S.Code Cong. Admin.News 1984 at 5722. Because “most firms, in fact, attempt to market and promote their products as unique in some way,” id. at 8, the public may understand that all goods with a certain term originate from the same producer and at the same time understand the term to indicate particular product characteristics (or a combination of characteristics) that distinguish the product from others. The Clarification Act therefore specifies, “A registered mark shall not be deemed to be the common descriptive name of goods or services solely because such mark is also used as a name of or to identify a unique product or service.” Pub.L. 98-620, Title I, §§ 102, 103 (codified at 15 U.S.C. §§ 1064, 1127).
 Notwithstanding the centrality of the primary significance test to the genericness doctrine, it does not enable us to determine the relevant product genus in this case because it only becomes applicable after we have distinguished product genus from product brand. Although a term may primarily signify source if it primarily signifies a product emanating from a single,
albeit anonymous, source, it does not primarily signify source if the product that emanates from a single source, e.g., shredded wheat, constitutes its own product genus. As the Senate Report on the Clarification Act states:
The important question is whether the primary significance of the term to the relevant consuming public is to identify a product which emanates from a single, albeit anonymous, source, or merely to identify the product itself. Of course, if the public primarily understands the term as identifying a product, rather than a product emanating from a particular, albeit anonymous, source, the term is generic.
 Senate Report at 8-9, U.S.Code Cong. Admin.News 1984 at 5725 (citing Kellogg).
 In other words, the primary significance test is generally satisfied if a term signifies a product that emanates from a single source, i.e., a product brand, but it is not satisfied if the product that emanates from a single source is not only a product brand but is also a product genus. The primary significance test does not, in and of itself, tell us how to differentiate a mere product brand from a product genus. That, however, is the crucial question in this case. Once that question is decided, the resulting question often decides itself.
 B. The Test of Consumer Understanding
 Although the primary significance test does not itself tell us how to distinguish a product brand from a product genus, some commentators have suggested that such a distinction can be made by direct examination of consumer understanding. See, e.g.,
Greenbaum, Ginsburg Weinberg, A Proposal for Evaluating Genericism After “Anti-Monopoly”, 73 Trademark Rep. 101, 117-22 (1984); Note, The Legislative Response to Anti-Monopoly: A Missed Opportunity to Clarify the Genericness Doctrine, 1985 U.Ill.L.Rev. 197. Congress has attributed “the classic test” of consumer understanding, see Senate Report at 3, U.S.Code Cong. Admin.News 1984, at 5719, to Bayer Co. v. United Drug Co.,
272 F. 505, 509 (S.D.N.Y. 1921), in which Judge Learned Hand stated: “The single question, as I view it, in all these [genericness] cases, is merely one of fact: What do the buyers understand by the word for whose use the parties are contending?” Some commentators suggest that we may determine this understanding and thereby determine whether the term signifies a product brand or a product genus, by directly asking members of the relevant public whether, in their view, a term is a brand name or a “common name,” i.e., the name of a product genus. See
Greenbaum, Ginsburg Weinberg, supra, at 118-22.
 If such a survey could determine the primary significance of “diet chocolate fudge soda,” we might give some credence to the study that Canfield commissioned for separate litigation in Chicago. See supra note 2. It asked surveytakers “what kind
of drink” chocolate fudge soda is or sounds like. (Emphasis supplied.) Only thirteen percent of those who knew of the soda mentioned the term “fudge” in their response. This result might suggest that “chocolate fudge soda” cannot be the kind of product, or product genus — although the variety of responses to this question also suggests that the respondents had little idea what the survey meant by the term “kind of drink.” If we believed in the
utility of such a survey, we might remand the case for a finding of fact about consumer understanding, or at least require that such a factual finding precede the issuance of any permanent injunction.
 In contrast to the advocates of this survey technique, we do not believe that a direct survey of public views can truly measure consumer understanding if a term identifies a product that arguably constitutes its own genus. As we have discussed above, generic marks signifying goods produced by only one manufacturer may function both as generic terms, signifying the product genus, and as brand names, indicating continuity of source. Faced with a mark like shredded wheat, the consumer has no reason to define it either as the name of a brand or as the name of a genus because the term functions most efficiently as both. Accordingly, a survey inquiring whether a designation like shredded wheat is a brand name or a product name forces respondents to make a false dichotomy. Because consumers will never have had a reason to consider the question before, such a survey might not elicit real attitudes but merely answers developed on the spot that would be highly susceptible to the influences of survey phraseology.
 Perhaps more significantly, directly surveying the public without first differentiating the product brand from the product genus — or at least without first offering a definition for the distinction — would not, in the context of this case, test the meaning of words to the public but would rather request a legal conclusion. A conscientious survey respondent, when asked whether a term is a brand name or product name, might ask, “Exactly what do you mean by product name and what do you mean by brand name?” Using the primary significance test and taking account of the anonymous source rule, we would be compelled to answer: “A brand name primarily signifies a product brand while a product name primarily signifies a product genus.” The respondent then asks: “How do I distinguish a product brand from a product genus?” If we seek to answer this question with the results of this survey, we can only answer in a circular fashion: “That question depends on the answers we obtain from you and other respondents.”
We simply cannot circumvent the requirement of defining the distinction between product brand and product genus by asking the public to tell us if a name is a brand name or the name of a product genus; we can only ignore it.
 Some commentators have suggested that, instead of providing a definition of product brand and product genus, a survey may simply provide examples of brand names and product names. See
Greenbaum, Ginsburg Weinberg, supra, at 118. We do not agree. Such a technique would presumably rest on the assumption that even though we, as courts, cannot
define the distinction between brand and genus names without first distinguishing between brand and genus categories, the public can. We can think of no reason to support this assumption. Such a technique might also rest on the assumption that a list of examples might convey the distinction between brand and genus categories. But examples would only do so where the distinction is obvious, a situation we do not face. Finally, such a technique would not eliminate the fact that members of the public have no reason to distinguish between brand and genus name when they contemplate a genus with only one brand.
 Although we agree that consumer understanding lies at the heart of any trademark evaluation, how it may aid courts in cases involving unique products is problematic. In Bayer, the parties agreed that the relevant product genus was pills of acetyl salicylic acid, a product manufactured by many companies. The court used consumer understanding to determine whether the designation of one producer, “aspirin,” had come to be understood by the public to refer to pills of acetyl salicylic acid produced by all other manufacturers. (It did, and the court held it generic.) In this case, however, determining the relevant product genus is our principal problem, for at least until recently no producers other than Canfield have manufactured a diet soda tasting like chocolate fudge. We therefore know that the public understands “diet chocolate fudge soda” to refer only to Canfield’s product. That fact alone, however, does not mean that the brand is not also its own genus and the brand name not also a generic term.
 Thus, we have come full circle. Although we recognize the importance of consumer understanding, and although we must develop a rule that uses it, the foregoing discussion demonstrates that a direct survey of the public alone furnishes us no more of a method for determining the relevant product genus, or of circumventing the need to make that determination, than that furnished by the primary significance test.
 C. A Rule Developed from Basic Principles
 Our analysis and research have revealed no established test for differentiating between
product brand and product genus. In order to resolve the issue we must therefore develop an analytical framework that is consistent with Congress’s demand that genericness reflect the “primary significance” principle and at the same time that accommodates Congress’s insistence that a term is not generic merely because it also identifies a unique product or service. We are not prepared to announce a test for all cases. We do believe, however, that a direct analysis of interests underlying trademark law and the doctrine of genericness leads to a result in this case.
 Trademarks provide a short-hand means of enabling the buyer to distinguish the goods of one producer from the goods of others (or even to distinguish the different lines of a producer’s goods from each other). By enabling buyers to rely on previous experience and the experience of others, trademarks thus reduce the cost of obtaining information about products. See generally, Folsom Teply, supra, at 1336. Buyers can be confident that a product with a particular name originates from the same source and possesses the same intangible qualities as goods previously purchased with the same name. From this essential function, trademarks produce a variety of benefits, as the Senate Report enumerates:
Among other things, trademarks (a) foster competition by enabling particular business entities to identify their goods or services and to distinguish them from those sold by others; (b) facilitate distribution by indicating that particular products or services emanate from a reliable though often anonymous source; (c) aid consumers in the selection process by denoting a level of quality relating to particular goods or services: (d) symbolize the reputation and good will of the owner, thereby motivating consumers to purchase or avoid certain trademarked products or services; and (e) protect the public from confusion or deception by enabling purchasers to identify and obtain desired goods or services.
 S.Rep. at 2, U.S.Code Cong. Admin.News 1984, at 5718, 5719 (quoting the Chairman of the United States Trademark Association’s Federal Legislation Committee). Courts and commentators try to capture all of these interests in the oft-repeated statement that trademarks serve to prevent consumer confusion. See, e.g., Scott Paper, 589 F.2d at 1228 (“The trademark laws exist not to `protect trademarks, but . . . to protect the consuming public from confusion, concomitantly protecting the trademark owner’s right to a non-confused public.'”); McCarthy, supra, at § 2:3 (“Today, the keystone of that portion of unfair competition law which relates to trademarks is the avoidance of a likelihood of confusion in the minds of the buying public.”).
 Underlying the genericness doctrine is the principle that some terms so directly signify the nature of the product that interests of competition demand that other producers be able to use them even if terms have or might become identified with a source and so acquire “de facto” secondary meaning. See, e.g., CES Pub. Corp. v. St. Regis Publications, Inc., 531 F.2d 11, 13 (2d Cir. 1975)(Friendly, J.) (“To allow trademark protection for generic terms, i.e., names which describe the genus of goods being sold, even when these have become identified with a first user, would grant the owner of the mark a monopoly, since a competitor could not describe his goods as what they are.”). Courts refuse to protect a generic term because competitors need it more to describe their goods than the claimed markholder needs it to distinguish its goods from others.
 In essence, this kind of genericness requires courts to balance different kinds of confusion. If only one manufacturer can use a designation for which there is no common alternative, buyers may be confused by their ignorance that other goods possess the characteristics they seek. If a particular designation is available to all, however, customers who believe that only one manufacturer makes a product with that name may be misled into buying the goods of others, at least in the short run, and may face disappointment because these alternatives differ in subtle characteristics
from the brand they wished to buy. See Altman, Callman Unfair Competition, Trademarks Monopolies § 18.03 (4th ed. 1983) (“The competitor’s right to use words of common speech is bounded by the limits of necessity. The need of competitors to use the word will be balanced against the need of the first claimant to that particular word.”) (hereinafter Callman). The doctrine of genericness reflects the Congressional determination that the interest in preventing the second kind of confusion does not normally justify creating the first, in large part because producers may identify their goods through the use of terms that are not necessary to competitors.
 The genericness doctrine prevents trademarks from serving as the substitutes for patents, and protects the public right to copy any non-patented, functional characteristic of a competitor’s product. See Sears, Roebuck Co. v. Stiffel Co., 376 U.S. 225, 230, 84 S.Ct. 784, 788, 11 L.Ed.2d 661 (1964) Compco Corp. v. Day-Brite Lighting, Inc., 376 U.S. 234, 84 S.Ct. 779, 11 L.Ed.2d 669 (1964). Trademark law seeks to provide a producer neither with a monopoly over a functional characteristic it has originated nor with a monopoly over a particularly effective marketing phrase. Instead the law grants a monopoly over a phrase only if and to the extent it is necessary to enable consumers to distinguish one producer’s goods from others and even then only if the grant of such a monopoly will not substantially disadvantage competitors by preventing them from describing the nature of their goods. See Ideal Toy Corp. v. Plawner Toy Mfg. Corp., 685 F.2d 78, 84 (3d Cir. 1982) (trademark law remedies must “avoid affording undeserved patent protection”). Accordingly, if a term is necessary to describe a product characteristic that a competitor has a right to copy, a producer may not effectively preempt competition by claiming that term as its own.
 This quick summary of the genericness doctrine, we think, illuminates the primary significance test and provides a framework for analyzing the distinction between product brand and product genus. Notwithstanding its acknowledgment of the legitimate role a trademark may have in identifying a product brand, the primary significance test provides that “to function as a trademark, a term must be . . . an indicator of source, sponsorship, approval or affiliation.” Senate Report, at 2, U.S.Code Cong. Admin.News 1984, at 5719. To the extent that a trademark also communicates functional characteristics, it does not function as a trademark. Thus, to be consistent with the primary significance test, whether a product brand with a name used by one producer constitutes its own genus must turn on the extent to which the brand name communicates functional characteristics that differentiate the brand from the products of other producers. In making these calculations, consumer understanding will determine the extent to which a term communicates functional characteristics and the significance of a term’s role in doing so because of a dearth or abundance of alternative terms that effectively communicate the same functional information.
 Although we do not believe we have derived a formula that will differentiate the product brand from the product genus in all cases, we believe that application of these general principles of trademark law to this case, which involves at best a descriptive word not a coined term, is straightforward. We hold as follows: If a producer introduces a product that differs from an established product class in a particular characteristic, and uses a common
descriptive term of that characteristic as the name of the product, then the product should be considered its own genus. Whether the term that identifies the product is generic then depends on the competitors’ need to use it. See CES Publishing Corp. v. St. Regis Publications, Inc., 531 F.2d 11, 15 (1975). At the least, if no commonly used alternative effectively communicates the same functional information, the term that denotes the product is generic. If we held otherwise, a grant of trademark status could effectively prevent a competitor from marketing a product with the same characteristic despite its right to do so under the patent laws.
 Although our reasoning differs in some respects, our rule is consistent with other cases that essentially have taken the view that some terms may be so descriptive that even when combined in a new way with another product name, the resulting combination is generic. In Devcon Corp. v. Woodhill Chemical Sales Corp., 455 F.2d 830, 832 (1st Cir. 1972), for example, the court considered whether the appellant could register the name “5 minute epoxy” for a new epoxy that set in five minutes. The court rejected the idea “that a party can seize upon one of the primary characteristics that make an unpatented commercial product marketable and preempt or limit competitor reference” by registering it. We believe the same principle applies to a common law mark even if the producer, as a sole manufacturer, could establish de facto secondary meaning in the mark. See United States Jaycees v. San Francisco Junior Chamber of Commerce, 513 F.2d 1226, 1230 (9th Cir. 1975) (Merrill, J., dissenting) (“One who first conceives a left-handed moustache cup has no title to the product name even though, through lack of competitive interest or otherwise, he is the only source of that amenity.”)
 Ces Publishing Corp. v. St. Regis Publications, Inc., 531 F.2d 11, 13-15 (2d Cir. 1975), and Technical Publishing Co. v. Lebhar-Friedman, Inc., 729 F.2d 1136 (7th Cir. 1984), also support our position. In these cases, the courts respectively considered the terms “Consumer Electronics Monthly” and “Software News” as applied to magazines. The courts found their terms generic essentially because the common understandings of “consumer electronics” and “software” when combined with the common understanding of the terms “monthly” and “news” directly communicated the nature of the products. As the Second Circuit wrote in reference to “Consumer Electronics Monthly”:
[I]t is hard to think of a name for a magazine, directed deliberately and effectively to industry personnel, which more accurately names the class of trade magazines within that industry than one which simply gives the name of the trade plus the word “Monthly.” . . . It would be
difficult indeed for other trade magazines to flourish and identify themselves to a relevant readership if they were forbidden to use the common name of the trade in their titles.
 CES Publishing, 531 F.2d at 14-15. In addition to these cases, a number of other judges and courts have found terms unprotectable because the direct manner in which they communicate functional information makes them so descriptive as to be generic.
 Our holding is also consistent with the jurisprudence governing product design, which permits protection of trade dress that has acquired secondary meaning only if the design feature is nonfunctional. See Freixenet, S.A. v. Admiral Wine Liquor Co., 731 F.2d 148, 151 (3d Cir. 1984). As this court stated recently in Standard Terry Mills, Inc. v. Shen Mfg. Co., 803 F.2d 778, 780-81 (3d Cir. 1986):
The functionality doctrine accommodates the twin purposes behind the Lanham Act. It protects the manufacturer (and the consumer) from the copying of those features that signify a product’s source (and quality) and encourages competition by preventing one manufacturer from acquiring a monopoly by attempting to trademark those features of a design essential to a successful product of that type.
 Just as one producer may not use unfair competition law to prevent others from copying functional characteristics of its product, it may not use that law to prevent others from describing those characteristics in its products.
 V. APPLYING THE TEST TO CHOCOLATE FUDGE SODA
 Because the district court did not apply the test we have just established, we are not bound by its finding that chocolate fudge as applied to diet soda is descriptive. See Inwood Laboratories v. Ives Laboratories, 456 U.S. 844, 855 n. 15, 102 S.Ct. 2182, 2189 n. 15, 72 L.Ed.2d 606 (1982). Although the failure of the district court to apply the correct legal standard should normally result in a remand, we need not remand if “the record permits only one resolution of the factual issue.” Pullman Standard v. Swint, 456 U.S. 273, 291-92, 102 S.Ct. 1781, 1792, 72 L.Ed.2d 66 (1982). At the present stage of this case, we believe that application of our holding to the present record plainly demonstrates that Canfield has not sustained its burden of proving that chocolate fudge, as applied to diet soda, is not generic. Canfield must make such a showing to demonstrate a likelihood of success on the merits that would justify a preliminary injunction in its
favor. Kershner v. Mazurkiewicz, 670 F.2d 440, 443 (3d Cir. 1982).
 Application of our rule to this case is straightforward. Canfield had the good fortune of originating a soda idea, a diet soda tasting like chocolate fudge, that proved eventually to have extraordinary appeal. Largely through the Bob Greene article, and through the press reports that followed it, abetted by Canfield’s own advertising, Canfield sparked a nationwide demand for its soda. All these forms of publicity have highlighted the functional difference between a soda tasting like chocolate fudge and a mere chocolate soda, an established product class, suggesting that much of the demand for this soda arises because of its flavor. The term “chocolate fudge” is a common descriptive explanation of that functional difference. Accordingly, the relevant product class is not diet chocolate sodas but diet sodas that taste like chocolate fudge.
 Having determined the relevant product genus, we must examine the need of Canfield’s competitors to use the term “chocolate fudge.” Often, this examination will require a factual analysis of alternative terms. Flavors, however, have unique characteristics, and we can imagine no term other than “chocolate fudge” that communicates the same functional information, namely, that this soda has the taste of chocolate fudge, a particular, full, rich chocolate taste. (The preliminary injunctive record reveals no alternative terms either.) Notwithstanding Canfield’s role in creating the demand for this kind of product, the trademark laws do not give it the right to be the sole producer of a diet soda tasting like chocolate fudge. Such a right could only come from the patent laws. Accordingly, we conclude on this record that the term “chocolate fudge” as applied to diet soda is generic and is available to all potential competitors.
 Any complication in this case arises only from Canfield’s claims that chocolate fudge is “incongruous” when combined with diet soda, (Appellant’s Br. at 39). Such a suggestion implies that no diet soda can taste like chocolate fudge so that chocolate fudge soda cannot be the name of the product genus. To the extent that Canfield claims that its soda does not taste like chocolate fudge because chocolate fudge does not denote a taste, we must uphold the district court’s finding to the contrary, for it is not clearly erroneous. Of course, the district court made those findings only in a preliminary injunction hearing, and Canfield is free to adduce additional evidence in its case on final hearing. To the extent Canfield claims that its soda simply does not have the taste of chocolate fudge, however, all such a fact would prove is that Canfield has misdescribed its product. Even if Canfield’s soda does not have the taste of chocolate fudge, Canfield does not have the right to prevent others from advertising their soda as having that taste.
 Because Canfield has not thus far made out a case that the designation chocolate fudge for diet soda is protectable, it has little likelihood of success on the merits. Accordingly, the judgment of the district court denying Canfield’s request for a preliminary injunction will be affirmed.
(7th Cir. 1976) (stating distinction between suggestive and descriptive often made on “intuitive basis” rather than “logical analysis susceptible of articulation”); Vision Center v. Opticks, Inc., 596 F.2d 111, 115 n. 11 (5th Cir. 1979) (noting difficulty of distinguishing between descriptive and generic) Abercrombie Fitch Co. v. Hunting World, Inc., 537 F.2d 4, 11 (2d Cir. 1976) (same). Compare American Aloe Corp. v. Aloe Creme Laboratories, Inc., 420 F.2d 1248, 1252-53 (7th Cir.) (“alo-” generic for cosmetics), cert denied, 398 U.S. 929, 90 S.Ct. 1820, 26 L.Ed.2d 91 (1970) with Aloe Creme Laboratories, Inc. v. Milsan, Inc., 423 F.2d 845, 848-49 (5th Cir.), cert denied, 398 U.S. 928, 90 S.Ct. 1818, 26 L.Ed.2d 90 (1970) (“alo-” descriptive of cosmetics).
Because of our doubts about this test, we will not be the first. In some situations, this test may be unfair to a manufacturer with strong brand loyalty. Consumers of a particular brand may consider it so superior to others that they are unwilling to shift to other brands despite a significant change in price. Under the cross-elasticity test, that brand would become its own product class, and other manufacturers could use its name. In other circumstances, we imagine that goods with clearly recognized independent product names might prove to be highly cross-elastic; we suspect, for example, that the demand for all sodas varies considerably with the price of others. Again, cross-elasticity would not supply a proper test. Accordingly, we are unwilling to endorse a cross-elasticity analysis unless and until we are shown a method of applying it suited to the subtleties of the trademark context.
Consumers will not benefit, however, if trademark law prevents competitors from using generic or descriptive terms to inform the public of the nature of their product. Were the first user of a generic or descriptive term, say “bicycle,” able to exclude later entrants from use of that term, the former would be able not only to identify itself as the maker of the bicycle and to capitalize on whatever good will it has built up — legitimate purposes of trademark protection — but also to impair the ability of competitors to describe their product as bicycles — a wholly counterproductive result so far as consumers are concerned.
American Cynamid v. Connaught Laboratories, Inc., 800 F.2d 306, 308 (2d Cir. 1986).
App. Bd. 1962) (“space electronics” generic for electronic equipment used for space navigation); see generally McCarthy supra, § 12:2(E) (“To be generic, a term need not relate directly to the name of the product, but may relate to some distinctive characteristic of the genus of products.”).