WILL, Senior District Judge.
This is an appeal from the dismissal of a complaint of trademark infringement and the entry of a declaratory judgment for the accused infringer. For the reasons set forth herein, we reverse and remand to the district court for further action consistent herewith.
I. Facts.
The parties are corporations that provide training seminars to business employees either in-house or in public settings. Both appellant and appellee give the employees they are training written materials and provide an instructor, or train an employee to act as an instructor. The seminars last from one to five days. The parties provide both generic or customized training programs, the latter being modified to fit the customer’s particular desires for its employees. Despite these similarities, the parties appeal to different sectors of the corporate community. Appellant markets its teaching seminars to all types of corporations, while appellee restricts its market to manufacturers, especially those in the fabricating, machining and assembly of products areas.
Appellant, The Forum Corporation of North America, is the older of the two companies. It was founded in 1971 and is based in Boston, although it has operations in a number of cities, including Chicago. Appellant’s generic training programs are designed for managers and salespeople. An example is the program “Influence,” which “is designed to help people be more effective in getting things done through people who don’t report to them.” Trial transcript (“Tr.”) 29. Appellant’s annual sales totaled around $25 million for the 1987-88 fiscal year, a significant growth over its 1985 revenue of $20 million.
Appellee is a Milwaukee company founded in 1980, nine years after appellant’s [437]*437genesis. Its teaching seminars are designed for management-level employees of manufacturers. Appellee’s current annual sales were estimated at the time of trial to be around $2.5 million as compared to about $1.5 million in 1985.
Appellant’s business stationery, marketing and program material include the marks “The Forum Corporation,” “The Forum,” and/or “Forum.” Appellant’s Chief Executive Officer, testified that appellant has used the mark “The Forum Corporation” since 1971, the year when appellant created its “Exceptional Manager Program.” The U.S. Patent and Trademark Office has issued a federal registration to appellant for the composite mark “The Forum Corporation: Helping People Do a Better Job.” Appellant’s application for registration of the mark “The Forum Corporation” was still pending at the time this appeal was filed.
Appellee commonly uses the mark “The Forum Ltd.” in a composite with the slogan “Your Quality Circle for Management Education.” However, some of appellee’s materials do not include the composite mark. For instance, appellee markets to prospective clients by mailing out preview tapes for one of its programs which are marked only by the words “The Forum Ltd.” Also, a catalog in which defendant advertises, the Deltak catalog, lists the appellee by the simple mark, “The Forum Ltd.” Appel-lee’s employees answer the phone, “The Forum Ltd.,” and sometimes use that identifier in brochures after first giving appel-lee’s composite mark.
Both parties market their products through personal contacts: appellant employs a direct sales force and appellee uses independent sales representatives. Both rely on direct mailings. Only appellee does advertising in publications, choosing those which are intended to reach manufacturing management. Both attend trade shows, but with the exception of one show, the parties have not attended the same ones. This lack of overlap is the result, in part, of appellee’s choice to attend trade shows restricted primarily to manufacturing management.
Appellant produced evidence that its Fortune 1000 company clients do not usually follow a “highly planned, highly systematic process” in choosing a training program. Tr. 286. It also produced evidence that, for twenty to thirty percent of its customers, the amount a typical first-time buyer spends on appellant’s services is “a drop in the bucket.” Tr. 281. Finally, appellant provided testimony that most customers rely on word-of-mouth recommendations from other companies and do not spend much time deciding whether or not to purchase the service. Appellee, on the other hand, presented testimony that it takes an average of almost four months to conclude a sale after a potential client is first contacted.
Appellant brought instances of actual confusion between the parties to the attention of the trial judge. A common carrier misdelivered a package in 1982 in a hotel in Long Beach, California in which both parties were giving public seminars. At a 1983 trade show, people came to appellee’s booth looking for representatives of appellant. John Connellan, president of a company which trains executives to communicate, testified that he believed that the listing in the Deltak catalog related to appellant, rather than appellee. Appellee’s Chief Executive Officer Schultz testified that, based on his experience with his customers, there was no confusion between the parties’ products.
In 1986 Elaine Fuerst, an employee of Motorola, was told by her boss to “call Forum and get their ‘Influence.’ They have a program called ‘Influence.’ ” Fuerst looked through her files, which she had not yet organized since taking them over from her predecessor, called appellant’s number in Florida which was disconnected and then called appellee in Milwaukee. She told appellee’s representative that she was looking for “Influence,” but the representative told her she should consider their “Business Requirements Planning” program. Fuerst agreed to review the “Business Requirements” program, but also expected to receive “Influence.”
[438]*438Fuerst reviewed appellee’s program which she also showed to upper-level managers and then sent the materials to another training manager who was in charge of manufacturing and materials. She later received a copy of appellant’s annual report, which described the “Influence” program, and figured out that there were two separate companies with the word “forum” in their names. She spoke to her boss, a training manager, and to the training manager for the manufacturing and materials area and to other training personnel and found that none knew there were two separate companies.
II. Standard of Review.
We review findings of fact under the clearly erroneous standard and findings of law de novo. Fed.R.Civ.P. 52(a); Pullman-Standard v. Swint, 456 U.S. 273, 287, 102 S.Ct. 1781, 1789, 72 L.Ed.2d 66 (1982). We have stated a number of times that the trial court’s ultimate conclusion on the likelihood of confusion is a finding of fact. G. Heileman Brewing Co. v. Anheuser-Busch, Inc., 873 F.2d 985 (7th Cir.1989); Int’l Kennel Club v. Mighty Star, Inc., 846 F.2d 1079, 1087 (7th Cir.1988); Scandia Down Corp. v. Euroquilt, Inc., 772 F.2d 1423 (7th Cir.1985), cert. denied, 475 U.S. 1147, 106 S.Ct. 1801, 90 L.Ed.2d 346 (1986) (White, J., dissenting because of need to resolve conflict in circuits). Other circuits have treated the final resolution of the confusion issue as legal. See, e.g., Charles of the Ritz Group, Ltd. v. Quality King Distrib., 832 F.2d 1317, 1321-23 (2d Cir.1987) (citing Plus Products v. Plus Discount Foods, Inc., 722 F.2d 999, 1004-05 (2d Cir.1983)) (applying the Plus Products two-level standard: fact findings in applying the test are reviewed under the clearly erroneous standard, while the ultimate conclusion is legal and reviewable de novo); Alpha Indus., Inc. v. Alpha Steel Tube & Shapes, Inc., 616 F.2d 440 (9th Cir.1980) (two-level approach).1
We have considered the arguments in favor of treating the ultimate finding of confusion or no confusion as legal and rejected them. Scandia Down, 772 F.2d at 1428-29. However, we review the district court’s statement of the law de novo for legal error and its conclusions for signs that the court’s application of the law was infected with legal error, i.e., an erroneous general principle about the way the test should be applied. See Schwinn Bicycle Co. v. Ross Bicycles, Inc., 870 F.2d 1176, 1187-88 (7th Cir.1989) (district court’s statement that there is a presumption of likelihood of confusion when a trademark has secondary meaning was an error of law and its mechanical application of the test, without considering the totality of the circumstances, was an error of law which prevented the court from exerting its sound discretion). See also James Burrough, Ltd. v. Sign of Beefeater, Inc., 572 F.2d 574, 577 (7th Cir.1978) (reversal because district court placed excessive importance on certain factors of test); Chevron Chem. Co. v. Voluntary Purchasing Groups, Inc., 659 F.2d 695, 696 (5th Cir.1981), cert. denied, 457 U.S. 1126, 102 S.Ct. 2947, 73 L.Ed.2d 1342 (1982) (“district court applied an erroneous legal standard in finding no likelihood of confusion, thereby stripping that determination of the protection normally accorded fact findings”).
The standard of review of a district court’s classification of a term along the spectrum of trademark protection — that is, as generic, descriptive, suggestive or arbitrary — is also clearly erroneous, so that the classification “may only be set aside if after considering all the evidence we are ‘left with the definite and firm conviction that a mistake has been committed.’ ” Walt-West Enterprises v. Gannett Co., 695 F.2d 1050, 1058 (7th Cir.1982) (quoting United States v. United States Gypsum Co., 333 [439]*439U.S. 364, 395, 68 S.Ct. 525, 542, 92 L.Ed. 746 (1948)). We must, therefore, toe a line between reweighing the evidence and disregarding our responsibility to make sure that the district court’s trademark classification was based on correct legal standards and that the confusion test was correctly applied.
III. Analysis.
In order to prove a claim pursuant to 15 U.S.C. § 1125(a), a plaintiff must show (1) that its trademark may be protected and (2) that the relevant group of buyers is likely to confuse the alleged infringer’s products or services with those of plaintiff. Int’l Kennel Club, 846 F.2d at 1084. A trademark may be protected if it is more fanciful or arbitrary than it is descriptive of an entire genus, or of one characteristic of a product or service. Park’n Fly, Inc. v. Dollar Park and Fly, Inc., 469 U.S. 189, 193-4, 105 S.Ct. 658, 661, 83 L.Ed.2d 582 (1985); Henri’s Food Prod. Co. v. Tasty Snacks, Inc., 817 F.2d 1303, 1306 (7th Cir.1987). The district judge found that “forum” could be protected since it was neither generic nor descriptive, but was suggestive. Memor. Op. at 3-6. He then concluded, however, that there was no likelihood of confusion. Id. at 19-20.
A. Whether the trial court erred in finding no likelihood of confusion.
Appellant urges us to affirm the trial court’s classification of “forum” as suggestive, but reverse its finding that there is no likelihood of confusion. Appellee asks us to review the suggestive classification only if we reverse the court’s conclusion that there is no likelihood of confusion. We, therefore, consider the likelihood of confusion issue first.
The district court applied the seven factor likelihood of confusion test set out in McGraw-Edison v. Walt Disney Prod., 787 F.2d 1163, 1167-68 (7th Cir.1986):
The degree of similarity between the marks in appearance and suggestion; the similarity of the products for which the name is used; the area and manner of concurrent use; the degree of care likely to be exercised by consumers; the strength of the complainant’s mark; actual confusion; and an intent on the part of the alleged infringer to palm off his products as those of another.
1. Whether the trial court erred in finding no similarity of marks.
The court considered the promotional materials, letterhead and advertisements of the parties and concluded that the marks are not likely to cause confusion. Necessary to that conclusion were the court’s findings that appellant mostly used the simple marks, “The Forum” or “The Forum Corporation,” while appellee had consistently used a composite form which included the words “The Forum Ltd.,” a design comprised of “a four columned, Romanesque building facade, with an A-frame roof,” and the slogan “Your Quality Circle For Management Education.” Memor. Op. at 7-8. Based on these differences, the court wrote, “I do not believe that a consumer, presented with defendant[-appellee]’s mark, is likely to confuse it with plaintiff[-appel-lantj’s.” Id. at 9.
The district court noted the rule that the proper method of comparison is conducted “in light of what occurs in the marketplace, not in the courtroom....” Int’l Kennel Club, 846 F.2d at 1088. Appellant contends that, in spite of its accurate statement of the law, the district court failed to make a proper marketplace comparison, committing the very error that it attempted to avoid. Appellant complains of three errors in the comparison of marks: that the court ignored evidence that appellee used the simple mark, “The Forum Ltd.,” without its slogan in many instances, compared only the full written manifestations of the marks and mistakenly placed responsibility for preventing confusion on appellant. The record shows that appellee does use a shortened form of its mark in some instances. And the district court’s finding that appellee’s mark had “appeared consistently ... in its composite form in its advertisements, letterhead and promotional literature,” memor. op. at 9, may be an over[440]*440statement. Nevertheless, the evidence supports the trial court’s distinction between appellant's and appellee’s marks on the basis of the form most often used. However, the court’s conclusion that there was no likelihood of confusion does not necessarily follow from that distinction alone.
We considered the marketplace comparison of goods or services whose trademarks are not normally seen side by side in James Burrough, Ltd. v. Sign of the Beefeater, Inc., 540 F.2d 266 (7th Cir.1976), a case in which a gin distiller sought to enjoin the use by a restaurateur of a sign which looked similar to the distiller’s label. We wrote:
The sign must ... be evaluated on the basis of its likely effect upon consumers who do not have Distiller’s label before them, but who may have a general, vague, or even hazy, impression or recollection of Distiller’s marks such as may trigger a belief that a relation exists between the Distiller and the enterprise identified by the sign.
Id. at 275. Because the district court had not placed due importance on the dominant feature of the parties’ marks, we reversed and remanded. Id. at 276.
Appellant argues that a marketplace analysis in this case demanded that the district court consider that the market for management training depends on word-of-mouth name recognition. Appellant points out that its expert, Mr. Connellan, testified that consumers of training products say, “We’re using Forum.” Tr. 322. Other courts have noted explicitly the importance of the sound of a protected mark, see, e.g., E. Remy Martin & Co. v. Shaw-Ross Int’l Imports, 756 F.2d 1525, 1526 (11th Cir.1985); Grotrian, Helfferich, Schulz v. Steinway & Sons, 523 F.2d 1331, 1340 (2d Cir.1975), or of the dominant terms in a trademark, Century 21 Real Estate Corp. v. Sandlin, 846 F.2d 1175, 6 U.S.P.Q.2d 2034, 2036 (9th Cir.1988) (Century Investments & Realty found confusingly similar to Century 21 because of the dominance of the word “Century”). Any memorable feature of a mark, which of course includes the way it sounds, should be considered in analyzing likelihood of confusion. Henri’s Food Prod. v. Kraft, Inc., 717 F.2d 352, 356 (7th Cir.1983); Blumenfeld Dev. Corp. v. Carnival Cruise Lines, Inc., 669 F.Supp. 1297 (E.D.Pa.1987) (Carnival Club likely to be shortened and confused with Carnival).
In the present case, the trial judge wrote: Plaintiff[-appellant] urges that consumers are not likely to notice defendant[-ap-pelleej’s ‘trade dress’ (i.e., its design and slogan), but will instead focus on its name, and particularly on the word ‘Forum.’ It may be that ‘Forum’ is the salient portion of plaintiff’s mark and should be accorded greater weight in the similarity analysis. But its potential for causing confusion is largely plaintiff's responsibility. In many instances, plaintiff has chosen to present itself without reference to its corporate status or without its slogan and design, all very helpful source identifiers. Defendant should not suffer from plaintiff’s attempt to lay claim to condensed versions of its registered mark.
Memor. op. at 9.
Placing the responsibility for preventing confusion on appellant, who was the first user, is wrong as shown by our decision in Wesley-Jessen Div. v. Bausch & Lomb, Inc., 698 F.2d 862 (7th Cir.1983). We wrote there that "one entering a field already occupied by another has a duty to select a trademark that will avoid confusion.” Id. at 867. See also Harold F. Ritchie, Inc. v. Chesebrough-Pond’s, Inc., 281 F.2d 755, 760 (2d Cir.1960). It is the second user’s responsibility to avoid confusion in its choice of a trademark, and that responsibility must include choosing a mark whose salient portion would not likely be confused with a first user’s mark. See Saks & Co. v. TFM Indus., 5 U.S.P.Q.2d 1762, 1764 (T.T.A.B.1987) (registration of “Folio by Fire Islander” refused because of evidence that applicant’s president abbreviated the mark as “Folio” which would likely be confused with opposer’s mark, “Folio”).
[441]*441Appellee responds that appellant, as the first user, should be found responsible for the confusion, because appellant started out by using its full (or composite) mark, which included a slogan, and then “gradually moved away from use of its full name ... to an abbreviated name.” Brief of Defendant-Appellee at 22 (citing Tr. at 50, 64-65; Pltf. Exs. 1A-1F). Therefore, according to appellee, “Plaintiff[-appellant] should not be rewarded nor Defendant punished because of post-1980 changes in the way Plaintiff chose to identify itself.” Id. at 23.
Neither the facts nor the law, however, support appellee’s contention or the district court’s conclusion. First, the testimony to which appellee points is testimony that appellant changed from its use of the mark “The Forum Corporation of North America” to “The Forum Corporation” or simply “The Forum.” We found, in International Kennel that such a distinction made little difference. 846 F.2d at 1088. Considering appellant’s annual reports for 1980-1985 (Pltf. Exs. 1A-1F), we find that appellant used its slogan in 1980-82 and not on the 1983-85 reports. We cannot, however, say that such a change had substantial import, since only in 1982 was the slogan clearly readable while in 1980 and 1981 the slogan was imprinted into the paper and not easily noticed. In addition, as the appellant points out, it has used the simple form “The Forum” on other materials since 1971, well before the reports to which appellee points. Tr. 38, Pltf. Exs. 2C, 11. Finally, even if appellant had shortened the form of its trademark throughout the years, appellee would not be freed from its responsibility to avoid confusion since the salient and memorable feature of appellant’s mark is “Forum.”
It is also important to note that the trial judge focused his analysis on the parties’ printed matter and not the full range of ways that the parties communicate their marks to the public. These include the way appellee answers its phone as “The Forum Ltd.” or “The Forum,” the display of appellee’s logo at public and in-house seminars, and the likelihood that much of the parties’ business comes by word-of-mouth contacts. Especially in this context of two service (rather than manufacturing) companies, name recognition would reasonably come about through word-of-mouth.
The court erroneously placed the responsibility for a likelihood of confusion between the salient marks of the two parties on appellant, the first user. In addition, the court’s analysis indicates that it ignored the non-visual ways in which the parties communicated their marks. These legal errors led the court to conclude that the marks were dissimilar and unlikely to be confused.
2. Whether the trial court erred by failing to find an overlap in plaintiff’s and defendant’s markets.
Although the trial court found that “the products produced by these parties are similar enough that the public may likely attribute them to a single source,” memor. op. at 11, the court concluded that there was insignificant overlap between plaintiff’s and defendant’s markets so that confusion was unlikely. The court wrote: “I find it significant that the parties operate through different marketing channels and approach different types of business consumers.” Memor. op. at 14. The trial court reasoned that the parties do not advertise in the same publications or attend the same trade shows. Although the court recognized that the parties share some of the same clients, it wrote that they “have little, if any, contact with each other in terms of promotion and competition for clients.” Id. at 13.
Appellant argues that the evidence that the parties shared clients and that both attracted customers through direct mail campaigns shows that the district court erred. There was, in fact, evidence that Motorola received direct mail from both companies. Appellants do not limit their list of potential customers to non-manufacturers. The parties’ products are substantially similar, as the district court found. In addition, appellant points out that both parties relied on public relation efforts to market their products.
[442]*442The “area and manner of concurrent use” factor requires us to consider whether there is a relationship in use, promotion, distribution, or sales between the goods or services of the parties. In re Superior Outdoor Display, Inc., 478 F.2d 1388, 178 U.S.P.Q. 151 (C.C.P.A.1973). The parties’ lines of business need not be the same, so long as their “ ‘products [or services] are the kind the public attributes to a single source.’ ” Int’l Kennel, 846 F.2d at 1089 (citation omitted). “ ‘[A] trade-mark protects the owner against not only its use upon the articles to which he has applied it but also upon such other articles as might naturally or reasonably be supposed to come from him.’ ” Helene Curtis Indus. v. Church & Dwight Co., 560 F.2d 1325, 1331 (7th Cir.1977), cert. denied, 434 U.S. 1070, 98 S.Ct. 1252, 55 L.Ed.2d 772 (1978) (quoting California Fruit Growers Exch. v. Windsor Beverages, 118 F.2d 149, 152-53 (7th Cir.1941). The parties need not be in direct competition and their goods and services need not be identical. Halliburton Co. v. Halliburton Pipe and Steel, 207 U.S.P.Q. 318, 320 (S.D.Tex.1980) (concurrent use between large diversified oil field service company and distributor of steel supplies to petroleum companies).
The district court’s finding of no concurrent use when the parties in fact share a number of clients and marketing methods and the district court’s reliance on the one way in which the parties’ marketing differed do not sufficiently support its conclusion which we find to be clearly erroneous.
3. Whether the trial court erred in finding that buyers of business training programs are sophisticated.
Appellant contends that the court erred in applying the “degree of care likely exercised by consumers” prong of the test, since the court focused only on the question whether appellee’s customers were discriminating. Appellant had produced evidence that clients typically did not spend much time considering which training program to purchase when the programs were as inexpensive as the ones offered by appellant. Appellee had produced contrary evidence, which the trial court apparently relied on, in finding that the buyers of these training programs are sophisticated and careful purchasers.
It serves no useful purpose to focus exclusively on the evidence about either a plaintiff’s or defendant’s customer group, especially in the present case in which the parties share a number of major customers. In order to assess the possibility of customer confusion, the court properly should have considered the group of potential purchasers of both products, since even those who only buy appellant’s services (i.e., non-manufacturers) could be confused by appellee’s mark as to the source of appellant’s services. See Fuji Photo Film v. Shinohara Shoji Kabushiki Kaisha, 754 F.2d 591, 596 (5th Cir.1985) (and cases cited therein). However, the district court did not simply ignore appellant’s evidence of lack of customer sophistication, but also found it unpersuasive. Memor. op. at 16. The district court is certainly in a better position than we are to judge the credibility of appellant’s witnesses. Accordingly, we are not persuaded that the district court’s emphasis on the evidence about appellee’s customers so infected its conclusions as to make them clearly erroneous.2
4. Whether the trial court reasonably discounted the importance of the actual instances of confusion shown by plaintiff.
With regard to the evidence produced by plaintiff of actual confusion, the trial court wrote that it “is neither substan[443]*443tial nor especially convincing. For example, Ms. Fuerst’s [from Motorola] confusion is just as likely the product of her own lack of organization as it is the similarity of the parties’ names.” Memor. op. at 19. Appellant cites the case of Harold F. Ritchie, 281 F.2d at 761, in which the reviewing court found that the trial court erred by discounting evidence of confusion by attributing it to carelessness of the purchaser. Appellant cites other cases in which the courts have emphasized the importance of any evidence of actual confusion. E.g., Fuji Photo, 754 F.2d at 597. See also McGraw-Edison, 787 F.2d at 1172 (“ ‘very little proof of actual confusion would be necessary to prove likelihood of confusion’ ”) (citation omitted). Of course, as appellee points out, de minimis evidence of confusion may be discounted. Union Carbide Corp. v. EverReady Inc., 531 F.2d 366, 383 (7th Cir.1976). Nevertheless, it does not seem reasonable to classify appellant’s evidence—including the testimony of John Connellan and Elaine Fuerst and evidence of other phone calls from other persons confused about the identity of appel-lee—as de minimis, since it was not based on a full survey of customers. In other words, this is not a case in which we could conclude that there was a statistically insignificant percentage of confusion, since the actual percentage is unknown.
In summary, we found that the district court applied the wrong legal principles to its comparison of the parties’ trademarks which infected its conclusion on that factor. We also find that the court clearly erred in finding no concurrent use between the parties’ services and erred by discounting appellant’s evidence of actual confusion. As a result, we reverse the district court’s finding of no likelihood of confusion.
B. Whether the trial court erred in classifying “forum” as suggestive.
Appellee argues that if we find that the trial court erred in applying the likelihood of confusion test, we should also reverse the trial court’s finding that “forum” may be protected. In order to determine whether a trademark is subject to protection, a court must determine the mark’s place on the spectrum from generic to arbitrary. A generic term can never function as a trademark, while a merely descriptive term may be protected only if it has acquired secondary meaning. Liquid Controls Corp. v. Liquid Control Corp., 802 F.2d 934, 935 (7th Cir.1986). Suggestive and arbitrary terms are protected, without a showing of secondary meaning. Miller Brewing Co. v. G. Heileman Brewing Co., 561 F.2d 75, 79 (7th Cir.1977), cert. denied, 434 U.S. 1025, 98 S.Ct. 751, 54 L.Ed.2d 772 (1978). The trial court found that “forum” is suggestive, because “it suggests a certain characteristic of the product but is not specific enough to enable the consumer to identify it.” Memor. op. at 16 (citing J. Gilson, Trademark Protection and Practice, § 2.04). Appellee argues that this finding is erroneous, that “forum” is either generic or descriptive.
A suggestive term is one that “suggests rather than describes an ingredient or characteristic of the goods and requires the observer or listener to use imagination and perception to determine the nature of the goods.” Miller Brewing, 561 F.2d at 79. “[T]he ‘imagination’ required to link a suggestive term with the corresponding product ‘refers to the mental process required to connect a name that is incongruous or figurative with the product (e.g., ‘Roach Motel’ with an insect trap or ‘TIDE’ with soap)....’” G. Heileman Brewing Co. v. Anheuser-Busch Inc., 873 F.2d 985, 997 (7th Cir.1989) (quoting district court opinion, 676 F.Supp. 1436, 1491 n. 46 (E.D.Wis.1987)).
The district judge in the present case did not explain how “forum” suggests the particular programs and methods used by the parties. That “forum” may be an unfamiliar word which requires a hearer to think about its meaning does not show that it is suggestive. “Forum” at most describes the teaching method used by the parties and is neither incongruous nor figurative in its method of description. We rejected in G. Heileman v. Anheuser-Busch the argument that “L.A.” is a sug[444]*444gestive term, because of the thought process required to translate initials into descriptive words. Id. If that thought process is not enough to make initials suggestive, then the thought process required to understand “forum” as suggestive of the parties’ programs is not either. The district court erred by classifying “forum” as suggestive.
Appellee urges us to find that “forum” is generic and, therefore, not subject to trademark protection. The district court considered Webster’s definition for “forum”:
3a: an organization that holds public meetings for the discussion of subjects of current interest b: a meeting that is held by such an organization and is frequently in the form of a question period following a lecture; broadly: a lecture followed by audience discussion and questioning of the lecturer....
Webster’s Third International Dictionary of the English Language Unabridged (G. & C. Merriam Company 1961) at p. 896; Def. Ex. 109. The court concluded that “consumers of business training programs” do not “consider ‘Forum’ to be a generic reference to that line of products.” Memor. op. at 5. We agree.
“A generic term is one that is commonly used as the name of a kind of goods.” Liquid Controls, 802 F.2d at 936. “In order to be generic ... the word in question must serve to denominate a type, a kind, a genus or a subcategory of goods [or services].” Heileman, 873 F.2d at 997. Because a generic term is a name in common usage, a dictionary has been used as evidence “about the meaning attached by a linguistic group to a particular ... symbol.” Id.
In Henri’s Food, we applied a similar analysis to the word “tasty” as it relates to salad dressing and reversed the district court’s finding that it is generic:
[U]nlike ‘light beer,’ [which is generic] ‘tasty salad dressing’ is not a kind, sort, genus or subcategory of salad dressing. ... There can be no suggestion that ‘tasty dressing’ is a kind or type or subcategory of dressing such as, for example, French dressing.
Henri’s Food, 817 F.2d at 1306.
Although “forum” describes a method for teaching business skills or any other subject — “a lecture followed by audience discussion and questioning of the lecturer” —it does not name any commonly used category or subcategory. “[C]onsumers of business training programs” do not “consider ‘Forum’ to be a generic reference to that line of products.” Memor. op. at 5. Moreover, the parties provide a service which includes more than a simple forum, since they sometimes customize teaching methods for companies and sometimes provide written materials for participants to use outside of the seminar. The service is also different from many other types of forums, because of the skills which are taught, the specific target audience and the fact that there is a fairly substantial charge for attendance at these business skills forums. “Forum” describes a characteristic of the services provided by the parties, not the service itself.
We are, therefore, led to the conclusion that “forum” is a descriptive term, in contrast to the trial judge’s conclusion that it is suggestive. He distinguished “forum” from other descriptive terms as follows:
‘Forum,’ ... does not specifically describe business training and education programs. Nor is use of the term necessary to describing similar services. While the term suggests something about the format of plaintiff[-appellant]’s programs, it is not ‘merely descriptive’ in the trademark sense.
Memor. op. at 6. We must disagree with the trial court’s reasoning and conclusion. First, it is not necessary that a descriptive term depict the service itself, but only that the term refer to a characteristic of the service. Miller, 561 F.2d at 79. Second, the word need not be necessary to the description of similar services. In Liquid Controls, we noted the fact that descriptive terms may be needed in order to describe “services of a similar nature,” 802 F.2d at 936, in order to show why such [445]*445words could not be removed from the public domain. We did not, either there or in any other case, find that a term must be necessary to describe similar products before it can be classified as descriptive. Such a rule would mean that any word which has a synonym is not descriptive, and that is clearly not the law in this circuit. If “forum” is descriptive, as we have concluded, then it is entitled to trademark protection only if it has acquired secondary meaning.
IV. Conclusion.
Because we have found that the district court clearly erred in applying the applicable legal principles, we reverse the court’s conclusion that there was no likelihood of confusion. In addition, we have found that the court clearly erred in classifying “forum” as suggestive rather than merely descriptive. Accordingly, we remand on the issue whether appellant can show evidence of secondary meaning. If the court concludes that there is secondary meaning attached to appellant’s mark, then it should reconsider the evidence of likelihood of confusion in light of our ruling.