OPINION

GILMAN, Circuit Judge.

Therma-Scan, Inc. (TSI), a Michigan corporation, brought this lawsuit for trademark infringement and unfair competition against Thermoscan, Inc., a Georgia corporation. TSI sought monetary damages and injunctive relief, as well as cancellation of Thermoscan's allegedly infringing trademark. The district court granted Thermoscan's motion to enforce a purported settlement that the parties reached during oral argument on Thermoscan's motion for summary judgment. We reversed the district court's order in a prior appeal because a material dispute existed as to whether a meeting of the minds had actually occurred regarding the purported settlement. The case was then remanded "for a ruling on Thermoscan's motion for summary judgment and, if necessary, for a trial on the merits." Therma-Scan, Inc. v. Thermoscan, Inc., 217 F.3d 414, 415 (6th Cir. 2000). After allowing the parties to file supplemental briefs, the district court granted summary judgment in favor of Thermoscan. For the reasons set forth below, we AFFIRM the judgment of the district court.

I. BACKGROUND

A. Factual background

TSI performs infrared thermal-imaging examinations of the human body. After analyzing the resulting images, TSI prepares diagnostic reports that are provided to its clients or their physicians.

Most of TSI's clients are residents of the greater Detroit, Michigan area. TSI performs approximately 28 examinations each week at its facility in Huntington Woods, Michigan. Although a few clients learn about TSI's services on their own, the large majority are referred by their physicians.

On November 1, 1988, TSI received the formal registration of its "THERMA-SCAN" trademark from the United States Patent and Trademark Office. TSI's trademark appears on its office door, stationery, literature, reports to physicians and patients, and informational and advertising materials.

Although TSI has sold various medical devices in the past, it no longer markets any products. TSI promotes its services primarily through maintaining and developing contacts with physicians. In addition, it seeks to cultivate relationships with individuals who conduct radio news programs on which TSI's services might be mentioned. TSI does not, however, advertise in any magazines, professional journals, or other publications.

Thermoscan, a Georgia corporation that began its operations in March of 1989, manufactures hand-held electronic ear thermometers. Sales of these devices increased from $3 million in 1991 to over $147 million in 1996. During the same period, Thermoscan's advertising and promotional expenses increased from $1.2 million to over $20 million.

On November 16, 1989, Thermoscan filed an application to register "THERMOSCAN" as a trademark for use on its thermometers. It began using the "THERMOSCAN" trademark on its thermometers in 1990, prior to the completion of the trademark registration process. The United States Patent and Trademark Office accepted Thermoscan's application and published it for opposition on January 15, 1991. A registration for THERMOSCAN was issued on September 24, 1991.

In October of 1992, TSI's attorney informed Philip Hoekstra, the president of TSI, that Thermoscan was using the THERMOSCAN trademark on its thermometers. The attorney recommended that TSI take action against Thermoscan for infringement of the THERMA-SCAN trademark. Hoekstra decided against pursuing any action at that time because he believed that the thermometers would not interfere with TSI's business.

Over a year later, in December of 1993, Hoekstra sent TSI's attorney a facsimile with a subject line reading "Ref: Infringement on trademark." That communication included a catalog advertisement for Thermoscan's thermometer. Hoekstra became concerned with Thermoscan's use of the THERMOSCAN trademark, and the possibility that it might create confusion with TSI's services, when he saw the thermometer being sold in drugstores and being advertised in magazines and on television.

In 1995, The Gillette Company acquired Thermoscan. Braun, Inc., a subsidiary of Gillette, began marketing Thermoscan's thermometers in April of 1996. The trademark "BRAUN" now appears on all Thermoscan packaging and on the thermometers themselves.

Despite its awareness of Thermoscan's products and the THERMOSCAN trademark, TSI did not take any action against Thermoscan until August of 1996, when TSI's attorney sent Thermoscan a protest letter. On September 10, 1996, TSI commenced a proceeding in the United States Patent and Trademark Office to cancel Thermoscan's registration. TSI's cancellation proceeding was suspended to await the disposition of the present lawsuit.

B. Procedural background

This lawsuit was filed in the United States District Court for the Western District of Michigan on January 26, 1998. TSI's complaint alleges trademark infringement and unfair competition, in violation of 15 U.S.C. §§ 1114 and 1125(a), respectively. In addition, TSI requested that the district court, pursuant to its authority under 15 U.S.C. § 1119, issue an order directing the Commissioner of Patents and Trademarks to cancel Thermoscan's trademark registration.

Thermoscan filed a motion for summary judgment in October of 1998. The district court conducted a hearing on the motion several months later. During oral argument, the parties reached a purported settlement. The district court then summarized for the record the understanding between the parties with respect to the settlement. Despite this development, the parties were unable to agree on the specific terms of their purported agreement in subsequent discussions. Thermoscan eventually filed a motion to enforce the settlement or, in the alternative, for a ruling on its motion for summary judgment. The district court granted the motion to enforce the settlement agreement as drafted by Thermoscan and dismissed the lawsuit with prejudice on April 20, 1999.

TSI appealed the district court's order. We reversed and remanded the case, concluding that "it was an abuse of discretion for the court to impose Thermoscan's version of the settlement upon TSI." Therma-Scan, Inc. v. Thermoscan, Inc., 217 F.3d 414, 420 (6th Cir. 2000).

The district court allowed the parties to file supplemental briefs on the motion for summary judgment after the remand. On October 25, 2000, the court granted Thermoscan's motion for summary judgment and entered judgment against TSI. This appeal followed.

II. ANALYSIS

A. Standard of review

A district court's grant of summary judgment is reviewed de novo. Holloway v. Brush, 220 F.3d 767, 772 (6th Cir. 2000). Summary judgment is proper where no genuine issue of material fact exists and the moving party is entitled to judgment as a matter of law. Fed.R.Civ.P. 56(c). In considering such a motion, the court must view the evidence and draw all reasonable inferences in favor of the nonmoving party. Matsushita Elec. Indus. Co. v. Zenith Radio Corp., 475 U.S. 574, 587, 106 S.Ct. 1348, 89 L.Ed.2d 538 (1986). The central issue is "whether the evidence presents a sufficient disagreement to require submission to a jury or whether it is so one-sided that one party must prevail as a matter of law." Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 251-52, 106 S.Ct. 2505, 91 L.Ed.2d 202 (1986).

B. Whether a likelihood of confusion exists — the eight-factor test

In order to prevail on its trademark infringement and unfair competition claims, TSI must establish that Thermoscan's trademark creates a likelihood of confusion regarding the origin of the goods or services offered by TSI and Thermoscan. Daddy's Junky Music Stores, Inc. v. Big Daddy's Family Music Ctr., 109 F.3d 275, 280 (6th Cir. 1997) ("The touchstone of liability under § 1114 is whether the defendant's use of the disputed mark is likely to cause confusion among consumers regarding the origin of the goods offered by the parties."); Holiday Inns, Inc. v. 800 Reservation, Inc., 86 F.3d 619, 622-23 (6th Cir. 1996) (explaining that a party seeking to avoid summary judgment in a case alleging trademark infringement and unfair competition, in violation of 15 U.S.C. §§ 1114 and 1125(a), "must establish that genuine factual disputes exist concerning those factors that are material to whether confusion is likely in the marketplace as a result of the alleged infringement").

This court has identified eight factors that are relevant in determining whether a likelihood of confusion exists: (1) the strength of the plaintiff's mark, (2) the relatedness of the goods or services offered by the plaintiff and the defendant, (3) the similarity of the marks, (4) any evidence of actual confusion, (5) the marketing channels used by the parties, (6) the probable degree of purchaser care and sophistication, (7) the defendant's intent in selecting its mark, and (8) the likelihood of either party expanding its product line using the marks. Daddy's Junky Music Stores, 109 F.3d at 280 (citing Frisch's Restaurants, Inc. v. Elby's Big Boy, Inc., 670 F.2d 642, 648 (6th Cir. 1982)). Not all of these factors will be relevant in every case, and in the course of applying them, "[t]he ultimate question remains whether relevant consumers are likely to believe that the products or services offered by the parties are affiliated in some way." Homeowners Group, Inc. v. Home Mktg. Specialists, Inc., 931 F.2d 1100, 1107 (6th Cir. 1991).

The significance of these factors also varies with the form of trademark infringement at issue. In the first and most common type of infringement, similar marks on directly competing goods or services cause confusion over their origin. This situation is known as "palming off," because the defendant junior user misleads the public about the source of its goods or services, leading consumers to purchase the defendant's products in the belief that they are buying the plaintiff's. Ameritech, Inc. v. American Info. Techs. Corp., 811 F.2d 960, 964 (6th Cir. 1987) (explaining that in "palming off," similar marks on competing goods "can cause the consumer to mistakenly buy the infringing defendant's product as the plaintiff's; the defendant tries to `palm off' his goods as the plaintiff's"). A second form of infringement — confusion of sponsorship — "occurs where the goods do not directly compete," but the trademarks are so similar that consumers might mistakenly believe that the junior user is associated with the senior user. Id. In this scenario, "the defendant seeks to capitalize on the plaintiff's goodwill and established reputation." Id. The final type of infringement involving confusingly similar trademarks is "reverse confusion of sponsorship." Id. Reverse confusion occurs where

the junior user saturates the market with a similar trademark and overwhelms the senior user. The public comes to assume the senior user's products are really the junior user's or that the former has become somehow connected to the latter. The result is that the senior user loses the value of the trademark — its product identity, corporate identity, control over its goodwill and reputation, and ability to move into new markets.

Id. In the present case, TSI's allegations concern both palming off and reverse confusion.

The determination of whether a likelihood of confusion exists is a mixed question of fact and law. Data Concepts, Inc. v. Digital Consulting, Inc., 150 F.3d 620, 624 (6th Cir. 1998). Any disputes about the evidence that pertains to the eight factors set forth above presents a factual issue. Id. at 624 (noting that "[i]f the facts relevant to the applicable factors are contested, factual findings must be made with respect to each of these factors"). In contrast, "the further determination of whether a given set of foundational facts establishes a likelihood of confusion is a legal conclusion." Homeowners Group, 931 F.2d at 1107; Data Concepts, 150 F.3d at 624 (explaining that "the balancing of [the factual] findings to determine the ultimate issue of likelihood of confusion is a question of law").

Because this appeal follows the district court's grant of summary judgment, we must first "determine whether the District Court correctly held that no genuine issues of material fact were presented regarding the likelihood of confusion factors." Homeowners Group, 931 F.2d at 1107. The second step of our review involves an inquiry into the legal question of whether the relevant facts create a likelihood of confusion. Daddy's Junky Music Stores, 109 F.3d at 279 (noting that after examining the relevant factual findings, the reviewing court must determine "whether those findings overall reveal a likelihood of confusion"). Pursuant to the standard that applies in reviewing a district court's grant of summary judgment, both stages of the review are conducted de novo. Id. at 280 (clarifying that a de novo standard of review applies to all aspects of the district court's grant of summary judgment in a trademark infringement case).

1. Strength of TSI's trademark

This factor focuses on the distinctiveness of a mark and its recognition among the public. Homeowners Group, 931 F.2d at 1107 (explaining that "[a] mark is strong if it is highly distinctive, i.e., if the public readily accepts it as the hallmark of a particular source") (internal quotation marks and citation omitted); Daddy's Junky Music Stores, 109 F.3d at 280 ("The more distinct a mark, the more likely is the confusion resulting from its infringement, and therefore, the more protection it is due.") (citation omitted). Generally, the strength of a mark is the result of its unique nature, its owner's intensive advertising efforts, or both. Daddy's Junky Music Stores, 109 F.3d at 280 (noting that public acceptance of a mark "can occur when the mark is unique, when it has received intensive advertisement, or both") (citation omitted).

A trademark's distinctiveness and resulting strength also depends partly upon which of four categories it occupies: "generic, descriptive, suggestive, and fanciful or arbitrary." Id. Whereas a descriptive trademark "specifically describes a characteristic or ingredient of an article," Champions Golf Club, Inc. v. The Champions Golf Club, Inc., 78 F.3d 1111, 1117 (6th Cir. 1996) (internal quotation marks and citation omitted), "[a]n arbitrary mark has a significance recognized in everyday life, but the thing it normally signifies is unrelated to the product or service to which the mark is attached, such as CAMEL cigarettes or APPLE computers." Daddy's Junky Music Stores, 109 F.3d at 280-81 (internal quotation marks and citations omitted).

Fanciful or arbitrary marks therefore represent the far extreme on "a spectrum of increasing strength" among the categories. Id. at 280. Even an arbitrary trademark is not a strong mark, however, if it does not achieve broad public recognition across product lines. Homeowners Group, 931 F.2d at 1107 ("HMS may indeed be arbitrary and hence inherently distinctive, yet have little customer recognition or `strength' in the market, or perhaps have high recognition which is limited to a particular product or market segment."); see also Daddy's Junky Music Stores, 109 F.3d at 281 (noting that "[a]ssigning a category to a mark constitutes only a single step in determining the strength of the mark").

The district court determined that TSI's trademark is descriptive because it describes the services that TSI performs. We agree with this conclusion, and so apparently does TSI. Hoekstra, TSI's president, testified that TSI chose the name because "it's the best description, in a succinct manner, as to what it is that we're doing."

Moreover, TSI does not conduct vigorous advertising, but instead seeks to establish contacts within the medical community and relies upon periodic discussions of its services on radio shows. A company's lack of advertising does not necessarily reduce the strength of its trademark. See The Morningside Group Ltd. v. Morningside Capital Group, L.L.C., 182 F.3d 133, 139 (2d Cir. 1999) (noting that "[w]hen a claimant has no need for traditional advertising because of the nature of its market, it should not feel compelled to advertise simply to protect its service mark"). The absence of any advertising, however, diminishes the likelihood that most people will be familiar with a company's mark, absent evidence that its goods or services have achieved broad public recognition.

TSI does not attempt to establish that its trademark is widely recognized among the general population. Instead, it relies upon the presumption that a trademark that has been registered and uncontested for five years is a strong mark. Data Concepts, 150 F.3d at 625 (stating that such a presumption exists); Wynn Oil Co. v. Thomas, 839 F.2d 1183, 1187 (6th Cir. 1988) (noting that "once a mark has been registered for five years, the mark must be considered strong and worthy of full protection").

TSI's reliance upon this presumption is misplaced. Even where a trademark is incontestable and "worthy of full protection," the significance of its presumed strength will depend upon its recognition among members of the public. Treating a valid, incontestable trademark as an exceptionally strong mark for the purpose of determining whether confusion is likely to occur, without examining whether the mark is distinctive and well-known in the general population, would shift the focus away from the key question of "whether relevant consumers are likely to believe that the products or services offered by the parties are affiliated in some way." Homeowner's Group, 931 F.2d at 1107. Although a trademark may be "strong and worthy of full protection" because it is valid and incontestable, Wynn Oil Co., 839 F.2d at 1187, that does not necessarily mean that its strength is particularly relevant to the ultimate issue of whether confusion is likely to occur.

TSI's trademark, although valid and incontestable, is not an especially strong mark. Not only is the mark descriptive, but it also lacks broad public recognition. As a result, this factor does not weigh strongly in TSI's favor.

2. Relatedness of the goods or services

This court has identified three categories regarding the relatedness of the goods or services with which trademarks are associated:

First, if the parties compete directly by offering their goods or services, confusion is likely if the marks are sufficiently similar; second, if the goods or services are somewhat related but not competitive, the likelihood of confusion will turn on other factors; third, if the goods or services are totally unrelated, confusion is unlikely.

The district court concluded that although TSI and Thermoscan might coexist "in a very broad industry of medical applications of thermology and infrared identification of heat," TSI's services and Thermoscan's goods are not so related that any confusion is likely to occur. We agree with this conclusion. TSI and Thermoscan offer goods and services that utilize similar technology in very different ways. Moreover, because they market their goods and services to different segments of the population, TSI and Thermoscan do not compete in the marketplace.

TSI contends that the district court prematurely determined that this factor weighs against a likelihood of confusion, and that it instead should have concentrated on the other factors. This position relies upon a belief that the present case implicates Daddy's Junky Music Stores's second category of related goods or services, where the goods are somewhat related but do not compete. We believe that any commonality between TSI's services and Thermoscan's thermometers is insufficient to establish that their products are related for the purpose of determining whether a likelihood of confusion exists. Homeowners Group, 931 F.2d at 1109 (explaining that companies that "operate at different levels in the broad real estate industry and sell to two completely distinct sets of buyers" are not similar, even though their services are not "totally unrelated"). As a result, the second factor supports a finding that confusion is not likely to occur.

3. Similarity of the marks

This factor entails more than a simple side-by-side comparison of the trademarks in question. Id. (explaining that "in evaluating similarity of marks, it is axiomatic in trademark law that side-by-side comparison is not the test") (internal quotation marks and citation omitted). Instead, the relevant inquiry is whether a particular trademark, when viewed alone, would lead to uncertainty about the goods or services that it identifies. Daddy's Junky Music Stores, 109 F.3d at 283 (noting that "courts must determine whether a given mark would confuse the public when viewed alone, in order to account for the possibility that sufficiently similar marks may confuse consumers who do not have both marks before them but who may have a general, vague, or even hazy, impression or recollection of the other party's mark") (internal quotation marks and citation omitted); see also Homeowners Group, 931 F.2d at 1109 ("A court must determine, in the light of what occurs in the marketplace, whether the mark will be confusing to the public when singly presented.") (internal quotation marks and citation omitted). As such, a detailed analysis of specific features of a trademark is not appropriate; rather, "courts must view marks in their entirety and focus on their overall impressions, not individual features." Daddy's Junky Music Stores, 109 F.3d at 283.

The district court found that "the degree of similarity between the marks is evident." As the district court noted, the two marks — "THERMA-SCAN" and "THERMOSCAN" — are identical except that TSI's trademark has an "a" rather than an "o" and includes a dash preceding the "s". Although the typeface and graphical design of the trademarks differ, these contrasts relate to the marks' "individual features" rather than the "overall impressions" that the marks convey. Id. These differences, therefore, do not negate the high degree of similarity between the marks.

Thermoscan contends that despite these similarities, the prominent display of the BRAUN name on all thermometers and packaging reduces the likelihood of confusion. Several cases support this position. See Vitarroz Corp. v. Borden, Inc., 644 F.2d 960, 968-69 (2d Cir. 1981) (noting that "[w]hen similar marks are always presented in association with company names, the likelihood of confusion is reduced"); McGregor-Doniger, Inc. v. Drizzle, Inc., 599 F.2d 1126, 1134 (2d Cir. 1979) ("The fact that a trademark is always used in conjunction with a company name may be considered by the trial court as bearing on the likelihood of confusion."). The presence of the BRAUN label on Thermoscan's thermometers, however, does not eliminate the similarity between the trademarks. Instead, this labeling diminishes the likelihood of confusion created by the comparable marks and reduces the importance of this factor.

Based upon these considerations, we believe that an average consumer might be unable to recall which trademark identifies TSI's services as opposed to Thermoscan's thermometers. The similarity of the marks thus increases the likelihood of confusion, although the presence of the BRAUN name on Thermoscan's products decreases the significance of this factor.

4. Evidence of actual confusion

"Evidence of actual confusion is undoubtedly the best evidence of likelihood of confusion." Daddy's Junky Music Stores, 109 F.3d at 284 (citation omitted). Where evidence of actual confusion exists, the weight to which such evidence is entitled varies depending upon both the type and amount of confusion that occurs. Homeowners Group, 931 F.2d at 1110 (noting that "it does not follow that any type or quantum of such evidence [of actual confusion] is entitled to significant weight"). This court has explained that "the existence of only a handful of instances of actual confusion after a significant time or a significant degree of concurrent sales under the respective marks may even lead to an inference than no likelihood of confusion exists." Id. Similarly, confusion that is brief or that occurs among individuals who are not familiar with the products in question is entitled to considerably less weight than are "chronic mistakes and serious confusion of actual customers." Id.

TSI offered 18 e-mail messages that were sent to its Internet address, info@thermascan.com, between November 20, 1998 and July 26, 2000 as evidence that consumers mistakenly believe that it manufactures ear thermometers. Six of these e-mail messages indicate the writer's belief that TSI produces and sells thermometers. Specifically, two prospective customers requested rebates or coupons for Thermoscan's ear thermometers, and a third consumer wanted to obtain a catalog or address of the nearest distributor for the thermometers. The fourth and fifth e-mail messages were inquiries about repairs or replacements from individuals who owned Thermoscan thermometers that were not working properly. Finally, someone wanted to know whether TSI was the manufacturer of "Thermascan thermometers" because of a desire to obtain "year 2000 compliance" information.

Most of the remaining 12 inquiries had international origins, including: (1) distributors of medical devices and research equipment in Austria, Eastern Europe, Japan, and The Philippines, (2) a person working with horses in Canada, (3) an individual planning to build an alternative hospital in Guatemala, (4) a computer hardware development company in China, (5) a company involved in the application of industrial thermography in the automobile industry and electronics in Mexico, and (6) a medical company in Turkey. The remaining e-mail messages were from a medical supply trading company, a physical therapist in Hawaii, and an individual with a general interest in TSI's medical infrared-imaging equipment. All 12 of these businesses and individuals requested information relating to the cost and application of thermography equipment and TSI's products.

The district court determined that only the six incidents where individuals contacted TSI to inquire about Thermoscan's thermometers might support a finding that the authors were confused. According to the court, the remaining 12 e-mail messages did not allow for a finding of actual confusion because none of these inquiries mentioned thermometers. Instead, their authors sought information pertaining to thermography equipment. We agree with this aspect of the district court's analysis. These 12 e-mail messages might support a finding that the authors were confused about TSI's services, but they do not provide evidence of any confusion regarding whether TSI and Thermoscan are associated with each other.

The district court then concluded that the six e-mail messages that specifically mentioned thermometers failed to present a genuine issue of material fact regarding whether any actual confusion occurred. To the extent that the district court based its conclusion upon a belief that TSI's evidence has no bearing on whether actual confusion occurred, we disagree with its analysis. TSI correctly emphasizes that in the context of a motion for summary judgment, any evidence of confusion, regardless of how minimal, weighs in its favor. Daddy's Junky Music Stores, 109 F.3d at 284-85 (concluding that the district court should reconsider the significance of a single incident of actual confusion on remand). As this court explained in Daddy's Junky Music Stores, "[b]earing in mind that a successful Lanham Act plaintiff only must show a sufficient potential of confusion, not actual confusion, the fact that some confusion already has occurred favors plaintiff at least to some extent." 109 F.3d at 284 (emphasis in original).

The six e-mail messages indicating uncertainty regarding whether TSI manufactures any products or only provides thermal-imaging services could support a finding that those people were actually confused. But the weight to which those e-mail messages are entitled for the purpose of determining whether a likelihood of confusion exists among the general public is a different issue. Specifically, several factors considerably reduce the legal significance of TSI's evidence of confusion.

First, the six misdirected e-mail messages become much less probative of actual confusion when measured against the number of people who have purchased Thermoscan's thermometers and who have requested information about them. Thermoscan sold 3,200,000 ear thermometers between January 1, 1997 and July 31, 2000, and it receives, on average, 11,000 calls each month on its toll-free consumer-information telephone number. Six confused customers is legally insignificant in light of the scale of Thermoscan's operations. Checkpoint Sys., Inc. v. Check Point Software, Tech., Inc., 269 F.3d 270, 299 (3d Cir. 2001) (explaining that the evidence of actual confusion, including the plaintiff's receipt of e-mails and customer inquiries pertaining to the defendant's products, was legally insignificant because of "the size of these companies, and the large number of e-mails, customer inquiries, and other communications they receive on a daily basis").

Second, the fact that the confusion occurred in e-mail messages raises the possibility that consumers sent the inquiries about the thermometers to TSI rather than to Thermoscan because they were inattentive or careless, as opposed to being actually confused. S.C. Johnson Son, Inc. v. Johnson, 266 F.2d 129, 141 (6th Cir. 1959) ("The owner of a trademark is not entitled to a guarantee against confusion in the minds of careless and indifferent buyers, and merely occasional cases of confusion or thoughtless error by very inattentive purchasers are of very little significance in trademark and unfair competition cases.") (internal citation omitted). A consumer using a popular Internet search engine to locate an e-mail address for Thermoscan could find TSI's website if he or she mistakenly conducted a search for "thermascan" rather than for "thermoscan." Duluth News-Tribune v. Mesabi Publ'g Co., 84 F.3d 1093, 1098 (8th Cir. 1996) (finding the plaintiff's receipt of the defendant's mail and phone calls "to be de minimis and to show inattentiveness on the part of the caller or sender rather than actual confusion"). The fact that TSI does not advertise increases the likelihood that the people who sent the e-mail messages to TSI were inattentive or careless when attempting to find the e-mail address for Thermoscan, rather than confused about the source of the ear thermometers.

Finally, all of the instances which allegedly support a finding of actual confusion occurred between November 20, 1998, and July 26, 2000. The absence of any evidence of confusion prior to this time period, despite the coexistence of the goods and services since 1990, supports a conclusion that consumers have not been confused about whether TSI is affiliated with Thermoscan. Homeowners Group, 931 F.2d at 1110 (noting that "the existence of only a handful of instances of actual confusion after a significant time or a significant degree of concurrent sales under the respective marks may even lead to an inference than no likelihood of confusion exists"). Moreover, the lack of any indication that consumers were confused prior to November of 1998 further bolsters the likelihood that the e-mail messages were due to carelessness rather than actual confusion.

Based upon all of these considerations, we conclude that the six e-mail messages provide only weak support for finding a likelihood of confusion. This factor therefore does not tilt the balance of determining whether a likelihood of confusion exists to a significant degree in either direction. Daddy's Junky's Music Stores, 109 F.3d at 284 (noting that the absence of evidence of actual confusion is generally entitled to little weight because of the difficulty in obtaining such evidence).

5. Marketing channels used

The fifth factor requires an analysis of the parties' predominant customers and their marketing approaches. Homeowners Group, 931 F.2d at 1110 (noting that this factor "consists of considerations of how and to whom the respective goods or services of the parties are sold"). Where the parties have different customers and market their goods or services in different ways, the likelihood of confusion decreases. Id. (explaining that dissimilarities between the parties' customers and their methods of marketing their products reduces the likelihood of confusion). This factor becomes particularly important where the other factors are not helpful, because it "is very significant in illuminating what actually happens in the marketplace." Id.

TSI argues that both parties promote their goods and services to health care professionals, and that they both make use of the Internet. The district court, however, found significant differences between the predominant customers of TSI and Thermoscan, and concluded that the parties' use of the Internet as a marketing channel was not entitled to significant weight. We agree.

With regard to the parties' customers, Thermoscan sells approximately 80 percent of its thermometers directly to consumers for household use, and the remaining 20 percent are sold to physicians and hospitals. TSI, in contrast, depends upon referrals from physicians for the vast majority of its business. These customer bases are sufficiently distinct so as to greatly reduce the likelihood of confusion.

Turning to the marketing channels actually used, neither party disputes that consumers can learn about their respective goods and services on the Internet. TSI presented no additional evidence of common marketing channels. Such proof was in fact unavailable because TSI does not conduct any formal advertising, but instead relies upon its relationship with physicians and the possibility of radio personalities publicizing its services.

Nor does the availability of information about the parties' goods and services on the Internet automatically lead to the conclusion that they use common marketing channels. Although we have not previously addressed this issue, the Court of Appeals for the Ninth Circuit has explained that " [s]ome use of the Internet for marketing, however, does not alone and as a matter of law constitute overlapping marketing channels." Entrepreneur Media, Inc. v. Smith, 279 F.3d 1135, 1151 (9th Cir. 2002) (emphasis in original). Instead, the relevant questions include: (1) "whether both parties use the Web as a substantial marketing and advertising channel," (2) "whether the parties' marks are utilized in conjunction with Web-based products," and (3) "whether the parties' marketing channels overlap in any other way." Id. (internal quotation marks and citations omitted) (emphasis in original). We believe that these considerations are appropriate, because a non-specific reference to Internet use is no more proof of a company's marketing channels than the fact that it is listed in the Yellow Pages of the telephone directory.

Applying the Entrepreneur Media factors to the present case, TSI does not contend that it uses the Internet to market its services, much less that the Internet is a substantial means for it to establish the personal contacts that it seeks to cultivate with physicians. The second consideration similarly does not support a finding that the parties' use of the Internet is significant for the purpose of determining whether they use common marketing channels, because neither party produces goods or services connected to the Internet. Finally, TSI does not contend that additional overlap exists between the marketing channels used by it and Thermoscan.

In light of these considerations, the fifth factor weighs against finding a likelihood of confusion. The possibility of health care professionals referring their patients to TSI and also purchasing Thermoscan's thermometers might allow for a weak inference that TSI and Thermoscan have some customers in common, but the complete absence of any similar marketing approaches between them discounts the significance of this possibility. 6. Likely degree of purchaser care

With respect to an ordinary buyer, the standard for determining whether he or she would differentiate between products with similar trademarks is the exercise of ordinary caution. A higher degree of care, in contrast, is appropriate where the buyer in question has a particular expertise or sophistication, or is making an expensive or unusual purchase. Homeowners Group, 931 F.2d at 1111 (differentiating between the appropriate standards of care for ordinary buyers and more sophisticated purchasers). This expectation of greater attention, where appropriate, diminishes the likelihood of confusion. Id. ("When services are sold to such [sophisticated] buyers, other things being equal, there is less likelihood of confusion.").

In considering this factor, the district court focused on clients choosing a provider of diagnostic medical services and physicians referring their patients for thermographic scanning. It concluded that such clients and physicians would exercise a high degree of care, and that this factor therefore decreases the likelihood of confusion.

The district court's analysis of this factor is problematic for two reasons. First, given the high degree of similarity between TSI's and Thermoscan's trademarks, the significance of the likely degree of purchaser care for determining whether a likelihood of confusion exists decreases considerably. Daddy's Junky Music Stores, 109 F.3d at 286 (explaining that "confusingly similar marks may lead a purchaser who is extremely careful and knowledgeable about the instrument that he is buying to assume nonetheless that the seller is affiliated with or identical to the other party").

The second problem with the district court's analysis is that it fails to recognize that Thermoscan's customers might mistakenly believe that the thermometers are somehow affiliated with TSI. Data Concepts, Inc. v. Digital Consulting, Inc., 150 F.3d 620, 626 (6th Cir. 1998) (explaining that an analysis of the "likely degree of purchaser care" factor requires a court to consider whether someone purchasing an inexpensive product might believe that it is affiliated with a company that provides sophisticated, expensive services, where the latter company alleges that the former has infringed its trademark). Nevertheless, given that (1) any of Thermoscan's customers who become interested in TSI's services are likely to be careful, (2) TSI does not advertise to the general public, and (3) TSI and Thermoscan do not compete, the potential for ordinary consumers to become confused about the origin of the parties' respective goods and services is relatively low.

In contrast to the district court's conclusion, therefore, we believe that the high degree of care that TSI's customers would presumably use does not weigh against finding a likelihood of confusion. This conclusion, however, does not result in the factor supporting TSI. Instead, its significance is minimal, and the other factors become more important. Daddy's Junky Music Stores, 109 F.3d at 285-86 (explaining that the significance of purchaser care "often will depend upon its relationship with the other seven factors").

7. Thermoscan's intent in selecting its trademark

If a party chooses a mark with the intention of creating confusion between its products and those of another company, "that fact alone may be sufficient to justify an inference of confusing similarity." Id. at 286 (citation omitted). Circumstantial evidence of copying, particularly "the use of a contested mark with knowledge of the protected mark at issue," is sufficient to support an inference of intentional infringement where direct evidence is not available. Id.; see Champions Golf Club, Inc. v. The Champions Golf Club, Inc., 78 F.3d 1111, 1121 (6th Cir. 1996) (instructing the district court to evaluate the credibility of one of the defendant's former employees, who testified that he had told the defendant's owner about the prior use of the mark by the plaintiff). Where there is no evidence that the defendant actually knew that a protected trademark existed, such knowledge can be presumed if evidence exists of the plaintiff's "extensive advertising and long-term use of a protected mark." Daddy's Junky Music Stores, 109 F.3d at 286.

TSI contends that Thermoscan had constructive knowledge of the "THERMA-SCAN" trademark because the mark was registered before Thermoscan's own mark was adopted. According to TSI, Thermoscan failed to investigate or obtain the opinion of counsel prior to seeking registration for its trademark. TSI presents no evidence to support these allegations. The existence of TSI's registered trademark prior to Thermoscan's adoption of its mark, moreover, does not support an inference that intentional copying occurred. Id. at 286-87 ("Plaintiff is incorrect to the extent that it is suggesting that the mere prior existence of a registered mark demonstrates that the alleged infringer intentionally copied that mark; otherwise, presumably all trademark infringement cases could result in a finding of intentional copying."). TSI's relatively small scale of operations in a limited geographical area and the fact that it has not conducted extensive advertising further preclude a presumption that Thermoscan had actual knowledge of TSI's mark.

The district court concluded that TSI had failed to present a genuine issue of material fact with respect to Thermoscan's intent. As the preceding discussion indicates, we agree with this conclusion. The district court erred, however, in finding that the absence of any evidence of intentional copying "weighs in favor of finding little likelihood of confusion between the marks." This factor, rather than tilting the balance in either direction, does not carry significant weight if no evidence of intentional infringement exists. Id. at 287 (explaining that a finding of no intentional copying "is largely irrelevant in determining if consumers likely will be confused as to source") (internal quotation marks and citation omitted).

8. Likelihood of expansion of the parties' product lines

TSI does not contend that it has any plans to expand its services beyond the specialized field in which it currently operates. In evaluating this factor, the district court concluded that the absence of evidence that TSI intended to expand its services decreased the likelihood of confusion. The district court erred in this respect.

This court has explained that although evidence that either party will likely expand its product lines supports finding a likelihood of confusion, "[a] finding that the parties will not expand their marks significantly . . . does not address the ultimate issue of likelihood of confusion." Id. at 287 (internal quotation marks and citation omitted). In contrast to the district court's conclusion, therefore, the eighth factor is not particularly relevant in the present case and does not tilt the balance in either direction.

C. Relative weight of the eights factors in the present case

The preceding discussion of the eight factors relevant to a determination of whether a likelihood of confusion exists indicates that the similarity between TSI's and Thermoscan's trademarks provides the strongest support for TSI's position. As this court has explained, however, even though similarity is "entitled to considerable weight," it does not compel a finding that confusion is likely to occur. Champions Golf Club, 78 F.3d at 1119. Furthermore, the presence of the BRAUN name that appears alongside Thermoscan's trademark on the thermometers and their packaging decreases the likelihood of confusion created by the similarity of the marks.

Three other factors contribute to a finding that a likelihood of confusion exists, albeit in very minimal ways. First, TSI's trademark is entitled to a presumption that it is strong by virtue of its having been registered and uncontested for five years. The facts of this case, however, rebut the significance of this presumed strength, because TSI did not present any evidence that its mark has broad public recognition. Second, TSI offered only a modicum of evidence of actual confusion. This proof provides weak support for finding a likelihood of confusion and therefore has little if any effect on the balancing scale. Finally, the possibility that Thermoscan's advertising might lead some of its Detroit-area customers to believe that TSI is affiliated with Thermoscan favors TSI. Once again, however, this support is relatively weak.

The two factors with the most relevance to the present case — the relatedness of the goods and services and the marketing channels used — both weigh strongly against finding a likelihood of confusion. As explained previously, TSI's diagnostic thermal-imaging examinations and Thermoscan's thermometers neither compete with each other nor are they related for the purpose of determining whether a likelihood of confusion exists. This factor therefore favors a determination that consumers are unlikely to confuse the origin of TSI's services and Thermoscan's products. Similarly, because TSI and Thermoscan have predominately different customers and because their use of the Internet is insufficient to establish common marketing approaches, a consideration of the marketing channels actually used also supports a finding that confusion is not likely to occur.

The final two factors — Thermoscan's intent in selecting its trademark and the likelihood that either party will expand its product line — are not relevant in the present case. Accordingly, they do not contribute to a determination of whether a likelihood of confusion exists.

The most significant considerations in the present case suggest that confusion is not likely to occur, whereas the factors that favor TSI's position provide only minimal support. Moreover, despite TSI's arguments to the contrary, the presence of a reverse-confusion claim does not alter this analysis. In Cohn v. Petsmart, Inc., 281 F.3d 837 (9th Cir. 2002), the Court of Appeals for the Ninth Circuit held that Cohn failed to establish a likelihood of confusion for its reverse-confusion infringement claim against Petsmart. Id. at 843. Cohn, who operated a veterinary clinic in Boise, Idaho, had received a state trademark registration for the phrase "Where Pets are Family," and Petsmart, which owns a national chain of pet supplies stores and leased space to a veterinarian in its Boise store, had obtained a federal trademark for the same slogan. Id. at 839.

The court recognized that both Cohn and Petsmart sell related goods and services, and that "Petsmart's extensive advertising gives it the ability to overwhelm any public recognition and goodwill that Cohn has developed in the mark," a consideration with particular relevance in reverse-confusion cases. Id. at 841-42. Several other factors, however, weighed against any likelihood of confusion. In particular, the court explained that consumers would likely encounter the slogans differently because Cohn and Petsmart (1) had distinctive business names ("Critter Clinic" and "Petsmart"), (2) concentrated their marketing efforts in different media (the Yellow Pages as opposed to television and newspaper advertising and an Internet website), and (3) had not experienced any actual confusion despite six years of coexistence in the same city. Id. at 842-43.

The Ninth Circuit's analysis in Cohn is equally applicable in the present case. Even though Thermoscan's extensive advertising could enable it to undermine the value of TSI's trademark, the parties' goods and services are related in only the most general sense. Moreover, consumers will encounter TSI's services and Thermoscan's thermometers in entirely different settings, one in specialized offices generally after physician referrals, the other in drugstores. TSI and Thermoscan, as noted above, do not use similar marketing channels, and the evidence of actual confusion is of minimal significance in the present case, especially given the many years of coexistence.

The dispute in the present case is not over the underlying facts, but over the legal question of whether this "given set of foundational facts establishes a likelihood of confusion. . . ." Homeowners Group, Inc. v. Home Marketing Specialists, Inc., 931 F.2d 1100, 1107 (6th Cir. 1991). As our discussion indicates, the relevant facts fail to raise a genuine issue as to whether consumers will be confused regarding the origin of TSI's examinations and Thermoscan's thermometers. We therefore conclude that the district court did not err in granting Thermoscan's motion for summary judgment.

III. CONCLUSION

For all of the reasons set forth above, we AFFIRM the judgment of the district court.