WREAL, LLC

v.

AMAZON.COM, INC.

In this case, the plaintiff is Wreal, LLC, a Miami-based pornography company, which has been using the mark “FyreTV” in commerce since 2008. The defendant is Amazon.com, Inc., the largest online purveyor of goods and services in the United States,

19-13285 Opinion of the Court 3 which has been using the mark “Fire TV” (or "fireTV") in commerce since 2012. Wreal does not claim that Amazon, by using the "Fire TV" mark, is attempting to profit off Wreal&#x²⁷;s good name, as would be typical in a forward-confusion case. Instead, Wreal contends that Amazon&#x²⁷;s allegedly similar mark is causing consumers to associate its mark—“FyreTV”—with Amazon.

The resolution of this appeal turns on the likelihood of confusing Amazon&#x²⁷;s “Fire TV” with Wreal&#x²⁷;s "FyreTV.” In forward-confusion cases, we determine likelihood of confusion by applying a well-established seven-factor test. See Welding Servs., Inc. v. For-man, 509 F. 3d 1351, 1360 (11th Cir. 2007). Applying those seven factors, the district court found that consumers were unlikely to confuse "Fire TV" with "FyreTV" and granted summary judgment to Amazon on Wreal&#x²⁷;s trademark infringement claims.

We have not had the opportunity to delineate how this seven-factor test applies in reverse-confusion cases. As discussed below, there are several important differences in how the seven likelihood-of-confusion factors apply in reverse-confusion cases versus forward-confusion cases. When applied specifically to the issues presented here, we conclude that the district court erred in granting summary judgment and should have allowed the case to proceed to trial. We therefore reverse the district court&#x²⁷;s order.

I. FACTUAL AND PROCEDURAL BACKGROUND

A. Wreal, LLC, and FyreTV

In order to make accessing its FyreTV service easier, Wreal also sells a set-top box, [fn²] called the FyreBoXXX, which allows consumers to access FyreTV on their television sets. To purchase a

Apart from the FyreBoXXX and FyreTV.com, Wreal’s customers also have other methods available to access the FyreTV service. For example, both Apple TV and Roku—two commercial set-top boxes that offer a host of general interest channels and media— support FyreTV. Thus, after signing up for an account at FyreTV.com, Wreal’s customers can watch its content from their television set through a computer, a smartphone, a FyreBoXXX, an Apple TV, or a Roku.

B. Amazon and “fireTV”

Amazon is the largest online purveyor of goods in the United States. In 2011, Amazon “started using the mark ‘Fire’ in connection with its Kindle tablets . . . to highlight the new model’s ability to stream video over the internet.” Id. at 1247. In late 2012 and early 2013, Amazon was gearing up to launch several new products, including a phone, a new tablet, and a set-top box. Id. It decided to use the “Fire” brand, as well as its housemark, “amazon,” on these products, with the set-top box being called

Amazon launched fireTV in April 2014 with a nationwide advertising campaign covered by major magazines and television networks. The fireTV is a streaming-only set-top box; it does not contain a DVD tray and cannot play DVDs. Amazon markets the product as a set-top box for general interest content, including "in-stant access to Netflix, Prime Instant Video, WatchESPN,” and more. It is not marketed as a device for streaming pornography. Amazon advertises the device on amazon.com, as well as on television, in print media, and using in-store displays at retailers like Best Buy and Staples. When Amazon began its search-engine-optimization efforts (to help fireTV appear on the internet), it bought ads for keywords related to fireTV, but not for FyreTV or anything related to pornography. Often—but not always—Amazon will market its “Fire” products with its housemark, “amazon.” In the graphics and advertisements for the device, the device is

Amazon&#x²⁷;s fireTV does not broadcast any hardcore porno-graphic material. [fn⁴] But the fireTV does have apps for Showtime and HBO GO, and both of those content providers broadcast softcore pornography as part of their after-hours programming. It is un-clear, however, whether those providers had any such material on their apps that link to fireTV at the time of the lawsuit.

It is undisputed that Amazon&#x²⁷;s policies for Amazon Prime Instant Video, which is Amazon&#x²⁷;s own streaming service and streams on the fireTV, prohibit the sale and consumption of hard-core pornography on the set-top box. However, the record evidence suggests that hardcore pornographic DVDs are available for purchase on amazon.com. The record evidence also suggests that two films with highly suggestive names were available for stream-ing on the fireTV through Amazon Prime Instant Video, though the record does not establish whether those films would be categorized as hardcore or softcore pornography.

C. Evidence of Confusion

In order to prevail on its trademark claims, Wreal must show a “likelihood of confusion.” Forman, 509 F. 3d at 1360. We therefore summarize the record evidence relevant to this issue, as presented by Wreal at the preliminary injunction hearing and by both parties as part of their summary judgment briefing. Below are screenshots of the marks at issue as they appear in internet advertising for the set-top boxes:

[fn 5] Wreal argues that the magistrate judge (and, by its adoption of the report and recommendation, the district court) erred when it considered, for purposes of summary judgment, evidence that was introduced by Wreal during the preliminary injunction hearing but that was not produced during subsequent discovery. This includes a consumer survey conducted by Wreal's ex-pert, Dr. Thomas Maronick. Wreal is incorrect in this assertion. Evidence introduced at a preliminary injunction hearing becomes part of the record, and it is properly before the district court and may be considered when ruling on motions for summary judgment. See Clinkscales v. Chevron U.S.A., Inc., 831 F. 2d 1565, 1570 (11th Cir. 1987) (“The affidavits appended to appellant's motion for a preliminary injunction were, however, part of the written record before the district judge at the time he ruled on the summary judgment motions. These affidavits were therefore properly before the district court.”). And Federal Rule of Civil Procedure 56(c)(3) allows a district court to consider any record evidence when ruling on summary judgment motions.

As noted above, the two products are neither advertised nor sold in the same outlets. A consumer cannot buy a fireTV at the same place where he could buy a FyreTV, and vice versa. Thus, no consumer will come across the products or marks in the same location—whether over the internet or in person at a brick-and-mortar location—save for an internet search engine like google.com. Additionally, Wreal&#x²⁷;s own evidence supports the proposition that mine-run internet consumers would not confuse Amazon&#x²⁷;s amazon.com website with Wreal&#x²⁷;s FyreTV.com web-site.

Over the course of the litigation, both Wreal and Amazon sought to present evidence relevant to the issue of actual consumer

Both parties also presented expert testimony regarding the level of confusion between the marks—Wreal at the preliminary injunction hearing and Amazon at the summary judgment stage. Amazon’s expert, Dr. Dan Sarel, conducted a consumer survey that

[fn⁶] The district court adopted the magistrate judge’s determination that, while this inquiry appears to show confusion, the sender was not confused. The magistrate judge based its conclusion solely on the text of the inquiry itself, and not on any other record evidence. In other words, the magistrate judge (and, by adoption, the district court) did not believe that the sender was con-fused. Credibility determinations like this, however, are inappropriate at the summary judgment stage. See Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 255 (1986) (“Credibility determinations, the weighing of the evidence, and the drawing of legitimate inferences from the facts are jury functions, not those of a judge, whether he is ruling on a motion for summary judgment or for a directed verdict.”). Here, for example, a reasonable juror could view the same evidence and come to the opposite conclusion reached by the magistrate judge and the district court. Because this credibility determination improperly in-vaded the province of the jury, it must be disregarded.

D. Procedural History

Wreal filed this lawsuit against Amazon about two weeks after the fireTV&#x²⁷;s product launch. In its complaint, Wreal sought treble damages and injunctive relief for reverse-confusion trade-mark infringement under the Lanham Act, the Florida Deceptive and Unfair Trade Practices Act, and Florida common law. Five

[fn⁷] Wreal complains about both studies, arguing that the Amazon study was conducted too early to be relevant to the issue of consumer confusion and that its own study was conducted for a separate purpose altogether. Absence of evidence for a proposition, however, is not affirmative evidence to the contrary. And the only survey evidence available to us is not in dispute both surveys show that there was no consumer confusion. Nevertheless, we accord this evidence relatively little weight, as “[t]his Circuit . . . has moved away from relying on survey evidence” in trademark cases. Frehling Enters., Inc. v. Int&#x²⁷;l Select Grp., Inc., 192 F. 3d 1330, 1341 n.5 (11th Cir. 1999).

[fn⁸] As noted by the district court, the protection that these three bodies of law provide is coextensive. See Gift of Learning Found., Inc. v. TGC, Inc., 329 F. 3d 792, 802 (11th Cir. 2003) (“[T]he analysis of the Florida statutory and com-mon law claims of trademark infringement and unfair competition is the same as under the federal trademark infringement claim.”). Because reaching that conclusion "is a question of state law that the parties do not challenge on appeal, we treat the district court&#x²⁷;s holding as correct and merely determine

After the close of discovery, Amazon moved for summary judgment. The district court again referred the motion to the magistrate judge for a report and recommendation, and the magistrate judge recommended granting the motion. Over Wreal&#x²⁷;s objections, the district court adopted the report and recommendation and granted summary judgment to Amazon. Wreal then timely appealed.

II. STANDARDS OF REVIEW

“We review a district court&#x²⁷;s grant of summary judgment de novo and apply the same legal standards as the district court.” Cus-tom Mfg. & Eng’g, Inc. v. Midway Servs., Inc., 508 F. 3d 641, 646 (11th Cir. 2007). Summary judgment is appropriate only when the record shows that there is no genuine issue of material fact such that the movant is entitled to judgment as a matter of law. Fed. R. Civ. P. 56(a). In conducting this inquiry, we view the evidence and reasonable inferences drawn therefrom in the light most favorable

whether the district court properly decided the Lanham Act count[s]." Tana v. Dantanna’s, 611 F. 3d 767, 772 (11th Cir. 2010); see also Jellibeans, Inc. v. Skating Clubs of Ga., Inc., 716 F. 2d 833, 839 (11th Cir. 1983) (“If we determine that the district court decided the Lanham Act count[s] properly, we will also affirm its decision on the [state] deceptive trade practices and [fraud] counts.”).

III. ANALYSIS

Wreal argues that Amazon’s use of the mark fireTV in-fringed its trademark FyreTV under a reverse-confusion theory—

[fn⁹] Amazon argues that Wreal has waived any challenge to the factual conclusions reached by the magistrate judge and adopted, without review, by the district court. In support, Amazon first directs our attention to Wreal’s statement of facts submitted in opposition to Amazon’s motion for summary judgment and the magistrate judge’s conclusion that Wreal did not present sufficient evidence of a genuine dispute for many of the facts at issue. Second, Amazon argues that Wreal failed to object to the magistrate judge’s recitation of facts in the objections Wreal filed with the district court. Our review of the record establishes that Wreal did, in fact, object to many of the conclusions the magistrate judge reached regarding the parties’ competing statements of facts. And, as a matter of law, we are obligated to review the district court’s grant of summary judgment de novo, applying the same legal standards as the district court. Custom Mfg. & Eng’g, 508 F. 3d at 646. This means we place ourselves in the shoes of the district court and apply whatever legal standard governed the district court in resolving the summary judgment motion. Be-cause Wreal did object to the many of the magistrate judge’s factual conclusions and resolution of certain factual disputes, we review those portions of the report and recommendation de novo. For factual conclusions in the report that were not objected to, we review for plain error. See Todd v. Fayette Cnty. Sch. Dist., 998 F. 3d 1203, 1214 n.6 (11th Cir. 2021); 11th Cir. R. 3-1.

In determining the likelihood of confusion, we consider the following seven factors: (1) distinctiveness of the mark alleged to have been infringed; (2) similarity of the infringed and infringing marks; (3) similarity between the goods or services offered under the two marks;

[fn¹⁰] In Bonner v. City of Prichard, 661 F. 2d 1206, 1209 (11th Cir. 1981) (en banc), we adopted as binding precedent all of the decisions of the former Fifth Circuit handed down prior to October 1, 1981.

Although likelihood of confusion generally is a question of fact, see Jellibeans, Inc. v. Skating Clubs of Ga., Inc., 716 F. 2d 833, 840 n.16 (11th Cir. 1983), in limited circumstances it may be decided as a matter of law via summary judgment, see Tana v. Dan-tanna’s, 611 F. 3d 767, 775 n.7. (11th Cir. 2010). In deciding whether to grant summary judgment, each of the seven factors must be considered. But because the weight to be given to each factor will vary depending on the circumstances of the case, summary judgment

In order to resolve this appeal, we must determine how these seven likelihood-of-confusion factors apply in the context of reverse-confusion trademark infringement. The “paradigm case [of reverse confusion] is that of a knowing junior user with much greater economic power who saturates the market with advertising of a confusingly similar mark, overwhelming the marketplace power and value of the senior user’s mark.”⁴ J. Thomas McCar-thy, McCarthy on Trademarks and Unfair Competition § 23:10 (5th ed.); see also Sands, Taylor & Wood Co. v. Quaker Oats Co., 978 F. 2d 947, 957 (7th Cir. 1992) (“Reverse confusion occurs when a large junior user saturates the market with a trademark similar or identical to that of a smaller, senior user. In such a case, the junior user does not seek to profit from the good will associated with the senior user’s mark.”) Because both the harm and the theory of in-fringement in a reverse-confusion case differ from what is claimed in a forward-confusion case, the analysis and application of the seven likelihood-of-confusion factors differ as well.

In a reverse-confusion case, the harms that can occur are varied. For example, consumers may come to believe the smaller, senior user of the mark is itself a trademark infringer, see Banff, Ltd. v. Federated Dep’t Stores, Inc., 841 F. 2d 486, 490 (2d Cir. 1988), or

[t]he 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 trade-mark—its product identity, corporate identity, control over its goodwill and reputation, and ability to move into new markets.

Ameritech, Inc. v. Am. Info. Techs. Corp., 811 F. 2d 960, 964 (6th Cir. 1987). In this case, Wreal contends that “Amazon’s use of Wreal’s mark creates a likelihood that consumers will believe that Amazon is the source of Wreal’s FyreTV service.”

With these principles in mind, we turn to the seven-factor test for likelihood of confusion and analyze each of the factors and their application in a reverse-confusion case. As with all species of trademark infringement, however, the “rule of reverse confusion is highly fact-specific and depends for its application on the presence of a critical mass of key facts.”⁴ McCarthy, supra, § 23:10. The seven factors used to assess likelihood of confusion—regard-less of what theory of infringement is implicated—should never be applied mechanically.

A. Distinctiveness of the Mark

In the typical forward-confusion case, this factor focuses only on the conceptual strength of the plaintiff&#x²⁷;s mark. [fn¹¹] See Frehling Enters., Inc. v. Int&#x²⁷;l Select Grp., Inc., 192 F. 3d 1330, 1335 (11th Cir. 1999) ("Classifying the type of mark Plaintiff has deter-mines whether it is strong or weak.”). This is because in a forward-confusion case, the plaintiff&#x²⁷;s theory is that the defendant—a newer user of the mark at issue—is attempting to profit off the plaintiff&#x²⁷;s goodwill and reputation. And here, the district court did assess the

[fn¹¹] We have summarized the inquiry into the strength of the mark as follows: There are four categories of marks: (1) generic, (2) descriptive, (3) suggestive, and (4) arbitrary. The categories are based on the relationship between the name and the service or good it describes. Generic marks are the weakest and not entitled to protection—they refer to a class of which an individual service is a member (e.g., “liquor store” used in connection with the sale of liquor). Descriptive marks describe a characteristic or quality of an article or service (e.g., “vision center” denoting a place where glasses are sold). “Suggestive terms suggest characteristics of the goods and services and require an effort of the imagination by the consumer in order to be understood as descriptive.” For instance, “penguin” would be suggestive of re-frigerators. An arbitrary mark is a word or phrase that bears no relationship to the product (e.g., “Sun Bank” is arbitrary when applied to banking services). Arbitrary marks are the strongest of the four categories. Frehling, 192 F. 3d at 1335–36 (citations omitted).

But in a reverse-confusion case, the plaintiff is not arguing that the defendant is attempting to profit off the plaintiff&#x²⁷;s good-will. Rather, the plaintiff asserts that the defendant—the junior but more powerful mark user—has been able to commercially over-whelm the market and saturate the public conscience with its own use of the mark, thereby weakening and diminishing the value of the senior user&#x²⁷;s mark. See, e.g., Checkpoint Sys., 269 F. 3d at 302-03. Thus, in this situation, the conceptual strength of the plaintiff&#x²⁷;s mark is necessarily less important to the analysis. See Com. Nat&#x²⁷;l Ins. Servs., Inc. v. Com. Ins. Agency, Inc., 214 F. 3d 432, 444 (3d Cir. 2000) (noting that “it is the strength of the larger, junior user&#x²⁷;s mark which results in reverse confusion"). Accordingly, when assessing the distinctiveness of the mark in a reverse-confusion case, the district court should consider both the conceptual strength of the plaintiff&#x²⁷;s mark and the relative commercial strength of the defend-ant&#x²⁷;s mark. See Visible Sys. Corp. v. Unisys Corp., 551 F. 3d 65, 74 (1st Cir. 2008) (“In a reverse confusion case, the focus is on the relative strengths of the marks so as to gauge the ability of the junior user&#x²⁷;s mark to overcome the senior user&#x²⁷;s mark."); Cohn v. Petsmart, Inc., 281 F. 3d 837, 841 (2d Cir. 2002) (noting that the defendant&#x²⁷;s "extensive advertising gives it the ability to overwhelm any public recognition and goodwill that [the plaintiff] has developed in the mark"); Walter v. Mattel, Inc., 210 F. 3d 1108, 1111 n.2 (9th Cir. 2000) (“In a reverse confusion cases . . . the inquiry focuses

Here, the district court did not consider the commercial strength of Amazon&#x²⁷;s mark because it found that Wreal waived the argument by failing to raise it in its response to Amazon&#x²⁷;s motion for summary judgment and instead raised it for the first time in its objections to the magistrate judge&#x²⁷;s report and recommendation. [fn¹²] The district court erred in that finding. At the summary judgement

[fn¹²] The district court also noted that the presence of Amazon&#x²⁷;s “amazon” housemark alongside “fireTV” in advertisements pushed the distinctiveness-of-the-mark factor further in Amazon&#x²⁷;s favor. As we discuss below, however, the presence of a housemark should be assessed in reference to the second factor in the analysis—the similarity of the marks. See A & H Sportswear, 237 F. 3d at 229-30.

The commercial strength of Amazon&#x²⁷;s mark is manifest and appears in the record. Amazon admitted in its answer that the fireTV was launched with a major advertising campaign, was covered by major magazines and television networks, and that it was a bestseller. Amazon also admits that it advertises the fireTV in multiple brick-and-mortar locations, as well as on amazon.com, one of the most visited online shopping sites in the United States. In short, Amazon&#x²⁷;s overwhelming commercial success with the fireTV mark, coupled with the conceptual strength of Wreal&#x²⁷;s mark, pushes this factor firmly in Wreal&#x²⁷;s favor.

B. Similarity of the Marks

The similarity-of-the-marks analysis is, with one exception related to housemarks noted below, the same in both forward-con-fusion and reverse-confusion cases. We compare “the marks and consider[] the overall impressions that the marks create, including the sound, appearance, and manner in which they are used." Frehling, 192 F. 3d at 1337. In doing so, we determine similarity based on "the total effect of the designation, rather than on a comparison of individual features.” Amstar Corp. v. Domino&#x²⁷;s Pizza,

The district court concluded that the marks at issue—fireTV and FyreTV—were not similar. It reached this conclusion mainly by focusing on the fact that the marks were spelled differently and used different fonts, as well as the fact that they were used differently in commerce. The district court also noted that one of Wreal&#x²⁷;s experts, Dr. Linda Williams, testified that visitors to FyreTV.com would not confuse it with amazon.com. The inquiry under this factor, however, is the similarity of the marks, not the similarity of the services or the similarity of the sales methods— each of which has their own factor and should thus be considered separately.

When the focus is on the similarity of the marks themselves, the result is clear—FyreTV and fireTV are nearly identical. “Fire”

[fn¹³] See supra note 10.

Moreover, the marks need not be identical, as the “purpose in considering the similarity of marks as an indicator of likelihood of confusion is that the closer the marks are, the more likely reasonable consumers will mistake the source of the product that each mark represents." Id. Thus, while "Fyre" and "fire" are spelled differently, and one is capitalized, the words have the same

The Ninth Circuit&#x²⁷;s decision in Dreamwerks Production Group, Inc. v. SKG Studio, 142 F. 3d 1127, 1130 (9th Cir. 1998) is instructive on this point. In that case, the court had to assess the similarity of the marks “Dreamwerks” and “DreamWorks,” which, like the marks at issue here, utilized different spellings and capitalization. Id. The Ninth Circuit concluded that the marks were similar, noting the obvious “perfect similarity of sound” and “similarity of meaning" while determining that even the similarity of sight also weighed in favor of a finding of similarity, as consumers “might shrug off the difference [in spelling and capitalization] as an intentional modification.” Id. at 1131. Our decision in Frehling is also instructive. There, we said that the marks “BELL&#x²⁷; OGGETTI” and "Tavola Collection by OGGETTI" were similar because the presence of the dominant and protected “OGGETTI” in both was likely to be confusing. Frehling, 192 F. 3d at 1337. Each of these conclusions applies here.

Amazon&#x²⁷;s pervasive use of its “amazon” housemark along-side "fireTV" in advertisements warrants separate discussion. In forward-confusion cases—where a commercially superior plaintiff with a strong conceptual mark sues a defendant for attempting to profit off its goodwill—the presence of a housemark is indeed likely to dispel confusion in ordinarily prudent consumers. See, e.g., Cus-tom Mfg., 508 F. 3d at 652 n.10. But in reverse-confusion cases, this presumption is reversed; because the harm is false association of

Amazon&#x²⁷;s use of its housemark alongside advertisements for the "fireTV" does exactly what one might expect it to do: it causes consumers to associate Amazon with fireTV. Because this is a re-verse-confusion case asserting that Amazon&#x²⁷;s use of fireTV causes consumers to associate FyreTV with Amazon instead of Wreal, Amazon&#x²⁷;s use of the housemark supports Wreal&#x²⁷;s theory of recovery. The district court erred in concluding otherwise.

In short, the parties&#x²⁷; marks are nearly identical. Both use the same words, are pronounced the same, and have the same mean-ing. While they are spelled slightly differently and use different fonts, this is not enough to conclude that the marks are dissimilar. Moreover, Amazon&#x²⁷;s pervasive use of its housemark alongside "fireTV" pushes this factor even further in favor of Wreal, as it is likely to confuse consumers into believing that Amazon is the origin of the FyreTV mark. Thus, the similarity-of-the-marks factor weighs heavily in favor of Wreal.

C. Similarity of the Products

The analysis of this factor is the same regardless of the theory of confusion, and "requires a determination as to whether the products are the kind that the public attributes to a single source, not whether or not the purchasing public can readily distinguish

Here, many pieces of record evidence are relevant to the question of whether the fireTV set-top box is similar to the Fyre-BoXXX. The record evidence presented in the district court established that consumers were already able to stream softcore pornography on Amazon&#x²⁷;s fireTV through content providers like HBO GO and Showtime. The record evidence also established that Am-azon Prime Instant Video—Amazon&#x²⁷;s own streaming service, which, like HBO GO and Showtime, is available on the fireTV— offered consumers softcore pornography. And the record evidence also established that: (1) Amazon already offered the sale of hard-core pornographic DVDs and magazines on its related consumer

The question therefore is whether this record evidence would suggest to an ordinarily prudent consumer that a doit-all giant like Amazon—which already sells a set-top box that streams softcore pornography and which competes against other set-top boxes that stream hardcore pornography—would "bridge the gap" to hardcore pornography streaming and release a set-top box that streams exclusively pornographic content. We answer that question in the affirmative. Amazon is a company that already sells hardcore pornography on its website and offers softcore pornography on its set-top box. And it competes in a market in which its direct competitors offer hardcore pornography streaming directly on their set-top boxes. Given this information, a reasonable juror could conclude that Amazon decided to "bridge the gap" and offer a standalone set-top box dedicated to streaming hardcore pornography. See id. The two products at issue therefore “are the kind the public attributes to a single source." E. Remy Martin, 756 F. 2d at 1530.

[fn¹⁴] In reaching the opposite result, the district court relied on our decisions in Tana and Ross Bicycles, Inc. v. Cycles USA, Inc., 765 F. 2d 1502 (11th Cir. 1985). But both cases are readily distinguishable. In Tana, we held that ordinary consumers would not confuse an “old-world-style Italian restaurant where mustached waiters dressed in tuxedos serve classic Italian dishes” with “an upscale sports restaurant, targeting sports enthusiasts and serving contemporary American cuisine in a modern setting decorated with flat-screen televisions.” 611 F. 3d at 778. And in Ross Bicycles, we stated that two bicycle companies— which, critically, both sold only bicycles—did not sell similar products due to the differences in the "size of the tubing, the style of wheels, pedals, seats, kickstands, and the difference in the frame angles.” 765 F. 2d at 1507. Both cases were forward-confusion cases in which the parties at issue competed in only one, identical market—restaurants in Tana and bicycles in Ross Bicycles. And because the parties competed in precisely the same market, we focused on granular differences between the products at issue to resolve the claim. Here, by contrast, the claim is one of reverse-confusion, and one of the parties (Amazon) offers products and services across a host of industries and media. The question here is thus whether a reasonable juror could conclude that Am-azon had "bridged the gap” into hardcore pornography streaming and attribute the fireTV and the FyreBoXXX to a single source.

In this case, it is not unreasonable to fathom that the goods emanate from the same source. Both products are furniture pieces, designed for the home, and both have the capability to house electronic equipment. In addition, both products are marketed as having an Italian design and thus a consumer could, on this basis alone, given that both marks con-note furniture, attribute the products to one source given the shared Italian theme.

Therefore, although the products are some-what dissimilar in composition, function, and design, they are similar in that they are both home furnishings sold under a very similar Italian label, and hence it seems possible that a consumer could attribute both products to a single source. While this possibility is perhaps not strong enough to suggest a likelihood of consumer confusion, neither is it so remote as to raise the opposite inference—that a reasonable consumer would likely not be confused. Thus, to the extent that the district court found that this factor favored ISG significantly, we find that the attribution of such

Decisions from our sister circuits in reverse-confusion cases lend further support to our conclusion here. In Attrezzi, the First Circuit held that the products of two “small electric appliance” manufacturers were similar even though one manufacturer also used the mark on its gourmet foods and dinnerware. 436 F. 3d at 39. In Dreamwerks, the Ninth Circuit concluded that a movie studio and a convention holder had similar products because it would not be unreasonable for consumers to presume that the production company behind Star Trek decided to bridge the gap to convention holding and had begun to host Star Trek conventions. See 142 F. 3d at 1131 (“[M]ovies and sci-fi merchandise are now as complementary as baseball and hot dogs. The main products sold at Dream-werks conventions are movie and TV collectibles and memorabilia; the lectures, previews and appearances by actors which attract customers to Dreamwerks conventions are all dependent, in one way or another, on the output of entertainment giants like DreamWorks.”).

Here, as in E. Remy Martin and Dreamwerks, a reasonable juror could conclude that Amazon was likely to market and sell a product like Wreal’s. Indeed, to see a doit-all giant like Amazon enter the pornographic streaming industry requires no more of an inferential leap than seeing a movie studio begin holding public conventions (as in Dreamwerks) or a liquor company begin selling

Finally, we note that “the more similar the marks are, the less necessary it is that the products themselves be very similar to create confusion.” Attrezzi, 436 F. 3d at 39. Accordingly, we conclude that this factor favors Wreal.

D. Similarity of Sales Outlets and Customer Bases

As for the "similarity of sales outlets” factor, we have held: This factor takes into consideration where, how, and to whom the parties&#x²⁷; products are sold. Direct com-petition between the parties is not required for this factor to weigh in favor of a likelihood of confusion, though evidence that the products are sold in the same stores is certainly strong. The parties&#x²⁷; outlets and customer bases need not be identical, but some degree of overlap should be present. Frehling, 192 F. 3d at 1339 (citations omitted). The analysis of this factor is the same in forward-confusion and reverse-confusion cases.

Here, the district court concluded that the "similarity of sales outlets" factor weighs in favor of Amazon. Amazon&#x²⁷;s fireTV

E. Similarity of Advertising

This similarity of advertising "factor looks to each party&#x²⁷;s method of advertising." Frehling, 192 F. 3d at 1339. "[T]he stand-ard is whether there is likely to be significant enough overlap in the readership of the publications in which the parties advertise that a possibility of confusion could result." Id. at 1340. This inquiry is the same in both forward-and reverse-confusion cases.

Wreal nonetheless argues that this factor favors it because, very broadly speaking, both the fireTV and the FyreBoXXX advertise through search engines, word of mouth, and social media. But Wreal presented no record evidence of audience overlap. Nor does Wreal identify any website (outside of search engines like Google) where both the fireTV and the FyreBoXXX are advertised. As we explained in Tana, rejecting a similar argument: “[T]he only similarity in the advertising channels used by the two parties is their maintenance of websites on the World Wide Web. This similarity would dispel rather than cause confusion, however, because the websites are separate and distinct, suggesting two completely un-related business entities.” 611 F. 3d at 778; see also Therma-Scan, Inc. v. Thermoscan, Inc., 295 F. 3d 623, 637 (6th Cir. 2002) (noting that the availability of information about the parties&#x²⁷; goods on the internet does not lead to the conclusion that they use the same marketing channels).

F. Amazon’s Intent

In the forward-confusion context, the intent factor asks whether the “defendant adopted [the] plaintiff’s mark with the in-tention of deriving a benefit from the plaintiff’s business reputation.” Frehling, 192 F. 3d at 1340. This is because in forward-con-fusion cases, “customers mistakenly think that the junior user’s goods or services are from the same source as or are connected with the senior user’s goods or services.”⁴ McCarthy, supra, § 23:10. Without precedent pointing in any other direction, the district court understandably applied this test for intent and found that Amazon did not adopt the fireTV mark with any intent to derive a benefit from Wreal’s FyreTV mark.

But reverse-confusion cases are different. In this context, the concern is that customers will “purchase the senior user’s goods under the mistaken impression that they are getting the goods of the junior user.” Id. In other words, that “the junior user’s advertising and promotion so swamps the senior user’s reputation in the market that customers are likely to be confused into thinking that the senior user’s goods are those of the junior user.” Id. In this case, Wreal is not suggesting that Amazon chose the fireTV mark with the intention of siphoning Wreal’s goodwill; instead, Wreal claims that, by Amazon’s use of the fireTV mark, Wreal has lost control over its own, more senior mark.

At one extreme, intent could be shown through evidence that a defendant deliberately intended to push the plaintiff out of the market by flooding the market with advertising to create reverse confusion. Intent could also be shown by evidence that, for example, the defendant knew of the mark, should have known of the mark, intended to copy the plaintiff, failed to conduct a reasonably adequate trademark search, or otherwise culpably disregarded the risk of reverse confusion. The tenor of the intent inquiry shifts when considering reverse confusion due to the shift in the theory of confusion, but no specific type of evidence is necessary to establish intent, and the

We agree with and adopt the Ninth Circuit’s approach. Ev-idence of a specific intent to deceive is not a prerequisite to establish intent in reverse-confusion cases, as it is in forward-confusion cases. Indicia of intent can come from a wide variety of sources, including a more generalized intent to obtain market saturation or to proceed with the adoption of a mark in circumstances where the defendant had constructive knowledge of the plaintiff’s mark. The facts of each case will vary, and district courts should accord the intent factor whatever weight it is due under the circumstances.

Here, applying this standard, the evidence of intent is strong. First, Amazon has admitted that, before launching the fireTV, it had actual knowledge of both the FyreBoXXX and Wreal’s FyreTV trademark registration. Wreal I, 840 F. 3d at 1247 (“Amazon was aware of Wreal’s FyreTV mark when it launched Fire TV but did not contact Wreal before launching Fire TV.”). Amazon’s Vice President of Marketing further testified in his deposition that Ama-zon not only chose to proceed with its usage of the fireTV mark after becoming aware of the FyreTV registration, but that his “goal was customers . . . if they search for Amazon Fire TV, if they search for our product I did not want them to first come across a porn site and have that experience.” The district court, upon reviewing that testimony, concluded that no reasonable juror could view it and

G. Actual Confusion

“[E]vidence of actual confusion is the best evidence of a likelihood of confusion.” Frehling 192 F. 3d at 1340. But the presence of such evidence is obviously not a prerequisite to a finding of likelihood of confusion, as it is one of seven factors considered in the likelihood-of-confusion determination. Id. Indeed, “it is not necessary to show actual confusion. One merely has to show that the likelihood of confusion exists.” World Carpets, Inc. v. Dick Lit-trell&#x²⁷;s New World Carpets, 438 F. 2d 482, 489 (5th Cir. 1971). [fn¹⁵] But in assessing the quantum of actual confusion required for a finding in the plaintiff&#x²⁷;s favor, even a "very little” amount of actual confusion is highly probative. See id.

"The strength of such evidence depends on &#x²⁷;the number of instances of confusion,&#x²⁷; &#x²⁷;the kinds of persons confused&#x²⁷; and &#x²⁷;the degree of confusion.” Sovereign Mil. Hospitaller Order of Saint John

[fn¹⁵] See supra note 10.

In reverse confusion cases, evidence of forward confusion will usually be probative. See Freedom Card, Inc. v. JPMorgan Chase & Co., 432 F. 3d 463, 473 (3d Cir. 2005). But even more relevant is direct evidence of reverse confusion—i.e., evidence that consumers of the plaintiff&#x²⁷;s more senior mark became confused as to its source following the launch of the defendant&#x²⁷;s more junior mark. See Sterling Drug, Inc. v. Bayer AG,¹⁴ F. 3d 733, 741 (2d Cir. 1994) (noting that, in a reverse-confusion claim, “the relevant issue is whether consumers mistakenly believe that the senior user&#x²⁷;s products actually originate with the junior user” and that “it is appropriate to survey the senior user&#x²⁷;s customers”). Survey

The record evidence here contains some evidence of actual confusion. For example, Wreal introduced evidence that one of its customers asked over Twitter, "Did you guys just merge with Am-azon?" And one of Amazon&#x²⁷;s customers communicated with Am-azon to ask whether he could access "adult content” on his Ama-zon "fyre” TV. Both instances directly suggest reverse confusion; the first consumer believed Amazon had purchased Wreal&#x²⁷;s trade-mark, and the second consumer contacted Amazon to inquire

[fn¹⁶] See Harvey S. Perlman, The Restatement of the Law of Unfair Competition: A Work in Progress, 80 Trademark Rep. 461, 472 (1990) (“Most surveys do not measure actual confusion. Surveys only give us information about a controlled and artificial world from which we are asked to draw inferences about the real world.").

Amazon and Wreal both also introduced survey evidence regarding the rate of confusion. Dr. Thomas Maronick, who testified for Wreal at the preliminary injunction hearing, conducted a preliminary survey in April 2014 and found "very low consumer confusion" between FyreTV and fireTV. Dr. Maronick also testified that awareness of the FyreTV mark was “very low.” In a similar vein, Dr. Dan Sarel, Amazon&#x²⁷;s expert, conducted a consumer survey and found a confusion rate of one percent, which he testified was "nonexistent” and “statistically insignificant."

[fn¹⁷] The magistrate judge (and, by its adoption, the district court) discounted both pieces of evidence, concluding that neither consumer was actually con-fused. As already discussed in footnote⁵, supra, this amounted to an improper credibility determination that invaded the province of the jury. See Mize v. Jefferson City Bd. of Educ., 93 F. 3d 739, 742 (11th Cir. 1996) (“Where the non-movant presents direct evidence that, if believed by the jury, would be sufficient to win at trial, summary judgment is not appropriate even where the movant presents conflicting evidence. It is not the court&#x²⁷;s role to weigh conflicting evidence or to make credibility determinations; the non-movant&#x²⁷;s evidence is to be accepted for purposes of summary judgment.”).

[fn¹⁸] Wreal identified other pieces of evidence to the district court, but our review of the record indicates that they do not represent direct evidence of actual confusion. For example, one of Wreal&#x²⁷;s customers said, “I plan to buy the new Amazon FireTV box (I know it is NOT related to you guys—although confusion over the name may bring Fyretv some more customers and maybe a domain name sale windfall—more power to you!) Will this new Amazon stream-ing device have a private channel installation of FyreTV in the near future?”

Our caselaw is clear that the "the quantum of evidence needed to show actual confusion is relatively small.” Jellibeans, Inc., 716 F. 2d at 845. But our caselaw imposes no hard-and-fast rule regarding the number of instances required to prevail. See Caliber Auto. Liquidators, 605 F. 3d at 937. “Rather, the court must evaluate the evidence of actual confusion in the light of the totality of the circumstances involved.” AmBrit, 812 F. 2d at 1543; accord World Carpets, 438 F. 2d at 489 (5th Cir. 1971). [fn¹⁹] “[R]eason tells us that while very little proof of actual confusion would be necessary to prove the likelihood of confusion, an almost overwhelming amount of proof would be necessary to refute such proof.").

Our previous decisions serve as guides. In Safeway Stores, when reviewing a district court&#x²⁷;s findings following a bench trial, we held that a mere two instances of confusion from relevant consumers was worthy of consideration. 675 F. 2d at 1166–67,

[fn¹⁹] See supra note 10.

Perhaps most analogous are our decisions in AmBrit and PlayNation. In AmBrit as in this case, the relevant products (ice cream novelties there and set-top boxes, here) were sold to the general public, not professional buyers, and had a “high volume of sales" (at least, such is the case for the fireTV here). See 812 F. 2d at 1544. The district court in AmBrit, after a bench trial, found that four instances of actual confusion supported a finding of actual con-fusion in favor of the plaintiff. See id. And we, reviewing for clear error, affirmed. See id. at 1544–45. Similarly, in PlayNation, the products at issue were playground equipment and pull-up bars which, like ice cream novelties and set-top boxes, are sold to the general public rather than to professional buyers. See 924 F. 3d at 1164. Following a bench trial, the district court found that just two instances of actual confusion—in which the plaintiff&#x²⁷;s customers contacted the defendant for customer support—were sufficient to

As in AmBrit and PlayNation, the reported instances of con-fusion in this case are relatively few. Even after years of litigation, Wreal is able to identify only two instances of potential or actual Wreal consumers being confused as to the source of its product. But the record also contains expert testimony that consumers of pornography are less likely to report their consumption than consumers of other media. Given that we are obliged to "evaluate the evidence of actual confusion in the light of the totality of the circumstances involved,” AmBrit, 812 F. 2d at 1543, we find it appropriate here to take that expert testimony into account when considering the number of reported instances of actual confusion. Alt-hough a close call, we conclude that the two reported instances of actual confusion here are sufficient to make the issue one of triable fact and thus weighs in Wreal&#x²⁷;s favor.

Turning to the survey evidence, both parties advance a number of arguments either for or against the consideration of the surveys. But given that we conclude that the instances of actual confusion present in the record are sufficient to push this factor in Wreal&#x²⁷;s favor, we conclude that it is unnecessary to also address the issue of survey evidence especially as a plaintiff need not pre-sent survey evidence in a trademark claim in order to escape sum-mary judgment. See PlayNation, 924 F. 3d at 1169 (“Lack of survey evidence does not weigh against the plaintiff when determining likelihood of confusion.”); Midwestern Pet Foods, Inc. v. Societe

IV. CONCLUSION

This case addresses the application of the seven likelihood-of-confusion factors to a reverse-confusion trademark infringement case. Although some of those factors are analyzed and applied in the same way in both reverse-confusion cases and the more familiar forward-confusion cases, there are important differences in how other factors are analyzed and applied that stem from the fact that the harm and the theory of infringement differ between for-ward and reverse confusion.

Here, the record evidence establishes that Amazon acquired actual knowledge of Wreal&#x²⁷;s registered trademark and still

As noted throughout our decision, there is no mechanical formula for applying the seven factors relating to likelihood of con-fusion. But when considering all seven factors as they apply to a theory of reverse confusion and taking all the circumstances of this case into account on the record before us, we conclude that they weigh heavily in favor of Wreal and that the district court erred when it entered summary judgment in Amazon&#x²⁷;s favor. We there-fore reverse the district court&#x²⁷;s order. This is not to say that Ama-zon may not ultimately prevail on the merits; rather, it must do so before a jury.

REVERSED AND REMANDED.

[fn 1]: On October¹⁴, 2008, Wreal registered both of its marks—“FyreTV” and “FyreTV.com”—with the U.S. Patent and Trademark Office. Wreal I, 840 F. 3d at 1246. [fn²]: A "set-top box” is “a device that is connected to a television so that the television can receive digital signals.” Set-top Box, Merriam-Webster Online Dic-tionary, https://www.merriam-webster.com/dictionary/set-top%20box (last visited June¹⁹, 2022). [fn 3]: The record shows that Amazon has alternatively used “Fire TV” or “fireTV” in its graphics and advertisements for its set-top box. For purposes of this opinion, we use the stylization of “fireTV,” because Wreal highlighted the in-consistency in its response disputing Amazon&#x²⁷;s statement of undisputed facts. However, we emphasize that we make no ultimate conclusion on whether Amazon&#x²⁷;s mark is stylized as “Amazon Fire TV” or “fireTV.” [fn⁴]: Generally, hardcore pornography refers to “scenes of actual sex acts.” Hard-core, Merriam-Webster Online Dictionary, https://www.merriam-web-ster.com/dictionary/hard-core (last visited June¹⁹, 2022). Softcore pornography refers to “scenes of sex acts that are less explicit than hard-core mate-rial.” Soft-core, Merriam-Webster Online Dictionary, https://www.merriam-webster.com/dictionary/soft-core (last visited June¹⁹, 2022). [fn⁵]: Wreal argues that the magistrate judge (and, by its adoption of the report and recommendation, the district court) erred when it considered, for purposes of summary judgment, evidence that was introduced by Wreal during the preliminary injunction hearing but that was not produced during subsequent discovery. This includes a consumer survey conducted by Wreal&#x²⁷;s ex-pert, Dr. Thomas Maronick. Wreal is incorrect in this assertion. Evidence introduced at a preliminary injunction hearing becomes part of the record, and it is properly before the district court and may be considered when ruling on motions for summary judgment. See Clinkscales v. Chevron U.S.A., Inc., 831 F. 2d 1565, 1570 (11th Cir. 1987) (“The affidavits appended to appellant&#x²⁷;s motion for a preliminary injunction were, however, part of the written record before the district judge at the time he ruled on the summary judgment motions. These affidavits were therefore properly before the district court.”). And Federal Rule of Civil Procedure 56(c)(3) allows a district court to consider any record evidence when ruling on summary judgment motions. [fn⁶]: The district court adopted the magistrate judge’s determination that, while this inquiry appears to show confusion, the sender was not confused. The magistrate judge based its conclusion solely on the text of the inquiry itself, and not on any other record evidence. In other words, the magistrate judge (and, by adoption, the district court) did not believe that the sender was con-fused. Credibility determinations like this, however, are inappropriate at the summary judgment stage. See Anderson v. Liberty Lobby, Inc., 477 U.S. 242, 255 (1986) (“Credibility determinations, the weighing of the evidence, and the drawing of legitimate inferences from the facts are jury functions, not those of a judge, whether he is ruling on a motion for summary judgment or for a directed verdict.”). Here, for example, a reasonable juror could view the same evidence and come to the opposite conclusion reached by the magistrate judge and the district court. Because this credibility determination improperly in-vaded the province of the jury, it must be disregarded. [fn⁷]: Wreal complains about both studies, arguing that the Amazon study was conducted too early to be relevant to the issue of consumer confusion and that its own study was conducted for a separate purpose altogether. Absence of evidence for a proposition, however, is not affirmative evidence to the contrary. And the only survey evidence available to us is not in dispute both surveys show that there was no consumer confusion. Nevertheless, we accord this evidence relatively little weight, as “[t]his Circuit . . . has moved away from relying on survey evidence” in trademark cases. Frehling Enters., Inc. v. Int&#x²⁷;l Select Grp., Inc., 192 F. 3d 1330, 1341 n.5 (11th Cir. 1999). [fn⁸]: As noted by the district court, the protection that these three bodies of law provide is coextensive. See Gift of Learning Found., Inc. v. TGC, Inc., 329 F. 3d 792, 802 (11th Cir. 2003) (“[T]he analysis of the Florida statutory and com-mon law claims of trademark infringement and unfair competition is the same as under the federal trademark infringement claim.”). Because reaching that conclusion "is a question of state law that the parties do not challenge on appeal, we treat the district court&#x²⁷;s holding as correct and merely determine [fn⁹]: Amazon argues that Wreal has waived any challenge to the factual conclusions reached by the magistrate judge and adopted, without review, by the district court. In support, Amazon first directs our attention to Wreal’s statement of facts submitted in opposition to Amazon’s motion for summary judgment and the magistrate judge’s conclusion that Wreal did not present sufficient evidence of a genuine dispute for many of the facts at issue. Second, Amazon argues that Wreal failed to object to the magistrate judge’s recitation of facts in the objections Wreal filed with the district court. Our review of the record establishes that Wreal did, in fact, object to many of the conclusions the magistrate judge reached regarding the parties’ competing statements of facts. And, as a matter of law, we are obligated to review the district court’s grant of summary judgment de novo, applying the same legal standards as the district court. Custom Mfg. & Eng’g, 508 F. 3d at 646. This means we place ourselves in the shoes of the district court and apply whatever legal standard governed the district court in resolving the summary judgment motion. Be-cause Wreal did object to the many of the magistrate judge’s factual conclusions and resolution of certain factual disputes, we review those portions of the report and recommendation de novo. For factual conclusions in the report that were not objected to, we review for plain error. See Todd v. Fayette Cnty. Sch. Dist., 998 F. 3d 1203, 1214 n.6 (11th Cir. 2021); 11th Cir. R. 3-1. [fn¹⁰]: In Bonner v. City of Prichard, 661 F. 2d 1206, 1209 (11th Cir. 1981) (en banc), we adopted as binding precedent all of the decisions of the former Fifth Circuit handed down prior to October 1, 1981. [fn¹¹]: We have summarized the inquiry into the strength of the mark as follows: There are four categories of marks: (1) generic, (2) descriptive, (3) suggestive, and (4) arbitrary. The categories are based on the relationship between the name and the service or good it describes. Generic marks are the weakest and not entitled to protection—they refer to a class of which an individual service is a member (e.g., “liquor store” used in connection with the sale of liquor). Descriptive marks describe a characteristic or quality of an article or service (e.g., “vision center” denoting a place where glasses are sold). “Suggestive terms suggest characteristics of the goods and services and require an effort of the imagination by the consumer in order to be understood as descriptive.” For instance, “penguin” would be suggestive of re-frigerators. An arbitrary mark is a word or phrase that bears no relationship to the product (e.g., “Sun Bank” is arbitrary when applied to banking services). Arbitrary marks are the strongest of the four categories. Frehling, 192 F. 3d at 1335–36 (citations omitted). [fn¹²]: The district court also noted that the presence of Amazon&#x²⁷;s “amazon” housemark alongside “fireTV” in advertisements pushed the distinctiveness-of-the-mark factor further in Amazon&#x²⁷;s favor. As we discuss below, however, the presence of a housemark should be assessed in reference to the second factor in the analysis—the similarity of the marks. See A & H Sportswear, 237 F. 3d at 229-30. [fn¹³]: See supra note 10. [fn¹⁴]: In reaching the opposite result, the district court relied on our decisions in Tana and Ross Bicycles, Inc. v. Cycles USA, Inc., 765 F. 2d 1502 (11th Cir. 1985). But both cases are readily distinguishable. In Tana, we held that ordinary consumers would not confuse an “old-world-style Italian restaurant where mustached waiters dressed in tuxedos serve classic Italian dishes” with “an upscale sports restaurant, targeting sports enthusiasts and serving contemporary American cuisine in a modern setting decorated with flat-screen televisions.” 611 F. 3d at 778. And in Ross Bicycles, we stated that two bicycle companies— which, critically, both sold only bicycles—did not sell similar products due to the differences in the "size of the tubing, the style of wheels, pedals, seats, kickstands, and the difference in the frame angles.” 765 F. 2d at 1507. Both cases were forward-confusion cases in which the parties at issue competed in only one, identical market—restaurants in Tana and bicycles in Ross Bicycles. And because the parties competed in precisely the same market, we focused on granular differences between the products at issue to resolve the claim. Here, by contrast, the claim is one of reverse-confusion, and one of the parties (Amazon) offers products and services across a host of industries and media. The question here is thus whether a reasonable juror could conclude that Am-azon had "bridged the gap” into hardcore pornography streaming and attribute the fireTV and the FyreBoXXX to a single source. [fn¹⁵]: See supra note 10. [fn¹⁶]: See Harvey S. Perlman, The Restatement of the Law of Unfair Competition: A Work in Progress, 80 Trademark Rep. 461, 472 (1990) (“Most surveys do not measure actual confusion. Surveys only give us information about a controlled and artificial world from which we are asked to draw inferences about the real world."). [fn¹⁷]: The magistrate judge (and, by its adoption, the district court) discounted both pieces of evidence, concluding that neither consumer was actually con-fused. As already discussed in footnote⁵, supra, this amounted to an improper credibility determination that invaded the province of the jury. See Mize v. Jefferson City Bd. of Educ., 93 F. 3d 739, 742 (11th Cir. 1996) (“Where the non-movant presents direct evidence that, if believed by the jury, would be sufficient to win at trial, summary judgment is not appropriate even where the movant presents conflicting evidence. It is not the court&#x²⁷;s role to weigh conflicting evidence or to make credibility determinations; the non-movant&#x²⁷;s evidence is to be accepted for purposes of summary judgment.”). [fn¹⁸]: Wreal identified other pieces of evidence to the district court, but our review of the record indicates that they do not represent direct evidence of actual confusion. For example, one of Wreal&#x²⁷;s customers said, “I plan to buy the new Amazon FireTV box (I know it is NOT related to you guys—although confusion over the name may bring Fyretv some more customers and maybe a domain name sale windfall—more power to you!) Will this new Amazon stream-ing device have a private channel installation of FyreTV in the near future?” [fn¹⁹]: See supra note 10.