Pool halls conjure up images of green felt tables, low-hanging lights, and billiard cues. But can a pool hall’s overall image constitute protectable trade dress? To answer that question in this dispute between two pool hall operators, Appellant Clicks Billiards, Inc. (“Clicks”), and its competitor, Sixshooters, Inc. (“Sixshooters”), we must address whether Clicks’ claimed trade dress was nonfunctional; whether Clicks’ trade dress had acquired secondary-meaning; and whether there was a likelihood of confusion between the two establishments. Because there are disputed issues of material fact on each of these three questions, we reverse the district court’s grant of summary judgment in favor of Sixshooters and remand for trial.

Background

Clicks operates a number of billiards establishments in the Southwest, including two in the Phoenix area. Sixshooters opened a single pool hall in the Phoenix area in December 1996, sometime after the two Clicks facilities were already in existence. Daryl Chester, a former Clicks manager who left to work for Sixshooters, was ostensibly involved in the design of Sixshooters both before and after he left Clicks. Clicks presented evidence that Chester and others affiliated with Six-shooters visited Clicks and engaged in detailed inspection and measuring of Clicks’ interior features while the design of Six-shooters was being formulated. Clicks also claims that Sixshooters was built in the path of Clicks’ planned expansion.

In April 1997 Clicks filed suit under the Lanham Act, 15 U.S.C. § 1051 et seq., alleging that Sixshooters infringed its trade dress. In summary judgment proceedings, Clicks asserted that the following individual features of its billiard halls, taken collectively, constitute its trade dress:

Both parties conducted extensive discovery, including depositions of employees and patrons of both establishments. Clicks submitted photographs of Clicks and Sixshooters, and the district court conducted visits to the two Clicks and one Sixshooters pool halls, though not in the presence of the parties or their attorneys. Clicks also submitted a survey purporting to provide support for a finding of secondary meaning and likelihood of confusion. The district court granted summary judgment to Sixshooters, finding that Clicks had not presented sufficient evidence to raise a triable issue of fact on the basic trademark issues of functionality, secondary meaning, and likelihood of confusion.

Standard of Review

We review de novo a grant of summary judgment. Balint v. Carson City, 180 F.3d 1047, 1050 (9th Cir.1999) (en banc). Viewing the evidence in the light most favorable to the nonmoving party (Clicks) and drawing all reasonable inferences in its favor, we must determine whether the district court correctly applied the relevant substantive law and whether there are any genuine issues of material fact. See id.

Analysis

Trade dress refers generally to the total image, design, and appearance of a product and “may include features such as size, shape, color, color combinations, texture or graphics.” Int’l Jensen, Inc. v. Metrosound U.S.A, Inc., 4 F.3d 819, 822 (9th Cir.1993) (internal quotation marks and citations omitted). It is well settled that restaurants and similar establishments may have a total visual appearance that constitutes protectable trade dress. See, e.g., Two Pesos, Inc. v. Taco Cabana, Inc., 505 U.S. 763, 765 , 112 S.Ct. 2753 , 120 L.Ed.2d 615 (1992) (describing protectable trade dress of Mexican restaurant as “a festive eating atmosphere having interior dining and patio areas decorated with artifacts, bright colors, paintings, and murals”); Fuddruckers, Inc. v. Doc’s B.R. Others, Inc., 826 F.2d 837 , 841 (9th Cir.1987) (“[A] restaurant’s decor, menu, layout and style of service may acquire the source-distinguishing aspects of protecta- ble trade dress . ."); cf. Shakey's Inc. v. Covalt, 704 F.2d 426 , 430 (9th Cir.1983) (claimed aspects of pizzeria's trade dress included menu content, prices, pizza ingredients, and "style" of preparation).

To sustain a claim for trade dress infringement, Clicks must prove: (1) that its claimed dress is nonfunctional; (2) that its claimed dress serves a source-identifying role either because it is inherently distinctive or has acquired secondary meaning; and (3) that the defendant's product or service creates a likelihood of consumer confusion. See Disc Golf Ass'n, Inc. v. Champion Discs, Inc., 158 F.3d 1002, 1005 (9th Cir.1998); Fuddruckers, 826 F.2d at 841. The district court held that Clicks did not present evidence sufficient to create a triable issue of fact on any of the three elements. We address them in turn.

I. FUNCTIONALITY

The first hurdle Clicks must overcome is functionality. Trade dress protection extends only to design features that are nonfunctional. As the Supreme Court explained, "A product feature is functional and cannot serve as a trademark if the product feature is essential to the use or purpose of the article or if it affects the cost or quality of the article, that is, if exclusive use of the feature would put competitors at a significant, non-reputation-related disadvantage." Qualitex Co. v. Jacobson Prods. Co., Inc., 514 U.S. 159, 165 , 115 S.Ct. 1300 , 131 L.Ed.2d 248 (1995); accord TrafFix Devices, Inc. v. Marketing Displays, Inc., - U.S. -, 121 S.Ct. 1255, 1259 , 149 L.Ed.2d 164 (2001). Functionality is a question of fact. Fuddruckers, 826 F.2d at 843 (citing Transgo, Inc. v. Ajac Transmission Parts Corp., 768 F.2d 1001, 1028-29 (9th Cir.1983)).

Although trade dress had already been recognized by courts as worthy of protection under the Lanham Act, in 1999 Congress amended the Act to make this statutory protection explicit. And, although the amendment took effect after the filing of this case, it confirms the functionality limitation on trade dress:

15 U.S.C. § 1125 (a)(3). As the Supreme Court observed: "This burden of proof gives force to the well-established rule that trade dress protection may not be claimed for product features that are functional." TrafFix Devices, 121 S.Ct. at 1259 .

This court in Fuddruckers-a case bearing many similarities to this one-shed considerable light on the precise meaning of functionality in the context of a restaurant's trade dress. The court stressed the importance of evaluating the establishment’s “combination of visual elements ‘that, taken together, ... may create a distinctive visual impression.’ ” Fuddruckers, 826 F.2d at 842 (quoting Falcon Rice Mill v. Cmty. Rice Mill, 725 F.2d 336, 346 (5th Cir.1984)). The fact that individual elements of the trade dress may be functional does not necessarily mean that the trade dress as a whole is functional; rather, “functional elements that are separately unprotectable can be protected together as part of'a trade dress.” Id. (quoting Le Sportsac, Inc. v. K. Mart Corp., 754 F.2d 71, 76 (2d Cir.1985)).

To illustrate, many of the elements that constitute part of Clicks’ claimed trade dress are indeed functional; the lamps illuminate the pool tables, and the counters provide a place for customers to place their drinks. Clicks cannot claim a monopoly in using the particular type of lamps or counters with which it furnished its pool halls. Nor can it prevent Sixshooters from copying purely functional aspects of the pool hall. It can, however, “ ‘claim[ ] as its mark the particular combination and arrangement of design elements’ that distinguish it from others using the same concept.” Id. (citing LeSportsac, 754 F.2d at 76 ). As the Fuddruckers court further explained,

Id. at 843 n. 7.

We emphasize here that, in evaluating functionality as well as the other elements of a trade dress claim, it is crucial that we focus not on the individual elements, but rather on the overall visual impression that the combination and arrangement of those elements create. Trade dress is the composite tapestry of visual effects. Courts have repeatedly cautioned that, in trademark-and especially trade dress-cases, the mark must be examined as a whole, not by its individual constituent parts. See, e.g., The Forschner Group, Inc. v. Arrow Trading Co., Inc. 124 F.3d 402, 409 (2d Cir.1997) (“[W]e are not so much concerned with dissecting the competing trade dress and enumerating discrete points of similarity, but rather we focus on the overall image created.”); Union Carbide Corp. v. Ever-Ready, Inc., 531 F.2d 366, 379 (7th Cir.1976) (“[Dissecting marks often leads to error. Words which would not individually become a trademark may become one 'when taken together.”); see also 1 J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 8:2 (4th ed. 2000) (“[T]he issue is not whether defendant’s package or trade dress is identical to plaintiffs in each and every particular. Rather, it is the similarity of the total, overall impression that is to be tested.... ”).

In addressing functionality in its order granting summary judgment, the district court stated, “The overall image of Click’s [sic] is Functional and Purely Aesthetic, not Source Identifying.” We disagree with this formulation of the functionality inquiry. First, trade dress cannot be both “functional and purely aesthetic.” Such a formulation is internally inconsistent and at odds with the commonly accepted view that functionality denotes utility. Nor has this circuit adopted the “aesthetic functionality” theory, that is, the notion that a purely aesthetic feature can be functional. See First Brands Corp. v. Fred Meyer, Inc., 809 F.2d 1378 , 1382 n. 3 (9th Cir.1987) (“In this circuit, the ‘aesthetic’ functionality test has been limited, Vuitton [et Fils S.A. v. J. Young Enters., Inc.], 644 F.2d [769], 773 [ (9th Cir.1981) ], if not rejected, Fa brica, Inc. v. El Dorado Corp., 697 F.2d 890, 895 (9th Cir.1983), in favor of the ‘utilitarian’ functionality test.”); 1 McCarthy § 7:81 (“‘Aesthetic functionality’ is an oxymoron. Ornamental aesthetic designs are the antithesis of utilitarian designs.”); see also Leatherman Tool Group, Inc. v. Cooper Indus., Inc., 199 F.3d 1009, 1013 (9th Cir.2000) (distinguishing functional from “ornamental” features).

We do not accept that there is a dichotomy between “functional” and “source identifying.” Although the evidence relevant to them may overlap, these are two separate aspects of the protecta-bility analysis. Whether a feature is functional has no direct bearing on whether it plays a source-identifying role. The functionality analysis should not be conflated with the inquiry into whether “the purchasing public associates the dress with a particular source.” Fuddruckers, 826 F.2d at 843 (conducting functionality analysis and then conducting separate secondary meaning analysis). Trade dress, of course, can be aesthetically pleasing and still play a source-identifying role; it can also be both aesthetically pleasing and functional. Trade dress cannot, however, be both functional and purely aesthetic.

We consider several factors in determining whether trade dress is functional: “(1) whether the design yields a utilitarian advantage, (2) whether alternative designs are available, (3) whether advertising touts the utilitarian advantages of the design, and (4) whether the particular design results from a comparatively simple or inexpensive method of manufacture.” Disc Golf, 158 F.3d at 1006 . These factors do not translate easily from application in the context of utilitarian product features such as a parabolic chain on a disc golf “hole,” see id., or the configuration of a Swiss Army knife-like tool, see Leatherman, 199 F.3d at 1009 , to application in the context of services, such as a restaurant or a pool hall. Neither is the Fuddruckers case much help in this regard. There the court did not actually evaluate whether the restaurant’s trade dress was functional; it merely held that it was error for the district court not to instruct the jury as to functionality. See Fuddruckers, 826 F.2d at 842-43.

The discussion of functionality in the Fifth Circuit’s Two Pesos opinion may be most instructive. See Taco Cabana Int’l, Inc. v. Two Pesos, Inc., 932 F.2d 1113, 1118-19 (5th Cir.1991), aff'd, 505 U.S. 763 , 112 S.Ct. 2753 , 120 L.Ed.2d 615 (1992). Acknowledging the difficulty of applying traditional functionality analysis in the restaurant context, the Fifth Circuit pointed out that the usual concern motivating the functionality inquiry (i.e., not to allow a merchant to obtain perpetual, patent-like protection for a product feature) is generally not present in this area: “The need to avoid monopolization of a design lessens ... in the area of distinctive trade dress. The wide range of available packaging and design options allows a producer to appropriate a distinctive identity without unduly hindering his competitors’ ability to compete.” Id. at 1119 (quoting Sicilia Di R. Biebow & Co. v. Cox, 732 F.2d 417, 425 (5th Cir.1984)). Thus, in that case, where plaintiff Taco Cabana claimed trade dress protection for the overall image of its restaurant, the fact that many of the individual features of its establishments were functional did not preclude protection, because “Taco Cabana’s particular integration of elements leaves a multitude of alternatives to the upscale Mexican fast-food industry that would not prove confusingly similar to Taco Cabana’s trade dress.” Id.

Here, then, the ultimate issue on functionality is whether Clicks’ “particular integration of elements leaves a multitude of alternatives to the” pool hall “industry that would not prove confusingly similar to” its trade dress. Clicks, we conclude, presented sufficient evidence of the arbitrariness and non-functional nature of its design decisions and the availability of alternative designs to clear the summary judgment hurdle. The district court, even as it came to the conclusion that Sixshoot-ers was entitled to judgment as a matter of law on the functionality issue, identified various arbitrary elements of Clicks’ trade dress, among them the size, placement, and layout of the pool tables; the color combination, including the contrast between the carpet and the dark wood; the lighting; the neon beer signs, bar tap handles, and the like; the cue racks; the selection of video games; the floor covering; the wall treatment; the drink rails; and the millwork. To be sure, many of these elements, considered in isolation, may be functional. The issue, however, is whether, taken as a whole, the overall look and feel of the establishment is functional.

As the district court itself stated, Clicks’ “alleged ‘trade dress’ is purely aesthetic, and part of an overall image of an upscale bar.... Within the overall image are several distinct decisions, such as the darkness of the wood, the placement of various items, color choices and patterns.” The district court concluded that “these decisions are decorative and aesthetic.” Although the court did go on to state as part of the same sentence that “these decisions” are not “source identifying,” as explained above, we address separately the secondary meaning analysis. That the design decisions were made for aesthetic reasons-and not, for example, because they were the only, the cheapest, or the most efficient way to design a pool hall-is evidence of nonfunctionality.

In sum, Clicks presented evidence sufficient to raise an issue of fact that the trade dress of its pool halls is nonfunctional.

II. SECONDARY MEANING

We turn next to analysis of secondary meaning, a term of art for identification of source. As the Supreme Court elaborated in Wal-Mart Stores, Inc. v. Samara Brothers, Inc., 529 U.S. 205, 211 , 120 S.Ct. 1339 , 146 L.Ed.2d 182 (2000), “a mark has acquired distinctiveness, even if it is not inherently distinctive, if it has developed secondary meaning.” “The trade dress of a product or service attains secondary meaning when the purchasing public associates the dress with a particular source.” Fuddruckers, 826 F.2d at 843. “[T]he elements making up the alleged trade dress must have been used in such a manner as to denote product source. Thus, a product feature whose only impact is decorative and aesthetic, with no source-identifying role, cannot be given exclusive rights under trade dress law.” 1 McCarthy § 8:1 (quoted in Stephen W. Boney, Inc. v. Boney Servs., Inc., 127 F.3d 821, 828 (9th Cir.1997)). “Whether a particular trade dress has acquired secondary meaning is a question of fact....” First Brands, 809 F.2d at 1383 .

Clicks introduced substantial evidence tending to show that the trade dress of its pool halls had attained secondary meaning. Chief among this evidence was a survey commissioned by Clicks purporting to demonstrate that patrons associated the look of Clicks with that establishment and not others. Such surveys, while often subject to criticism and varying interpretations, are a routine and well-established feature of trademark practice. “Surveys in trademark cases may be considered so long as they are conducted according to accepted principles.” Stuhlbarg Int'l Sales Co., Inc. v. John D. Brush & Co., Inc., 240 F.3d 832, 840 (9th Cir.2001); see also E. & J. Gallo Winery v. Gallo Cattle Co., 967 F.2d 1280, 1292 (9th Cir.1992) (“surveys in trademark cases are to be admitted as long as they are conducted according to accepted principles.” (emphasis added)); Vision Sports, Inc. v. Melville Corp., 888 F.2d 609, 615 (“An expert survey of purchasers can provide the most persuasive evidence of secondary meaning.”). “Technical unreliability goes to the weight accorded a survey, not its admissibility.” Prud ential Ins. Co. v. Gibraltar Fin. Corp., 694 F.2d 1150, 1156 (9th Cir.1982).

Clicks submitted a survey prepared by Dr. Sandra Cogan, president of Cogan Research Group. According to Dr. Cogan’s resume, she has conducted surveys in over eighty trademark cases and has testified in eight cases within the Ninth Circuit. In this survey, patrons at a Clicks establishment were asked, “Do you associate the visual look and appearance of this billiard parlor with Clicks Billiards only or with other billiard parlors too?” Eighty percent of respondents said that they associated the look and feel with Clicks only; eighteen percent said, “Other billiard parlors too.”

The district court stated, “This survey is not without flaws,” and ultimately concluded, “Put bluntly, the survey is of little or no value.” The court was troubled by the fact that in response to a follow-up question, respondents often used vague and general language in describing Clicks’ image and made reference to unprotectable features of Clicks’ interior, such as its cleanliness, upkeep, good service, and music. This is a valid observation and may indeed be true, but the issue is not one subject to resolution on summary judgment. It is true that some respondents used such vague language and referred to unprotectable elements of Clicks’ trade dress, but other responses tend to support Clicks’ contention that patrons specifically associate its image with Clicks and not with other pool halls. Among the responses associating Clicks with elements, that, when taken as a whole, may constitute protectable trade dress: “Different from other billiard parlors”; “nice layout”; “lots of room between tables”; “general decor, pictures, signs”; “the matchbooks, the, uh, pictures on the wall.... Logos, the well-placed logos”; “the tables, more space, the TVs”; “The appearance and position of the bar”; “They have high ceilings, pool tables in a line, and the size. Large size. There are two levels.”; “Step down to play pool. Bar is round and stools”; “Two levels, bar is unique, entrance.”; “[T]he felt on the wall.... Felix the Cat. The matches on top of the ashtrays. No other place does that.... These black dots, on the floor, you know, like metal, but this is a rubber floor. ... The way the lights hang down. Most places have track lighting, one long light.”

Ultimately these responses may not add up to much, and a reasonable juror might ultimately agree with the district court that the survey did not prove that Clicks’ trade dress had acquired secondary meaning. But the juror could only have done so after considering conflicting evidence and deciding what weight to accord the survey and the follow-up interviews. This undertaking describes the proper role for a trier of fact; it is not the role of a district court at the summary judgment stage where the issue is whether a triable issue of fact even exists. See, e.g., Summers v. A. Teichert & Son, Inc., 127 F.3d 1150, 1152 (9th Cir.1997) (“The court must not weigh the evidence or determine the truth of the matters asserted but must only determine whether there is a genuine issue for trial.”).

Here, the district court did not exclude the survey on the ground that it was irrelevant or that it was undermined by some other fatal flaw. Rather, the court appeared to admit the survey into evidence and then to analyze what it considered to be its deficiencies. Treatment of surveys is a two-step process. First, is the survey admissible ? That is, is there a proper foundation for admissibility, and is it relevant and conducted according to accepted principles? See Wendt v. Host Int’l, Inc., 125 F.3d 806 , 814 (9th Cir.1997); E. & J. Gallo Winery, 967 F.2d at 1292-93 . This threshold question may be determined by the judge. Once the survey is admitted, however, follow-on issues of methodology, survey design, reliability, the experience and reputation of the expert, critique of conclusions, and the like go to the weight of the survey rather than its admissibility. Stuhlbarg, 240 F.3d at 840 . These are issues for a jury or, in a bench trial, the judge. In laying this guide for analysis of survey evidence, we are not endorsing the survey here nor are we prejudging its fate at trial. Nor are we suggesting that the presence of a survey will always preclude summary judgment. We are simply saying that the survey admitted by the district court raises questions of material fact with respect to secondary meaning that may not be disposed of by the court alone.

Aside from the survey, Clicks presented evidence that Sixshooters intention-álly copied Clicks’ trade dress. “Our cases recognize that evidence of deliberate copying is relevant to a determination of secondary meaning. E.g. Transgo, 768 F.2d at 1015-16 .... Indeed, in appropriate circumstances, deliberate copying may suffice to support an inference of secondary meaning. See [id] at 1016 (dictum).” Fuddruckers, 826 F.2d at 844; Lisa Frank, Inc. v. Impact Int’l, Inc., 799 F.Supp. 980, 989 (D.Ariz.1992); see also Vision Sports, 888 F.2d at 615 (“[W]e have held that proof of copying strongly supports an inference of secondary meaning.”).

As the district court stated in its summary judgment order, “One former Six-shooters employee, employed by Clicks prior to joining Sixshooters, testified that Daryl Chester stated that he wanted Six-shooters to look like Clicks because the ‘look worked.’ ” Clicks also presented evidence that Chester, while working at Six-shooters, greeted a Clicks manager who came to visit him, “Welcome to Clicks West.” When asked if he knew what Chester meant, the manager said the comment referred to “the look” and that he thought Sixshooters looked like Clicks. Clicks employee Shawnna Gereg stated in her deposition that the Forbes brothers (the owners of Sixshooters) visited Clicks and actually took measurements of various features, including the foosball areas, the bar, and the fighting. They also examined other details of the interior, such as the rails and the woodwork. Clicks employee Roger Collett testified that the Forbes brothers brought samples of coun-tertops into Clicks to compare them with those at Clicks and took measurements of cabinets above the bar. Clicks manager Tanya Mueller stated that she once saw the Forbes brothers in Clicks before opening time (apparently accompanied by Chester) examining and measuring fighting fixtures, counter tops, shelves, and the distance between pool tables. A Clicks patron confirmed that she saw the Forbes brothers come into Clicks and examine various aspects of the pool hall, specifically the tables, walls, and bar stools.

The district court’s response to such extensive evidence of intentional copying was to deem it “slight” and “contested,” and to note that it “was controverted in the record.” No doubt, the statements at issue are subject to various interpretations. But the very fact that the court acknowledged that the evidence of copying was “contested” and “controverted” should have precluded summary judgment on this point. On summary judgment, the district court is prohibited from weighing evidence and deciding issues of contested fact.- Even if we accept that Sixshooters presented a version of events that cast considerable doubt on Clicks’ allegations of copying, it is not the role of the district court at this stage to decide which party has the more compelling argument; rather, it is simply to determine whether a triable issue of fact exists. Clicks has come forward with sufficient evidence of secondary meaning in the form of both a consumer survey and testimony from various witnesses to defeat summary judgment on this point.

III. Likelihood of Confusion

Likelihood of confusion, a question of fact, Levi Strauss & Co. v. Blue Bell, Inc., 778 F.2d 1352 , 1356 n. 5 (9th Cir.1985), is “the most important element” of the three components of a protectable mark. Kendall-Jackson Winery Ltd., v. E. & J. Gallo Winery, 150 F.3d 1042, 1048 (9th Cir.1998). Indeed, we have observed that “trial courts disfavor deciding trademark cases in summary judgments because the ultimate issue is so inherently factual.... Additionally, the question of likelihood of confusion is routinely submitted for jury determination as a question of fact.” Levi Strauss, 778 F.2d at 1355 n. 5 (citations omitted); see also 5 McCarthy § 32:119.

“Likelihood of confusion exists when customers viewing the mark would probably assume that the product or service it represents is associated with the source of a different product or service identified by a similar mark.” Fuddruckers, 826 F.2d at 845 (internal quotation marks and emphasis omitted). “The factual elements that make up likelihood of confusion include evidence of actual confusion, the defendant’s intent in adopting the mark, similarity of marks, similarity of goods and marketing channels, and the strength of the mark.” Id. (citing Park ‘N Fly, Inc. v. Dollar Park and Fly, Inc., 782 F.2d 1508, 1509 (9th Cir.1986)).

Clicks has presented significant evidence of actual confusion, which, according to our precedent, constitutes “persuasive proof that future confusion is likely.” Id. at 845. First, the Cogan survey tends to support a conclusion of actual confusion. Respondents were shown a series of photographs of Sixshooters and asked, “Do you think that the billiard parlor in these photos is owned or operated by Clicks Billiards or do you think that the billiard parlor in the photos is owned and operated-by a different company?” Seventy-two percent stated that they believed Clicks owned or operated the establishment (Sixshooters) shown in the photos. On top of that seventy-two percent, thirteen percent believed that the pool hall in the photos was “licensed by, endorsed by, authorized by or in [some] way connected with Clicks Billiards.” Of the eighty-five percent of respondents expressing apparent confusion as to ownership or affiliation, twenty-eight percent had actually visited Sixshooters.

Clicks also introduced testimony from a variety of people reporting their impressions about the similarities between Clicks and Sixshooters. Several witnesses made statements that appear to indicate actual confusion. For example, one Clicks patron was asked, “Had you ever before anybody talked to you about this ever made any assumption or thought process that the two of them [Clicks and Sixshooters] might have been the same place?” He responded, “The first time I walked into Sixshoot-ers.” Another patron was asked, “While you were actually in Sixshooters those two times, did you actually think at any time, ‘Boy, I might be in a Clicks’?” He responded, “Yeah.” A third patron was asked, “When you went to Sixshooters the first time, did you have any initial impressions at Sixshooters?” He responded, “I thought it was Clicks under a different name.”

In sum, Clicks presented sufficient evidence of likelihood of confusion between the trade dress of Clicks and Sixshooters to defeat summary judgment on the confusion issue. A jury, of course, may reject such evidence or determine that the witnesses are not credible, but a court on summary judgment may not make such determinations; instead it must draw all reasonable inferences in favor of the non-moving party, i.e. Clicks.

In addition, there was evidence of intent to copy, yet another of the factors tending to show likelihood of confusion. Fuddruckers, 826 F.2d at 854. “A showing that the defendant intended to adopt the plaintiffs trade dress is ... entitled to great weight because a defendant is presumed able to accomplish this purpose.” Id. at 845-46 (internal quotation marks and citation omitted). But the district court failed to accord “great weight” to the evidence presented by Clicks suggesting that such copying did occur. The district court brushed aside the extensive evidence presented by Clicks that Sixshooters employees conducted detailed inspection of Clicks and even went so far as to measure various features of the interior. Given that, on summary judgment, the court was required to draw all reasonable inferences in Clicks’ favor, see Balint, 180 F.3d at 1050 , the court was required to draw the inference that Sixshooters did engage in copying. The district court did acknowledge in its summary judgment order that “[t]here is conflicting evidence of intent to deprive Clicks of the benefit of its reputation.” .Consequently, no conclusion can be drawn from such “conflicting evidence.” Perhaps at trial, Sixshooters would be able to convince the jury of an innocent explanation for such conduct, or successfully dispute that it ever actually took place, but that is a factual determination reserved for trial, not for resolution on summary judgment.

The other likelihood of confusion factors-similarity of marks, similarity of goods and marketing channels, and the strength of the mark-do not bear greatly on our analysis in light of the factors discussed above. Clicks did present testimony from a variety of patrons who stated that they believed that the interiors of Clicks and Six-shooters looked similar. The “goods,” or, in this case, services, are indisputably similar; they are both billiards halls. Clicks did not present significant relevant evidence regarding marketing. And, finally, as the district court noted, Clicks has conceded that, given the nature of the business at issue, its mark is not particularly strong. Based on a review of the evidence, we conclude that Clicks has cleared the threshold of demonstrating that there is genuine issue of fact with respect to likelihood of confusion.

IY. Judicial Visit

Finally, we address Clicks’ argument that the district court’s visit to the three establishments at issue, without the parties or counsel present, constituted a denial of Clicks’ right to due process.

We first wish to make clear that, in a trade dress case, there is nothing wrong with a judge examining the actual dress at issue. Surely if this were a case about the outside of a cereal box instead of the inside of a pool hall, no one would object to the judge examining the actual box instead of pictures of it. If a picture is worth a thousand words, then the real thing is worth a thousand pictures. And if the judge were to take the box from the courtroom back to his chambers for additional study, he would be under no obligation to invite counsel to accompany him. Neither is there anything wrong with official excursions by a judge or jury to view evidence that simply cannot, because of physical limitations, be brought into a courtroom.

The situation here, however, is more troublesome. When a judge takes exhibits back to chambers to examine them, the parties have seen the same exhibits and know exactly what the court is reviewing. Here, however, because counsel were not informed when the judge would be conducting his inspection, the parties had no way of knowing exactly what the court looked at, or how the judge went about his visit. In short, the court’s site visit improperly exposed him to factual evidence not part of the record. See, e.g., Pina v. Henderson, 752 F.2d 47, 50 (2d Cir.1985) (“A court should not go outside the record to supply a fact that is an essential part of a party’s case unless the fact is clearly beyond dispute.”); May hue v. City of Plantation, 375 F.2d 447 , 451 n. 6 (5th Cir.1967) (“A judge is not to use ... those facts that he may know only as an individual observer outside of the courtroom.”); Alvary v. United States, 302 F.2d 790, 794 (2d Cir.1962) (“Although it is proper, and all too frequently necessary, for a judge to do independent research on questions of law, the [matters at issue] are questions of fact. On fact questions the court should not ... go outside the record unless the facts are matters of common knowledge or are capable of certain verification.”).

This concern is especially heightened in this case, given Clicks’ accusations in the record of spoliation of evidence, specifically that Sixshooters changed the appearance of its interior during the course of the litigation. And the court itself challenged the photographic depictions offered by Clicks. Thus, though the district court stated that the conclusions in its summary-judgment order were based at least in part on its inspection of the pool halls, we have no way of knowing precisely what the pool halls looked like at the time of the visit and how that view compared to the evidence in the record. All of this is another way of saying that the record on appeal is incomplete. At a minimum, the district judge should have given the parties and their counsel the opportunity to accompany him.

In this case, there is conflicting evidence whether Clicks actually consented to the court’s visit to the pool halls. For that reason, and because we conclude for other reasons that summary judgment was not appropriate, we decline to reach the issue whether Clicks’ due process rights were violated under these circumstances.

Conclusion

This case underscores our warning that “trial courts disfavor deciding trademark cases in summary judgments because the ultimate issue is so inherently factual.” Levi Strauss, 778 F.2d at 1355 n. 5. Clicks has presented evidence sufficient to clear the summary judgment hurdle. We emphasize that we do not hold here that Clicks has necessarily established nonfunc-tionality, secondary meaning, and likelihood of confusion, but only that there remain genuine disputes of material fact as to those issues.

REVERSED AND REMANDED.

. That trade dress can be "inherently distinc.-tive" without acquiring secondary meaning was established in Two Pesos, 505 U.S. at 776 , 112 S.Ct. 2753 . Clicks has not claimed that its trade dress is inherently distinctive, only that it has acquired secondary meaning. . Fuddruckers is a national chain of "upscale" hamburger restaurants. Its claimed trade dress, as characterized by the court, includes elements such as: a visible food preparation area; glass display cases; the disolay of food in bulk nackaees: stacked cartons of beer, produce, and other food items; iced tea served in containers that "look[ ] suspiciously like a garbage can"; a self-serve condiment bar featuring "black, round crocks"; extensive use of tiling; neon signs; a yellow awning; director's chairs; mirrors; a bakery area labeled "Mother Fuddruckers"; a restaurant "newspaper" at each table; paper-lined beer trays used as serving platters; an ice-cream display case; French doors between the dining rooms; large ceiling lamps; potted floor lamps; "black oversized napkin holders stacked on top of each other”; "a stainless steel bin used to dispense iced beverages”; and ceiling fans. Fuddruckers, 826 F.2d at 839-40, 840 n. 1. . The Supreme Court recently decided an appeal in a related case, Cooper Indus., Inc. v. Leatherman Tool Group, Inc., - U.S. -, 121 S.Ct. 297 , 148 L.Ed.2d 238 (2001) (certio-rari granted with respect to punitive damages issue only). . The Supreme Court in Two Pesos did not address functionality but "made the explicit analytic assumption that the trade dress features in question ... were not functional." TrafFix Devices, 121 S.Ct. at 1262 . . The Fifth Circuit upheld a jury instruction that read: The law allows the copying of functional features in the public interest of enhancing competition.... Even if the trade dress is made up of individual elements, some of which serve a functional purpose, the trade dress may be protectable so long as the combination of these individual elements which define Taco Cabana’s trade dress taken is arbitrary. On the other hand, if you find that Taco Cabana's trade dress taken as a whole must be used by others in order to compete in the Mexican fast-food restaurant business, then you should find that Plaintiff's trade dress is functional and not protectable. [T]he inquiry into whether Taco Cabana's trade dress is functional or non-functional should not be addressed to whether individual elements fall within the definition, but whether the whole collection of elements taken together are functional or non-functional. Two Pesos, 932 F.2d at 1119 . Though it generally approved this instruction, the Fifth Circuit expressed concern about the statement that protection must be denied on functionality grounds where the particular combination "must be used by others in order to compete.” Id. (emphasis added). The proper standard, rather, is that “It should suffice for a finding of functionality if protecting the trade dress threatens to eliminate a substantial swath of competitive alternatives in the relevant market.” Id. at 1119 n. 6 (citing Sicilia, 732 F.2d at 427 ) ("A design would be considered de jure functional if it is the best or one of a few superior designs available.” (internal quotation marks omitted)). . "What is it about the visual look and appearance of this billiard parlor that makes you associate it with Clicks Billiards only?” This question was only asked of respondents who said that they associated the look of Clicks only with Clicks and not with other pool halls. . Much of this testimony delivers much less than promised. The vast majority of the statements consist solely of the deponent’s belief that Clicks and Sixshooters look similar. That is not the same as the deponent being confused as to the source of the trade dress; if two products look somewhat the same, but consumers suffer no confusion, there can be no successful trademark infringement claim. Though most of the cited deposition testimony does not indicate actual confusion, it still may, however, be probative of likelihood of confusion, because it indicates that significant numbers of patrons perceived a strong similarity between the trade dress of the two establishments. . We wish to underscore that we do not determine here whether such accusations have any merit. . It appears that the court and counsel discussed this matter on a conference call that was not transcribed. On appeal, the parties dispute whether Clicks consented to the court’s visit.