NEWMAN, POOLER and KATZMANN, Circuit Judges.

POOLER, Circuit Judge.

Plaintiff-appellant Star Industries appeals from a judgment for defendants-ap-pellees, entered in the United States District Court for the Southern District of New York (Harold Baer, Jr., Judge), on January 30, 2004, after a bench trial, dismissing its federal claims under the Lan-ham Trade-Mark Act, 15 U.S.C. §§ 1051-1129, and pendent state law claims under N.Y. General Business Law §§ 360-1 & 349 and New York common law. Star also appeals from the district court’s order entered March 23, 2004, denying its motion to reopen the trial record and to amend the findings of fact and conclusions of law. For the reasons stated below, we affirm.

BACKGROUND

Plaintiff-appellant Star Industries, Inc. (“Star”) produces alcoholic beverages including, of particular relevance to this appeal, the “Georgi” brand of vodkas. Geor-gi vodka is sold primarily in New York state, and is one of the top selling vodkas in the New York metropolitan area. It is generally cheaper than the leading nationally distributed vodkas. Defendants-ap-pellees Bacardi & Co. Ltd. and Bacardi U.S.A. (collectively “Bacardi”) produce, import, and distribute “Bacardi” brand rums. Bacardi is the largest selling brand of hard liquor in the United States. Defendant-appellee Anheuser-Busch, Inc. is one of the leading producers of beers and malt beverages in the United States. In 2001 Anheuser-Busch and Bacardi entered into an exclusive trademark licensing agreement under which Anheuser-Busch produces malt beverages bearing the Bacardi logo.

In June 1996, inspired by the success of flavored vodkas introduced by leading international companies such as Stolichnaya, Star’s president decided to develop an orange-flavored Georgi vodka. A new label was designed, consisting of the traditional Georgi label, which contains a coat of arms and a logo consisting of stylized capital letters spelling ‘Georgi’ on a white background, together with three new elements: an orange slice, the words “orange flavored,” and a large elliptical letter “O” appearing below the “Georgi” logo and surrounding all of the other elements. The “O” was rendered as a vertical oval, with the outline of the “O” slightly wider along the sides (about one quarter inch thick) and narrowing at the top and bottom (about one eighth inch thick); the outline of the “O” is colored orange and decorated with two thin gold lines, one bordering the inside and one bordering the outside of the outline. Star was apparently the first company to distribute an orange-flavored alcoholic beverage packaged in a bottle bearing a large elliptical orange letter “O.”

Star’s hope was that, just as consumers of orange-flavored Stolichnaya vodka had begun referring to the vodka as “Stoli 0,” consumers would come to refer to orange-flavored Georgi vodka as “Georgi 0.” Accordingly, Star applied in 1996 to register “Georgi 0” as a word trademark with the *380 U.S. Patent and Trademark Office (“PTO”). Reasoning that consumers viewing the Georgi 0 label were likely to perceive the word “Georgi” as separate from the “0” design, and not as constituting a composite phrase, the PTO rejected Star’s application. Between 1996 and 2002 Star promoted its new orange-flavored vodka, although it is unclear how vigorously, and it apparently had little success. Sales volume remained low even in New York state, and practically nil elsewhere. No evidence in the record suggests that Georgi vodka has ever been commonly referred to by consumers as “Georgi 0.”

In 2000 Bacardi began to develop an orange-flavored rum, which it ultimately introduced nationally in 2001 under the name “Bacardi O.” Bacardi’s line of flavored rums originated in 1995 with “Bacardi Limón.” Unlike Bacardi’s other flavored rums, however, Bacardi 0 was produced and marketed bearing a distinct label consisting of the Bacardi logo and bat symbol above a large elliptical letter “0” against a clear background. This Bacardi “0” design was developed by a New York design firm, whose President and Creative Director supervised the design of the new label and claimed not to have had any awareness of the Georgi “0” design until after the instant lawsuit was filed. Like the Georgi “0,” the Bacardi “0” was decorated with gold bordering, was colored in orange, and spanned most of the height of the bottle. Its appearance differed in certain respects from the Georgi “0” design, including lesser elongation, greater thickness, variable shading, and the use of a brighter variety of orange. Star presented some evidence at trial suggesting that Bacardi’s local sales representatives and regional managers were aware of Geor-gi’s product and its appearance prior to the development of the Bacardi “0” design, but no evidence of such awareness by those involved in the development of the new “0” design.

In March 2003 Anheuser-Bush launched a new orange-flavored version of its “Bacardi Silver” malt beverage, under the name “Bacardi Silver O 3 .” This product also contained the Bacardi “0” design on the label. Under New York state law, malt beverages and hard liquors may not be sold in the same stores, malt beverages being sold primarily in grocery stores while hard liquors are sold in liquor stores.

In support of its contention that consumers and tradespeople were confused by the similar labels on Bacardi’s and Georgi’s orange-flavored beverages, Star presented the following evidence at trial: the testimony of a Star Vice President that she once overheard someone standing outside a Dunkin Donuts casually comment to a friend that Georgi’s orange-flavored vodka was a Bacardi product; the testimony of an employee of Star’s distributor that he himself had assumed on first seeing the Bacardi label that Bacardi must have obtained Star’s permission to use its “0” design, and that he once had a conversation with a liquor store owner about the potential for consumers to be confused by the similar labels; and the testimony of a Star employee that he once saw a consumer at a liquor store select a bottle of Bacardi 0 in the belief that it was Georgi orange-flavored vodka. Star presented no consumer surveys or studies, while Bacardi and Anheuser-Busch each proffered a survey suggesting that there was no consumer confusion.

After sending a cease and desist letter to Bacardi in September 2001, Star filed the instant lawsuit in May 2002, amending its complaint to add Anheuser-Busch as a defendant in May 2003. In June 2003 Star applied with the PTO to register its “0” *381 design. 1 A bench trial was held in the district court in July and August of 2003. At closing argument on August 14, 2003, the district court inquired as to the status of Star’s PTO application, which remained pending. The district court rejected Star’s claims in their entirety by order dated December 31, 2003, holding Star’s “O” design not protectable as a trademark, and holding in the alternative that, assuming the Georgi “O” design was a protectable trademark, such trademark was not infringed because Star’s product was not likely to be confused with Bacardi’s despite the similar “O” designs. Judgment was entered for defendants-appellees on all claims on January 30, 2004. On February 6, 2004, the PTO approved Star’s application to register its “O” design. On February 17, 2004, Star moved in the district court to reopen the trial record to include evidence of the PTO’s decision, seeking new conclusions of law that the design was protectable as a trademark and that there was a likelihood of confusion. The district court denied Star’s motion by order dated March 19, 2004. Star filed notices of appeal from the judgment on February 17, 2004, and from the denial of its motion to reopen on April 2, 2004. Star asks this Court to enter a permanent injunction and remand for calculation of damages, or in the alternative to remand for reconsideration after supplementing the record with evidence of the PTO decision.

We conclude that the district court erred in holding Star’s “O” design not protecta-ble as a trademark. We therefore need not consider whether the district court erred in denying Star’s motion to reopen the trial record. However, as we agree with the district court that Star has not established a likelihood of confusion with Bacardi’s product, we affirm the district court’s judgment for appellees.

DISCUSSION

“[T]o succeed in a Lanham Act suit for trademark infringement, a plaintiff has two obstacles to overcome: the plaintiff must prove that its mark is entitled to protection and, even more important, that the defendant’s use of its own mark will likely cause confusion with plaintiffs mark.” Gruner + Jahr USA Publ’g v. Meredith Corp., 991 F.2d 1072, 1074 (2d Cir.1993). We discuss these two elements of trademark infringement in turn.

I. Protectability as a Trademark

At the time of judgment, Star’s “O” design had not yet been approved for registration as a trademark by the PTO. An unregistered mark is entitled to protection under the Lanham Act if it would qualify for registration as a trademark. See Courtenay Communications Corp. v. Hall, 334 F.3d 210, 214 n. 2 (2d Cir.2003). To qualify for registration a mark must be sufficiently “distinctive” to distinguish the registrant’s goods from those of others. See id. Such distinctiveness may be demonstrated in either of two ways. The mark may be “inherently distinctive” if its intrinsic nature serves to identify its particular source. Alternatively, even if not inherently distinctive, the mark may be distinctive by virtue of having acquired a “secondary meaning” in the minds of consumers. See Two Pesos, Inc. v. Taco Cabana, Inc., 505 U.S. 763, 768-79, 112 S.Ct. 2753, 120 L.Ed.2d 615 (1992); Hall, 334 F.3d at 214 n. 2.

On appeal, Star does not purport to rely on a “secondary meaning” acquired *382 by its “0” design, relying instead on inherent distinctiveness. “Common basic shapes” or letters are, as a matter of law, not inherently distinctive. See Seabrook Foods, Inc. v. Bar-Well Foods, Inc., 568 F.2d 1342, 1344 (C.C.P.A.1977) (implying that “ ‘common’ basic shape or design” is not inherently distinctive); Permatex Co. v. California Tube Prods., Inc., 175 U.S.P.Q. 764, 766 (T.T.A.B.1972) (stating that “common basic shapes ... used as vehicles for the display of word or letter marks” are not inherently distinctive). However, stylized shapes or letters may qualify, provided the design is not commonplace but rather unique or unusual in the relevant market. See, e.g., id. (holding swirly circle design protectable). 2 The guiding principle in distinguishing protect-able from unprotectable marks is that no one enterprise may be allowed to attain a monopoly on designs that its competitors must be able to use in order to effectively communicate information regarding their products to consumers. See Restatement (Third) of Unfair Competition § 13 cmts. b & d; cf. Virgin Enters. Ltd. v. Nawab, 335 F.3d 141, 147 (2d Cir.2003) (explaining that a merchant who sells pencils labeled “Pencil” and seeks to exclude other merchants from using that word “is seeking an advantage the trademark law does not intend to offer. To grant such exclusivity would deprive the consuming public of the useful market information it receives where every seller of pencils is free to call them pencils.”). Trademark protection of a sufficiently stylized version of a common shape or letter will not hamper effective competition because competitors remain free to use nonstylized forms or their own alternative stylizations of the same shape or letter to communicate information about their products.

A design used in conjunction with other marks is separately protectable in its own right if it creates a separate and distinct impression from the impression created by the other marks. This will be true if either the mark is itself inherently distinctive, see, e.g., In re Servotronics, Inc., 156 U.S.P.Q. 592, 592 (T.T.A.B.1968) (holding stylized letter “S” separately protectable from word in which it appeared); In re W.B. Roddenbery Co., 135 U.S.P.Q. 215, 216 (T.T.A.B.1962) (holding combination of gold circle and colored rectangle, utilized as background to applicant’s advertisement for sour pickles, registrable separately from foreground elements), or if the consuming public has come to associate the separate mark in itself with the particular product, vesting it with its own distinct secondary meaning, see Restatement (Third) of Unfair Competition § 13 cmt. d.

The district court held Star’s “0” design not protectable on two grounds. First, the court held that the mark was not protectable separately from the other material appearing on the Georgi 0 label on the ground that there was a lack of evidence that it produced a separate impression among consumers. Second, the court held that the mark was not protectable because it was not inherently distinctive and Star had not demonstrated secondary meaning. A district court’s findings that a mark is not protectable as inherently distinctive is a finding of fact that we generally review for clear error. See Bristol-Myers Squibb Co. v. McNeil-P.P.C., Inc., 973 F.2d 1033, 1039-40 (2d Cir.1992). However, where the district court “base[d] *383 its findings upon a mistaken impression of applicable legal principles,” we are “not bound by the clearly erroneous standard.” Inwood Labs., Inc. v. Ives Labs., Inc., 456 U.S. 844, 855 n. 15, 102 S.Ct. 2182, 72 L.Ed.2d 606 (1982).

The district court erred when it described the Star “O” as a basic geometric shape or letter, and therefore rejected inherent distinctiveness and required a showing of secondary meaning. The Star “0” is not a “common basic shape” or letter, and the district court’s holding to the contrary was premised on a misunderstanding of this trademark law concept. Unshaded linear representations of common shapes or letters are referred to as “basic.” They are not protectable as inherently distinctive, because to protect them as trademarks would be to deprive competitors of fundamental communicative devices essential to the dissemination of information to consumers. However, stylized letters or shapes are not “basic,” and are protectable when original within the relevant market. See Hall, 334 F.3d at 215 n. 3 (distinguishing case of mark consisting of word displayed with distinctive “typeface, color, and other design elements,” which was protectable, from cases holding generic words not protectable); compare W.B. Roddenbery Co., 135 U.S.P.Q. at 216 (holding design consisting of colored circle attached to differently colored rectangle protectable as inherently distinctive) with In re Hillerich & Bradsby Co., 40 C.C.P.A. 990, 204 F.2d 287, 288 (1953) (noting that applicant conceded that unshaded line oval was not inherently distinctive). Star’s “O” is sufficiently stylized to be inherently distinctive and therefore protectable as a trademark. It is stylized with respect to shading, border, and thickness, and each of these design elements distinguishes it from the simple or basic shapes and letters that have been held unprotectable.

The Star “0” design had sufficient shape and color stylization to render it slightly more than a simply linear representation of an ellipse or the letter “0.” It was, furthermore, a unique design in the alcoholic beverage industry at the time it was introduced. This suffices to establish its inherent distinctiveness and thus its protectability. Furthermore, the Star “0” design is protectable separately from the other design elements on the Georgi orange-flavored vodka label precisely because the “0” design is itself inherently distinctive. See In re E.J. Brach & Sons, 45 C.C.P.A. 998, 256 F.2d 325, 327 (1958); W.B. Roddenbery, 135 U.S.P.Q. at 216. However, the extent of stylization was marginal at best. The outline of the “0,” though not uniform, is ordinary in its slightly varying width, and the interior and exterior borders are also ordinary. The result is a “thin” or weak mark, which will be entitled to only limited protection. See Libman Co. v. Vining Indus., 69 F.3d 1360, 1363 (7th Cir.1995).

II. Likelihood of Confusion

To prevail in a trademark infringement action under the Lanham Act, a plaintiff must prove, in addition to pro-tectability of the mark, “a probability of confusion, not a mere possibility,” affecting “numerous ordinary prudent purchasers.” Gruner + Jahr, 991 F.2d at 1077. “Likelihood of confusion includes confusion of any kind, including confusion as to source, sponsorship, affiliation, connection, or identification.” Guinness United Distillers & Vintners B.V. v. Anheuser-Bush, Inc., 64 U.S.P.Q.2d 1039, 1041 (S.D.N.Y.2002), citing McDonald’s Corp. v. McBa-gel’s, Inc., 649 F.Supp. 1268, 1273 (S.D.N.Y.1986). “In order to be confused, a consumer need not believe that the own *384 er of the mark actually produced the item and placed it on the market. The public’s belief that the mark’s owner sponsored or otherwise approved the use of the trademark satisfies the confusion requirement.” Dallas Cowboys Cheerleaders, Inc. v. Pussycat Cinema, Ltd., 604 F.2d 200, 204-05 (2d Cir.1979) (internal citations omitted).

Star asserts not that consumers were likely to be confused into believing that its product was Bacardi rum or that Bacardi O or Bacardi Silver O3 was Georgi vodka, but rather, that consumers were likely to be confused into believing that there was an affiliation, sponsorship, or other connection between the companies’ products. The district court’s analysis did not distinguish confusion as to identification or source from the kind of confusion relied on by Star, as to affiliation, sponsorship, or connection. Thus the district court held incorrectly that confusion was impossible as to Bacardi Silver O3 because it was not sold in the same stores as Georgi vodka. The district court also relied excessively on its observation that no sober consumer would purchase rum labeled “Bacardi” believing it to be vodka labeled “Georgi,” despite the presence of similar orange ovals on both labels. The district court’s analysis misses the point of an affiliation confusion claim: if Star’s “O” design is, as we have held above, inherently distinctive and therefore a unique and recognizable identifier of Georgi’s product, then the inclusion of a similar design on another company’s product may lead consumers to infer a relationship between the new product and Star’s.

To determine whether there is a likelihood of confusion, we apply the eight-factor Polaroid balancing test introduced by Judge Friendly in Polaroid Corp. v. Polarad Electronics Corp., 287 F.2d 492 (2d Cir.1961). The district court’s resolution of each separate factor is treated as a finding of fact which we review for clear error, while its balancing of the factors is treated as a matter of law subject to de novo review. See Gruner + Jahr, 991 F.2d at 1077. Our analysis is not mechanical, but rather, focuses on the ultimate question of whether, looking at the products in their totality, consumers are likely to be confused. See Nora Beverages, Inc. v. Perrier Group of Am., Inc., 269 F.3d 114, 119 (2d Cir.2001). The eight factors are: (1) strength of the trademark; (2) similarity of the marks; (3) proximity of the products and their competitiveness with one another; (4) evidence that the senior user may “bridge the gap” by developing a product for sale in the market of the alleged infringer’s product; (5) evidence of actual consumer confusion; (6) evidence that the imitative mark was adopted in bad faith; (7) respective quality of the products; and (8) sophistication of consumers in the relevant market. We discuss these factors in turn.

A. The Polaroid Factors

1. Strength

The district court did not explicitly discuss the “strength” factor of the Polaroid test, and so there appears to be no finding of fact on this issue to which we must defer. The strength of a mark is determined by its tendency to uniquely identify the source of the product. This tendency is strong to the extent that the mark is distinctive, either inherently or by virtue of having acquired secondary meaning. See Savin Corp. v. Savin Group, 391 F.3d 439, 457 (2d Cir.2004). Determination of strength therefore begins with inquiry as to whether the mark has the inherent distinctiveness that would entitle it to protection in the absence of secondary meaning. Marks are classified, in ascending order of strength, as “(1) generic; (2) descriptive; (3) suggestive; [or] (4) arbi *385 trary or fanciful.” TCPIP Holding Co. v. Haar Communications, Inc., 244 F.3d 88, 93 (2d Cir.2001). Generic marks are those consisting of words identifying the relevant category of goods or services. They are not at all distinctive and thus are not protectable under any circumstances. Id. Descriptive marks are those consisting of words identifying qualities of the product. They are not inherently distinctive, but are protectable provided they have acquired secondary meaning, which we sometimes refer to as “acquired distinctiveness.” Id. at 94. Suggestive marks and arbitrary or fanciful marks are each inherently distinctive. Id. Suggestive marks are those that are not directly descriptive, but do suggest a quality or qualities of the product, id., through the use of “imagination, thought and perception,” Time, Inc. v. Petersen Publ’g Co., 173 F.3d 113, 118 (2d Cir.1999) (quoting Stix Prods., Inc. v. United Merchants and Mfrs., Inc., 295 F.Supp. 479, 488 (S.D.N.Y.1968)). Arbitrary or fanciful marks are ones that do not communicate any information about the product either directly or by suggestion. See generally Abercrombie & Fitch Co. v. Hunting World, Inc., 537 F.2d 4, 9-11 (2d Cir.1976) (discussing scale of inherent distinctiveness).

Among marks relying solely on inherent distinctiveness for their protecta-bility, arbitrary or fanciful ones are the strongest. See Virgin Enters., 335 F.3d at 147 (explaining that only marks that can be considered clearly strong based exclusively on inherent distinctiveness are those that are arbitrary or fanciful). In the absence of any showing of secondary meaning, suggestive marks are at best moderately strong. See Patsy’s Brand, Inc. v. I.O.B. Realty, Inc., 317 F.3d 209, 217 (2d Cir.2003) (explaining that suggestive words, when rendered in sufficiently stylized script, may be strong even in absence of showing of secondary meaning); Lang v. Retirement Living Publ’g Co., 949 F.2d 576, 581 (2d Cir.1991) (explaining that “suggestiveness is not necessarily disposi-tive of the issue of the strength of the mark,” and that presence or absence of secondary meaning is relevant factor in determining suggestive mark’s origin-indicating quality and therefore its strength); accord Sports Auth., Inc. v. Prime Hospitality Corp., 89 F.3d 955, 961 (2d Cir.1996).

Once a mark has been classified, the second step in determining strength is to consider its degree of distinctiveness, an inquiry that concerns both the “inherent inventiveness of the mark itself and the amount of third-party usage of the term as a mark, especially in the market in question.” 2 J. Thomas McCarthy, McCarthy on Trademarks and Unfair Competition § 11:81, at 11:159 (2005); see Playtex Prods., Inc. v. Georgia-Pac. Corp., 390 F.3d 158, 164 (2d Cir.2004) (explaining that term “Wet Ones” was suggestive, not descriptive, because “the term Wet Ones,’ without more, does not itself conjure up the image of a towelette”). Star’s mark is suggestive, but just barely. The “O” requires some thought to identify it as the first letter of “orange,” but, with an orange slice depicted on the label, the degree of thought is minimal. Moreover, first letters to suggest descriptive words are near the bottom of the range of suggestive marks, if includable at all. Compare Anheuser-Busch, Inc. v. Stroh Brewery Co., 750 F.2d 631 (8th Cir.1984) (holding “L.A.” suggestive of low alcohol beer), with G. Heileman Brewing Co. v. Anheuser-Busch, Inc., 873 F.2d 985 (7th Cir.1989) (holding “L.A.” descriptive of low alcohol beer). Furthermore, the stylizing characteristics of the mark that achieve its placement in the suggestive category are minimal. Finally, Star has made no showing of secondary meaning. Thus, although the inherent dis *386 tinctiveness of Star’s mark invests it with some strength, see, e.g., Patsy’s, 317 F.3d at 217, all of the relevant factors combine to indicate that the mark is weak.

2. Similarity

The district court found that the marks lacked similarity. This holding was not clearly erroneous. The court reasoned that while the two “0” marks appear very similar when viewed in isolation, this similarity is tempered by the fact that the respective packaging is very different: the significance of the similarity of the “0” designs is undercut by the dissimilarity of the products’ respective labels as a whole. “In assessing similarity, courts look to the overall impression created by the • logos and the context in which they are found and consider the totality of factors that could cause confusion among prospective purchasers.” Gruner + Jahr, 991 F.2d at 1078; accord, Estee Lauder, Inc. v. Gap, Inc., 108 F.3d 1503, 1512 (2d Cir.1997); see also King Research, Inc. v. Shulton, Inc., 454 F.2d 66, 68 (2d Cir.1972) (finding similarity factor weighed against finding likelihood of confusion despite similarity of marks themselves because packaging was very dissimilar). Bacardi’s label displays the Bacardi “O” design alone against a clear background, whereas Georgi’s label displays the Star “0” alongside a number of other elements and against a white background. Furthermore, each label prominently displays the brand logo-the stylized “Bacardi” logo and bat symbol on the Bacardi 0 label, and the stylized “Georgi” logo on the Georgi 0 label. In light of these differences, it cannot be said that the similarity factor clearly favors Star, and the district court’s finding that it favored Bacardi was not clear error.

3. Competitive Proximity

Georgi 0 and Bacardi 0 are in moderate competitive proximity with one another. While the district court did not clearly state a finding that this factor favored Star, it did state that “Georgi 0 vodka and Bacardi 0 rum and malt beverage may be in competitive proximity.” Star Indus., Inc. v. Bacardi & Co. Ltd., 2003 WL 23109750, at *3 (S.D.N.Y. Dec.31, 2003). While this seems to imply a finding that the proximity factor favored Star, ap-pellees see it differently, see Appellees’ Br. at 46, and rely on the facts, found by the district court, that rum and vodka are distinct products typically displayed in separate areas of liquor stores, and that malt beverages and vodka are never sold in the same stores in New York. While the district court did find these facts, they do not require a finding that the respective products were not in competitive proximity, and the district court did not expressly state that they required such a legal conclusion. The district court did seem to imply that the fact that malt beverages and hard liquors are not sold in the same stores in New York requires that the proximity factor be held to favor Anheuser-Busch as a matter of law. However, this would be erroneous under Patsy’s, which held products to be in competitive proximity despite the fact that they were never sold in the same stores, where “[t]he products appealed] to the same consumers, and sale locations [we]re geographically close,” and consumers who visited stores selling the one product “[we]re reasonably likely to visit nearby retail stores where [the other was] sold, creating the opportunity for confusion.” 317 F.3d at 218. Similarly, were we to read the district court as making a factual determination that Geor-gi vodka and Bacardi rum were not in competitive proximity, we would find that holding to be clearly erroneous, for those two products are sold in the same precise locations, frequented by the same precise consumers.

*387 Appellees contend that Georgi is a “third tier” brand of cheap vodka that cannot be considered to be in competitive proximity with Bacardi’s “premium” rum products. The district court made no finding of fact one way or the other on this contention, and the record, while it does make clear that Georgi products are much less expensive than Bacardi’s, does not seem to support appellees’ contentions regarding the relative quality of the products. Appellees rely also on differences in the way the respective products are marketed, contending that Georgi vodkas are commonly used as well drinks and stocked under the bar or on the bottom shelf of liquor stores, while Bacardi products are allegedly displayed behind the bar and on eye level shelves at liquor stores. The district court did not accept these contentions in its findings of fact, and the record does not seem to clearly support them. In any event, even accepting these contentions as facts, we would still find the proximity factor to favor Star. Georgi and Bacardi products are marketed to the same consumers in the same bars and stores. That Georgi vodka costs half as much as Bacardi rum and is displayed on different shelves in the same store does not imply that the products are not in competition. The cases relied on by appellees are not binding precedents for this Court and are in any event inapposite. See National Distillers Prods. Co. v. Refreshment Brands, Inc., 198 F.Supp.2d 474, 481-82 (S.D.N.Y.2002) (holding $3/bottle low-alcohol vodka cooler sold in supermarkets not in competitive proximity to $25/bottle “premium” vodka sold in bars and liquor stores); Jim Beam Brands Co. v. Beamish & Crawford, Ltd., 852 F.Supp. 196, 200 (S.D.N.Y.1994) (holding beer and whiskey not in competitive proximity because whiskey not sold on tap at bars).

Appellees are correct to point out that vodka, rum, and malt beverages do reside in distinct submarkets of the market for alcoholic beverages. Vodka and rum, while sold in the same stores to the same consumers for similar purposes, are distinct varieties of product; and this is more so with malt beverages, which are not sold in the same New York stores as vodka and which contain a much smaller concentration of alcohol. Our finding of competitive proximity is thus tempered, and the factor does not overwhelmingly favor Star.

4. Bridging the Gap

“Bridging the gap” refers to the likelihood that the senior user will enter the junior user’s market in the future, or that consumers will perceive the senior user as likely to do so. See Sports Awth., 89 F.3d at 963. The district court determined that the evidence that appellees were seeking to “bridge the gap” between their product markets and Georgi’s was weak and therefore held that this factor favored appellees. Star contends that the court should have concluded that this factor favored Star, based on evidence that Bacardi was seeking to capture a share of the flavored vodka market.

Because, as we have held above, Star’s and appellees’ products are already in competitive proximity, there is really no gap to bridge, and this factor is irrelevant to the Polaroid analysis in this case. See Patsy’s, 317 F.3d at 218 (citing Paddington Corp. v. Attiki Imps. & Dists., Inc., 996 F.2d 577, 586 (2d Cir.1993)).

5. Actual Consumer Confusion

The district court “actual consumer confusion” factor militated against finding a likelihood of confusion. In light of the record, this holding was not clearly erroneous. The court explained that Star’s evidence of actual confusion was extremely weak while appellees’ evi *388 dence, though not flawless, tended to prove that actual confusion was minimal at worst and nonexistent at best. Star’s evidence of actual confusion consisted entirely of testimony by several interested witnesses recounting a handful of anecdotes, including a number of hearsay statements by unidentified and unidentifiable declar-ants. Appellees, on the other hand, submitted consumer surveys tending to rebut charges of actual consumer confusion. Star points out, and the district court noted, various flaws with these surveys. But as appellees correctly point out, Star’s failure to present its own consumer survey weighs against a finding of consumer confusion. See Merriam-Webster, Inc. v. Random House, Inc., 35 F.3d 65, 72 (2d Cir.1994). Weighing appellees’ consumer survey evidence, however flawed, against the kind of evidence relied on by Star, we cannot say that the district court committed clear error in finding that this factor favored appellees.

6. Bad Faith

Star failed to meet its burden to prove that Bacardi adopted its “0” design in bad faith. Bad faith generally refers to an attempt by a junior user of a mark to exploit the good will and reputation of a senior user by adopting the mark with the intent to sow confusion between the two companies’ products. See, e.g., Lang, 949 F.2d at 583. “Selection of a mark that reflects the product’s characteristics, request for a trademark search and reliance on the advice of counsel are factors that support a finding of good faith.” Id. Star concedes Anheuser-Busch’s good faith, but contends that Bacardi adopted its “0” design in bad faith. Star does not contend that Bacardi intentionally copied Star’s design, but rather that it was “willfully blind” in that its decision makers knew or should have known of Star’s “0” but failed reasonably to investigate.

*389 Bad faith may be inferred from the junior user’s actual or constructive knowledge of the senior user’s mark. See Paddington, 996 F.2d at 587. The district court made no factual finding on whether any Bacardi employee or agent was aware of the Star “0” design prior to the introduction of the Bacardi “0” design. Star presented evidence that northeast regional vice presidents of Bacardi had been exposed to the design. But in light of (1) Bacardi’s evidence that its “0” design was developed by a third party design firm retained by its marketing executive in charge of the Bacardi 0 project; (2) the lack of evidence that either the design firm or any Bacardi officials connected to the Bacardi 0 project had any awareness of the Star “0” design; (3) the lack of direct evidence tending to prove an intent to confuse or to exploit Star’s reputation or good will; and (4) the implausibility of the notion that a premier international rum manufacturer would seek to conflate its products with those of a regional discount vodka manufacturer-as appellees argued, “[g]iven [Bacardi’s] name recognition and good will, and [Star’s] relative obscurity, any confusion would [likely] have redounded to the plaintiffs, rather than the defendant’s, benefit,” W.W.W. Pharm. Co., 984 F.2d at 575-Star has not met its burden of proving bad faith.

7. Quality

The quality factor does not support either party. The district court made no explicit finding as to the respective quality of the parties’ products, aside from noting that Georgi’s vodka was much cheaper than Bacardi’s rum. The record is insufficient to support a finding that either product is markedly superior in quality to the other. Each party claims that its product is “premium.” Appellees disparage Georgi as a “third tier” vodka, but the record evidence does not support their disparagement of its quality; rather, it merely demonstrates its inexpensiveness. On the other hand, Star has not even alleged that Bacardi 0 is of inferior quality to Georgi 0. This makes some sense, as a very marked difference in quality between the two products would militate against finding a likelihood of confusion as consumers “will be less likely to assume that the senior user whose product is high-quality will have produced the lesser-quality products of the junior user,” Savin Corp., 391 F.3d at 461. But if there is no reduction in quality in the Bacardi product, then Star is less likely to have suffered any harm, as the quality factor of Polaroid “is primarily concerned with whether the senior user’s reputation could be jeopardized by virtue of the fact that the junior user’s product is of inferior quality.” Sports Auth., 89 F.3d at 965. In any event, absent factual findings or evidence one way or the other, this factor is at most evenly balanced.

8. Consumer Sophistication

The “consumer sophistication” factor militates against finding a likelihood

*388 As noted by the district court, Bacardi conducted a trademark search prior to adopting its “0” mark. As noted above, the fact that Bacardi conducted a trademark search would normally be held to militate for a finding that Bacardi adopted its “0” design in good faith. Star contends, however, that the trademark search was obviously flawed and that Bacardi’s failure to remedy it by ordering a new trademark search is evidence of “willful blindness.” Star contends, accordingly, that Bacardi’s reliance on the advice of counsel should not be held, as it normally would be, to support a finding of good faith, because counsel relied on the clearly flawed search. Bacardi responds that any flaw in its trademark search is somewhat beside the point, as Star made no attempt to register its “0” design as a trademark until after the commencement of the instant litigation-no trademark search, no matter how perfect, would have discovered Star’s “0” design trademark.

This Court has never held adoption of a mark with no knowledge of a prior similar mark to be in bad faith even in the total absence of a trademark search, much less on the basis of an allegedly flawed trademark search. See Savin Corp., 391 F.3d at 460. Furthermore, in some cases even where a trademark search resulted in knowledge of the earlier mark, in the absence of additional evidence indicating an intent to promote confusion or exploit good will or reputation, this Court has found the junior user to be in good faith. See W.W.W. Pharm. Co. v. Gillette Co., 984 F.2d 567, 575 (2d Cir.1993), limited on other grounds by Deere & Co. v. MTD Prods., Inc., 41 F.3d 39, 46 (2d Cir.1994). 3 *390 of confusion. The district court appears to have relied heavily on' this factor in its overall Polaroid analysis, noting that alcoholic beverage consumers are certainly sophisticated enough to distinguish rum from vodka. While humorous, and ultimately correct in its conclusion, the court’s analysis of this point was flawed. As we noted above, Star asserts assoeiational confusion, not direct confusion, and so it is irrelevant whether consumers are capable of distinguishing rum from vodka. At issue is whether consumers in this market, confronted with products bearing similar stylized letter “0” designs on their labels, are generally sophisticated enough to understand that Bacardi’s use of a large elliptical orange “0” on its orange-flavored rum bottles is not indicative of a licensing or sponsorship agreement with Star.

Our analysis of consumer sophistication “considers]' the general impression of the ordinary purchaser, buying under the normally prevalent conditions of the market and giving the attention such purchasers usually give in buying that class of goods.” Sports Auth., 89 F.3d at 965 (internal brackets omitted). Consumer sophistication may be proved by direct evidence such as expert opinions or surveys. In addition, in some cases a court is entitled to reach a conclusion about consumer sophistication based solely on the nature of the product or its price. See Patsy’s, 317 F.3d at 219 (noting that consumer sophistication is generally low in dealing with cheaper products or products sold in the rough-and-tumble world of the supermarket). In the instant case, the parties have not submitted any relevant expert opinions or surveys, and we are left to rely solely on such indirect indications of sophistication.

The record reveals that Georgi orange-flavored vodka retails for about $12 a liter, and Bacardi 0 costs about twice as much. Liquor stores, which sell a limited variety of products, are not as cluttered and frantic as the supermarkets described in Patsy’s. In comparison with the items available at such supermarkets, $24 bottles of liquor are relatively expensive. We conclude that the balance tips in favor of Bacardi with respect to consumer sophistication. Unhurried consumers in the relaxed environment of the liquor store, making decisions about $12 to $24 purchases, may be expected to exhibit sufficient sophistication to distinguish between Star’s and Bacardi’s products, which are differently labeled. See McGregor-Doniger Inc. v. Drizzle Inc., 599 F.2d 1126, 1137-38 (2d Cir.1979) (explaining that greater sophistication is required to distinguish identical products than to distinguish products that are differently labeled); cf. Playtex Prods., 390 F.3d at 162 (noting that “consumers cannot be expected to invest much time or effort in distinguishing among inexpensive bath tissues”). Consumers may be expected to educate themselves sufficiently to recognize the respective brand names, to understand the respective stature of the two companies, and thus to understand how fanciful the notion is that Bacardi would seek the benefit of Star’s reputation and good will by entering into a licensing agreement to secure utilization of its “0” design. Rather, as the district court assumed, such consumers would generally be expected to understand that Bacardi’s use of the letter “0” was intended to indicate orange flavor, and not a marketing or sponsorship arrangement with Star. The district court’s finding to this effect, while relying on some flawed analysis, was ultimately correct.

B. Balancing

The district court concluded correctly that, balancing all of the Polaroid factors, confusion was not likely. To sum *391 marize our findings above: five factors (strength of mark, similarity, actual confusion, bad faith, and consumer sophistication) favor appellees, one favors Star (competitive proximity), and two are neutral (bridging the gap and quality). Furthermore, Star’s showing of proximity is not overwhelming. Because the various products reside in distinct identifiable sub-markets, we found only moderate competitive proximity. On the other hand, as we have discussed, Star’s evidence of actual consumer confusion and bad faith was extremely weak, and these factors therefore clearly tipped in appellees’ favor. Thus, Star made a strong showing on none of the Polaroid factors, and a weak showing on one, while five of the six that were other than neutral came out in favor of appel-lees, two of them strongly. Reviewing the district court’s balancing de novo, we find that the district court’s balancing was correct and that the overall balance does favor appellees. Star has therefore not demonstrated a likelihood of consumer confusion, and has not met its burden to prove trademark infringement.

CONCLUSION

For the foregoing reasons, we affirm the district court’s judgment for defendants-appellees.

. Unlike its 1996 PTO application which sought to register the word "Georgi O” as a trademark, Star now sought to register its stylized "O” design separately from the word "Georgi.”

. “As we have noted on many occasions!,] the decisions of the [Trademark Trial and Appeals Board (TTAB) ], while not binding on courts within this Circuit, are nonetheless to be accorded great weight." Hall, 334 F.3d at 216 (internal quotation marks, brackets, and ellipsis omitted).

. Star relies primarily on cases from other circuits dealing with so-called "reverse confu *389 sion.” A "reverse confusion” situation exists where the junior user is able to amass such trademark strength in its imitative mark that the senior user's products become associated with the junior user in the minds of consumers. In this situation, some courts of appeals have held that a junior user’s failure to conduct an adequate trademark search may by itself be sufficient to prove that an imitative mark was adopted in bad faith. See, e.g., Fisons Horticulture, Inc. v. Vigoro Indus., Inc., 30 F.3d 466, 480 (3d Cir.1994). However, this particular holding is at odds with this circuit's precedents, which preclude finding bad faith on the basis of an inadequate trademark search, at least in the absence of evidence that the inadequate design or the failure to correct inadequacies in the search was motivated by an intent to sow consumer confusion or to exploit the good will or reputation of the senior user. See Sunenblick v. Harrell, 895 F.Supp. 616, 633 (S.D.N.Y.1995).