The word "capsule" can describe many things-a pharmaceutical pill, a space pod, or a short synopsis, for example. What about a cellphone case?

The U.S. Patent and Trademark Office has given conflicting answers. It has, on a few occasions, found that "capsule" was "merely descriptive" of cellphone cases, a finding that precludes the mark from registration on the Principal Register. Yet the Office has also found otherwise and allowed one manufacturer-Uncommon, LLC-to register "capsule."

Rival case manufacturers still use the term, though. Uncommon sued one of those manufacturers, Spigen, Inc., for trademark infringement and unfair competition. In discovery, Spigen produced a survey to prove that consumers did not associate "capsule" with Uncommon's cases, and it disclosed the person who conducted the survey as a "non-testifying expert." That was a problem: without foundational expert testimony to explain the survey's methodology, it was inadmissible. But when the litigation reached summary judgment, the district court excused Spigen's error. It then granted Spigen summary judgment on the merits.

We affirm. Spigen's disclosure was inaccurate, but calling it harmless, as the district court did, was reasonable. With the survey, there was no genuine issue of material fact as to the mark's invalid registration. Nor was there an issue of fact regarding the unlikelihood of consumer confusion. Summary judgment was therefore appropriate.

I. Background

Uncommon and Spigen are manufacturers and retail sellers of cellphone cases. Starting in December 2009, Uncommon began marketing a line of cases it called "CAPSULE." Its first sale came in July 2010, and Uncommon applied to register the mark with the Patent Office in September 2012. In May 2013, the Patent Office issued Trademark Registration No. 4,338,254 . The registration protected "capsule" for "cases specifically adapted for protection and storage of consumer electronics, namely, cellular phones and mobile media players."

That registration was not the first or last time the Patent Office would address the mark "capsule" for cellphone cases. In July 2008, the Office had issued a "capsule" registration to Vatra, Inc., which had claimed to first use the term in September 2007. (The Office later canceled Vatra's trademark in February 2015, after Vatra failed to certify its continued use.) Capsule Case Corp. likewise attempted to register "CAPSULE CASE" in 2017, though the record does not say what came of that application.

*415 Spigen joined the "capsule" ranks relatively early, in 2010. It sold its first "capsule"-branded case in June of that year, when, all agree, Spigen was unaware of Uncommon's trademark. Spigen then grew the brand into a line: the "Capsule-family" cases. The family includes the "Capsule Capella," the "Rugged Capsule," the "Capsule Ultra Rugged," the "Capsule Solid," and the "Air Capsule" cases.

Spigen applied to register the family of marks between 2014 and 2016. It was able to register three of those marks, "Capsule Capella," "Capsule Solid," and "Air Capsule." But for both "Capsule Solid" and "Capsule Capella" the Office decided that the marks were descriptive, at least in part. "Capsule Solid" was therefore put on the Supplemental (as opposed to Principal) Register, and the Office required the "Capsule" in "Capsule Capella" to be disclaimed by Spigen as descriptive and nonexclusive. The Office initially declined to register the other two marks, "Rugged Capsule" and "Capsuled Ultra Rugged," concluding that both were descriptive and likely to be confused with Uncommon's mark. These two registration applications have been suspended pending this litigation.

Uncommon's and Spigen's "capsule" cases have more in common than the name. Both sell their cases online, through their own websites and online retailers, like Amazon and eBay; both have a national market; both use the "capsule" name alone and in conjunction with other descriptors; and both use dark-colored, sans-serif font for the mark on their packaging. But there are differences too. Uncommon's cases can be customized with personal images, or they come "ready-made" with colorful designs or artwork. Spigen's cases cannot be customized and are generally solid in color. Uncommon, moreover, does not generally make cases for iPhones after the 5/5s generation (as far as this record is concerned); Spigen does-and, unlike Uncommon, it makes cases for Samsung cellphones as well. The two companies' packaging is also different, as shown by the record examples below.

*416 As illustrated, Uncommon's cases generally come in some combination of a blue, grey, white, and/or black package. Spigen's come in a distinctive orange and black package.

Seeing enough similarities, however, Uncommon filed this suit in December 2015, alleging that Spigen's use of "capsule" constituted trademark infringement and unfair competition under federal and state law. See 15 U.S.C. §§ 1114 , 1125(a). 1 Spigen countersued. It sought, among other things, to cancel Uncommon's registration.

The case entered discovery. Spigen disclosed its experts to Uncommon in December 2016. There were two: Doug Bania, who would testify to the lack of confusion, the descriptiveness of Uncommon's mark, secondary meaning, and damages; and Kirk Martensen, who, according to Spigen's disclosure, was a "non-testifying expert who will conduct a consumer survey" that may be relied upon by Bania. The disclosure added that Martensen may "be called to testify on the methodology of the survey if needed."

Spigen later sent Uncommon a two-page summary of Martensen's credentials. Bania's report, produced shortly after, relied extensively on Martensen's survey, which was attached to the report with its underlying data. In his deposition, Bania admitted that he was not an expert on survey methodology. Uncommon never asked to depose Martensen.

The parties cross-moved for summary judgment in June 2017. Spigen relied on the Martensen survey in its opening. Uncommon responded that the survey could not be considered at summary judgment, primarily because no expert could testify to its methodology. Spigen tried to fill that evidentiary gap in its reply. It submitted a declaration from Martensen explaining the survey's methodology-information which, by and large, had been disclosed in the earlier production of the survey itself. Uncommon moved to strike the declaration as untimely and inadequate. It did not seek leave to file a surreply responding on the merits.

The district court ruled in March 2018. It held that Spigen's failure to disclose Martensen as a testifying expert was harmless under Federal Rule of Civil Procedure 37(c)(1). The court further found that the survey was sufficiently reliable and supported by the declaration. Turning to the merits, the district court disposed of the case on two alternative grounds. First, the court found no issue of material fact as to whether "capsule" was descriptive, per the word's dictionary definitions, and it found no issue as to whether the mark lacked a secondary meaning. The court therefore canceled the registration, which necessarily defeated Uncommon's infringement and unfair-competition claims. Second, the court decided that there was no issue of fact regarding the lack of confusion posed by Spigen's use of "capsule," which also defeated both of Uncommon's claims.

The district court accordingly entered summary judgment in Spigen's favor. Uncommon appeals.

II. Discussion

Uncommon submits that the district court erred in three ways: by considering the Martensen declaration; by canceling the mark's registration; and by deciding, alternatively, that there was no likelihood of confusion between Uncommon's cases and Spigen's. Because the Martensen declaration *417 colors the summary-judgment analysis, we start there.

A. The Martensen Survey and Declaration

At summary judgment, Spigen relied on a piece of evidence familiar in trademark litigation-a consumer survey. See generally Robert C. Bird and Joel H. Steckel , The Role of Consumer Surveys in Trademark Infringement: Empirical Evidence from the Federal Courts , 14 U. Pa. J. Bus. L. 1013 (2012). Like all summary-judgment evidence, the survey, which Martensen conducted, needed to be admissible (or capable of taking admissible form) for the court to consider it. See, e.g. , Cehovic-Dixneuf v. Wong , 895 F.3d 927 , 931 (7th Cir. 2018).

To be admissible, a consumer survey almost always requires expert testimony. This is because conducting a survey entails expertise and the exercise of professional judgment. See Muha v. Encore Receivable Mgmt., Inc. , 558 F.3d 623 , 625-26 (7th Cir. 2009) ; Spraying Sys. Co. v. Delavan, Inc. , 975 F.2d 387 , 394 (7th Cir. 1992). No survey is "foolproof," and examiners make a range of decisions about sample size, question design, survey format, and statistical analysis-all of which can impact reliability. See Kraft Foods Grp. Brands LLC v. Cracker Barrel Old Country Store, Inc. , 735 F.3d 735 , 741-42 (7th Cir. 2013) ; see also 6 McCarthy on Trademarks and Unfair Competition § 32:158 (5th ed. 2019) ("McCarthy on Trademarks"). Expert testimony is therefore necessary to explain the survey's methodology and to attest to its compliance with principles of the profession. See Fed. R. Evid. 702 ; Evory v. RJM Acquisitions Funding L.L.C. , 505 F.3d 769 , 776 (7th Cir. 2007).

Because Spigen's survey evidence required expert testimony, Spigen had to comply with Federal Rule of Civil Procedure of 26(a). It did not. Rule 26(a)(2) obligates parties to timely disclose any expert who may testify and to provide a related expert report. Until the back half of its summary-judgment briefing, Spigen treated Martensen like a non -testifying expert. It designated him as such and failed to supply a formal, written report on his behalf.

The district court, however, excused Spigen's disclosure failures under Rule 37. Rule 37(c)(1) states that if a party fails to comply with Rule 26(a), the evidence is excluded "unless the failure was substantially justified or is harmless"-and the district court thought Spigen's failures were harmless. See Karum Holdings LLC v. Lowe's Cos., Inc. , 895 F.3d 944 , 951 (7th Cir. 2018). We review that decision for an abuse of discretion and affirm as long as it was reasonable. King v. Ford Motor Co. , 872 F.3d 833 , 838 (7th Cir. 2017). The following factors, we have said, should guide district courts in making Rule 37 determinations: "(1) the prejudice or surprise to the party against whom the evidence is offered; (2) the ability of the party to cure the prejudice; (3) the likelihood of disruption to the trial; and (4) the bad faith or willfulness involved in not disclosing the evidence at an earlier date." Tribble v. Evangelides , 670 F.3d 753 , 760 (7th Cir. 2012) (quoting David v. Caterpillar, Inc. , 324 F.3d 851 , 857 (7th Cir. 2003) ).

Spigen was remiss in framing Martensen's role, but the district court's excusal was not an abuse of discretion. There was ample reason to conclude, as the district court did, that Uncommon did not suffer any surprise. Spigen timely disclosed Martensen's survey with its underlying data; a two-page document summarizing Martensen's credentials; and Bania's report, which relied heavily on the survey.

*418 And the later-submitted Martensen declaration added little to those substantial disclosures. The survey itself explained Martensen's methodology and reliance on accepted research principles.

Spigen therefore supplied (in some form) nearly all the information Rule 26(a)(2) requires, with two notable exceptions. The first, of course, is the designation of Martensen as a testifying expert, although Uncommon knew that Martensen's testimony was on the table. The disclosure stated that he may "be called to testify on the methodology of the survey if needed." The second piece of information, which was included in the declaration but not the survey, was the fact that the survey employed a " Squirt format," as opposed to an " Eveready " one. See generally McCarthy on Trademarks § 32:173.50 (explaining the difference between the two types of surveys). Uncommon, however, does not explain why this distinction matters here or how the revelation caught it off guard.

Spigen's failures also caused no disruption to the proceedings. Perhaps there would have been a minor one had Uncommon sought to extend discovery or depose Martensen-but it never made either request. (More on this below.) And there was no bad faith on Spigen's part. As the district court explained, Spigen's failures resulted from a misunderstanding about how to admit certain expert evidence. It was not a "tactic of surprise." Sherrod v. Lingle , 223 F.3d 605 , 613 (7th Cir. 2000) ; see also Brown v. Columbia Sussex Corp. , 664 F.3d 182 , 191 (7th Cir. 2011) (concepts of "willfulness" and "bad faith" usually require "intentional or reckless behavior").

The remaining considerations, prejudice and the ability to cure it, present closer calls. Uncommon believes that the misleading disclosure harmed it in a few ways: Uncommon could not question Martensen about his methodology; it missed its chance to find a rebuttal expert; and it could not respond to the substance of the survey, having believed that it was inadmissible. That view is slightly overstated. The "non-testifying expert" designation did not prohibit Uncommon from deposing Martensen. Any expert who may testify at trial can be deposed, Fed. R. Civ. P. 26(b)(4)(A), and the fact that Bania relied so heavily on Martensen would likely have made Martensen fair game for discovery. Uncommon also had reason enough to order a survey of its own, knowing, as it must have, that Spigen's survey was a centerpiece of its case. And the fact that Uncommon believed that the survey was inadmissible in its then-form did not preclude Uncommon from responding to its substance during summary judgment. Litigants are permitted alternative arguments, after all.

But set all that aside and assume real prejudice. The question in this case becomes: prejudice inflicted by whom? No doubt Spigen was to blame in the first instance, but Uncommon seems to have adopted a strategic aversion to self-help.

Again, Uncommon knew the survey was key to Spigen's case. It also knew after Bania's deposition that Spigen would have an evidentiary gap in both admitting the survey and defending Bania's reliance on it. It did not flag the error for Spigen, which was fair; litigation is adversarial. But it also made no effort to conduct further discovery into the survey's methodology or Martensen. Instead, Uncommon awaited summary judgment to argue that the survey and Bania's reliance on it were inadmissible. And when Spigen, in reply, offered the Martensen declaration, Uncommon moved only to strike. It forewent the obvious, alternative option: request additional expert discovery or, at a minimum, a deposition of Martensen. See *419 David , 324 F.3d at 857 ( Rule 26 error was harmless, in part, because the other party "did not seek a continuance so that it could obtain additional information" relating to the evidence). Either measure would have allowed Uncommon to further scrutinize, and then respond to, the survey and Martensen if the district court disagreed with Uncommon's main argument. Yet Uncommon took a strategic risk and stuck to its guns-the survey could not be admitted, Uncommon argued, and the district court needed to reject the Martensen declaration under Rule 37(c)(1).

Rule 37, however, provides recourse for parties actually harmed by a litigant's noncompliance with disclosure obligations. It does not safeguard a party's decision to sense an error, seize on it, and then, when it is resolved, claim incurable harm in the face of apparent remedies. Litigation is adversarial, not a game of gotcha.

We should be clear: Spigen could have been faulted for its Rule 26 failures. Parties who fall short on their disclosure obligations generally lose out on their expert evidence, as Rule 37(c) and plenty of caselaw make plain. See, e.g. , Karum , 895 F.3d at 951-53 ; Tribble , 670 F.3d at 759-60 ; Ciomber v. Coop. Plus, Inc. , 527 F.3d 635 , 643 (7th Cir. 2008). But here, with this record's unusual facts, the district court acted within its discretion. There was no surprise to Uncommon and no delay in the proceedings, and while there was likely some prejudice, it was neither unforeseeable nor incurable-even if the cure was not Uncommon's strategic preference. 2

B. Summary Judgment Merits

With the Martensen survey in play, we can turn to the merits. Uncommon's claims for trademark infringement and unfair competition both require (1) that Uncommon's mark be validly registered and (2) that Spigen's use be likely to cause confusion among consumers. H-D Mich., Inc. v. Top Quality Serv., Inc. , 496 F.3d 755 , 759 (7th Cir. 2007) ; Packman v. Chicago Tribune Co. , 267 F.3d 628 , 638 & n.8 (7th Cir. 2001). Spigen's trademark-cancellation counterclaim requires that Uncommon's registration be invalid. 3 See TE-TA-MA Truth Found.-Family of URI, Inc. v. World Church of Creator , 297 F.3d 662 , 665 (7th Cir. 2002). The district court gave two grounds for granting Spigen summary judgment: the registration was invalid and there was no likelihood of confusion.

We review those decisions de novo. We construe the facts and draw all reasonable inferences in Uncommon's favor. Arlington Specialties, Inc. v. Urban Aid, Inc. , 847 F.3d 415 , 418 (7th Cir. 2017).

1. The Trademark's Validity

For a mark registration to be valid, it must be distinctive in one of two ways. See Wal-Mart Stores, Inc. v. Samara Bros. , 529 U.S. 205 , 210, 120 S.Ct. 1339 , 146 L.Ed.2d 182 (2000) ; see also 15 U.S.C. §§ 1052 (e), (f).

First, the mark can be " inherently distinctive," meaning, it identifies the maker of a product, but it does not describe the product itself.

*420 Two Pesos, Inc. v. Taco Cabana, Inc. , 505 U.S. 763 , 768-69, 112 S.Ct. 2753 , 120 L.Ed.2d 615 (1992) (emphasis added). To decide which is which, we look to Judge Friendly's continuum: marks can be "fanciful, arbitrary, suggestive, descriptive, or generic." H-D Mich. , 496 F.3d at 759 ; see Abercrombie & Fitch Co. v. Hunting World, Inc. , 537 F.2d 4 , 9 (2d Cir. 1976) (Friendly, J.). Fanciful ("Exxon" gasoline), arbitrary ("Apple" computers), and suggestive terms ("Tide" detergent) are inherently distinctive. Generic ("Apple" apples) and descriptive ("Tasty" apples) terms are not, and therefore they cannot generally be registered. See 15 U.S.C. § 1052 (e). This rule ensures that competitors can market their products by describing them. Custom Vehicles, Inc. v. Forest River, Inc. , 476 F.3d 481 , 483 (7th Cir. 2007). It also reflects that descriptors are a poor means to distinguish among competing products. Platinum Home Mortg. Corp. v. Platinum Fin. Grp., Inc. , 149 F.3d 722 , 727 (7th Cir. 1998).

Second, and alternatively, descriptive terms can become distinctive by acquiring a "secondary meaning," that is, the mark has grown distinctly and "uniquely associated with the original seller." Custom Vehicles , 476 F.3d at 483 . Think, for example, of "5-Hour Energy," which describes the caffeinated shot, but is readily associated by most customers with one maker. See Innovation Ventures, LLC v. N2G Distrib., Inc. , 779 F. Supp. 2d 671 , 675-78 (E.D. Mich. 2011). Descriptive marks with secondary meanings can be registered, see 15 U.S.C. § 1052 (f), a rule which protects against competitor freeloading and market confusion.

We, of course, do not write on a clean slate in deciding if Uncommon's "capsule" trademark is valid. The Patent Office granted it, and that registration is prima facie evidence of validity. 15 U.S.C. §§ 1057 (b), 1115(a). More specifically, the registration affords Uncommon "one of two" presumptions: (a) that its mark is at least suggestive and not "merely descriptive"; or (b) that if it is descriptive, the mark has a secondary meaning. Packman , 267 F.3d at 638 ; see also Liquid Controls Corp. v. Liquid Control Corp. , 802 F.2d 934 , 936-37 (7th Cir. 1986). Uncommon did not offer evidence of secondary meaning to the Office in applying for registration, according to the record. And Uncommon submits on appeal that the Office granted the registration believing that "capsule" was "not descriptive of cases." So we assume as much and apply the first presumption. See Zobmondo Entm't, LLC v. Falls Media, LLC , 602 F.3d 1108 , 1113-14 (9th Cir. 2010) ("Where the PTO issues a registration without requiring proof of secondary meaning, the presumption is that the mark is inherently distinctive."); Money Store v. Harriscorp Fin., Inc. , 689 F.2d 666 , 673 (7th Cir. 1982) (same). 4

a. Descriptive or Suggestive?

The first and main dispute is whether "capsule" describes, or instead suggests, cellphone cases. Inherent distinctiveness is a question of fact, and summary judgment is appropriate only if, reading the record in Uncommon's favor, no reasonable factfinder could decide for it. See Sands, Taylor & Wood Co. v. Quaker Oats Co. , 978 F.2d 947 , 952 (7th Cir. 1992).

A mark is descriptive when "it describes the product category to which the brand belongs." Custom Vehicles , 476 F.3d at 483 . The mark need not describe the product completely, but it must depict *421 an important characteristic of the product. Id. ; Quaker Oats , 978 F.2d at 952-53 ; see also Spraying Sys. Co. v. Delavan, Inc. , 975 F.2d 387 , 392 (7th Cir. 1992). A suggestive mark, on the other hand, "does not directly and immediately describe" an "aspect of the goods." McCarthy on Trademarks § 11:62. Instead, a suggestive mark "requires the observer or listener to use imagination and perception to determine the nature of the goods." G. Heileman Brewing Co., Inc. v. Anheuser-Busch, Inc. , 873 F.2d 985 , 996 (7th Cir. 1989).

In contending that "capsule" is suggestive, Uncommon trumpets its presumption of validity. But the presumption is just that-a presumption, open to rebuttal. And Spigen has carried its burden not only to demonstrate that "capsule" is indeed descriptive, but also to show that there is no material issue of fact on the question. 5 See, e.g. , Liquid Controls , 802 F.2d at 936 .

The line between descriptiveness and suggestiveness can be difficult to draw. But a few markers lead the way. See McCarthy on Trademarks § 11:66. First, we can look to how, and how often, the relevant market uses the word in question. See Bliss Salon Day Spa v. Bliss World LLC , 268 F.3d 494 , 497 (7th Cir. 2001) ; Spraying Sys. , 975 F.2d at 393 . Starting with Vatra in 2007, many of Uncommon's competitors-at least ten-use the word in their case names. A competitor is even named Capsule Case Corp., and the appendix contains more than 100 pages of cases that use "capsule" in some fashion. This use shows that the cellphone-case market understands "capsule" to describe cases. See, e.g. , Sec. Ctr., Ltd. v. First Nat'l Sec. Ctrs. , 750 F.2d 1295 , 1300 (5th Cir. 1985) (mark was descriptive because many in the market used it "in some form").

Examples illustrate why. In Bliss Salon Day Spa , we explained that even if the mark "BLISS" did not depict haircare products as a linguistic matter, the fact that many players in the beauty-care market use "BLISS" in association with their products and services meant that the mark was functionally descriptive. 268 F.3d at 497 . In Platinum Home , we reasoned similarly *422 with respect to "platinum" and financial products. 149 F.3d at 730 . Perhaps linguistically, "platinum" only suggests, rather than describes, financial products. "But because so many firms use the word ... for financial products, it cannot be inherently distinctive." Bliss Salon Day Spa , 268 F.3d at 497 (citing Platinum Home Mortg. , 149 F.3d at 730 ). So, too, for "capsule" and cellphone cases. The undisputed and prevalent use of the word in the market means that it is not inherently distinctive.

Uncommon wants us to draw a different inference, but it is neither reasonable nor supported by the record. Uncommon believes that the market's use of "capsule" can be explained by its competitors' freeloading on Uncommon's good reputation with consumers. There is no evidence of that good reputation, let alone evidence of copying, and Uncommon has conceded that Spigen marketed its "capsule" cases without knowing about Uncommon's use. Uncommon needs more than conjecture to survive summary judgement. Carmody v. Bd. of Tr. of Univ. of Ill. , 893 F.3d 397 , 401 (7th Cir. 2018).

Uncommon makes another, related argument. It submits that in the age of Amazon, eBay, and Google, in which consumers shop via search bars, sellers string together words in their product names to ensure the most hits. Uncommon gives a good example: the "iPhone 6 Case, Accez(SoftFlex) Capsule Premium Flexible Soft TPUSlim Case for iPhone 4.7 (Yellow)." As a result of this practice, Uncommon says, we should give little weight to the fact that competitors use the mark. That argument does not track, and we need to look no further than Uncommon's example to see why. Setting aside "Accez" (the seller's name), "capsule" is listed with the terms SoftFlex, premium, flexible, soft, and TPUSlim-all of which are unquestionably descriptive. (TPU means thermoplastic polyurethane, a durable material.) Again, the market's use of the term demonstrates its descriptiveness.

A second, and more common, way courts determine if a mark is descriptive is to employ the so-called degree-of-imagination test. "[I]f a mark imparts information directly it is descriptive. If it stands for an idea which requires some operation of the imagination to connect it with the goods, it is suggestive." Platinum Home Mortg. , 149 F.3d at 727 ; see also, e.g. , G. Heileman Brewing , 873 F.2d at 996 . Employing the test here, "capsule" lands on the descriptive side of the line. The mark directly conveys an idea: coverage, storage, protection-en capsul ation-all of which the cellphone cases provide, according to Uncommon's registration. This conclusion does not depend on how one technically defines "capsule" (which we address in the next paragraph). The word may draw from a pill, pod, and almost any other thing the word "capsule" can describe, and here, as with those items, the point is that the cellphone cases cover and protect a valuable. "Capsule" communicates that point directly.

As just alluded to, courts also look to dictionaries to see if a mark is descriptive. See, e.g. , Door Sys. , 83 F.3d at 171 ; Liquid Controls Corp. , 802 F.2d at 937-38 ; see also McCarthy on Trademarks § 11:69. To be sure, we have been cautious about this approach. Courts "cannot choose between the 'descriptive' and 'suggestive' categories on the basis of a dictionary," we have warned, because dictionaries often cite historical and dated definitions. Bliss Salon Day Spa , 268 F.3d at 497 ; see also World Church of Creator , 297 F.3d at 666 . Here, however, consulting the dictionary quickly reveals what the market's use and imagination test already tell us: "capsule" describes cellphone cases. Spigen put into *423 evidence various dictionary definitions of the word, including: "a small case, envelope, or covering," "any object that can be used to hold things," and a "small container." A cellphone case obviously fits the bill.

Uncommon does not respond by arguing that these definitions are obscure, functionally irrelevant, or out-of-date. It instead faults the district court for relying too heavily them. On that score, we agree-the district court, consistent with our previous warnings, needed to do more than cite the dictionary to find descriptiveness. But there is no prohibition on using dictionaries as a piece of the puzzle. So if, hypothetically, consulting dictionaries revealed that they do not define "capsule" in a manner that so easily fits cellphone cases, Uncommon might have evidence in its favor. A factfinder could potentially infer from such omissions that "capsule" does not describe cellphone cases. Cf. Liquid Controls Corp. , 802 F.2d at 938 . The opposite happens to be true here-dictionaries provide more evidence that "capsule" in fact describes cellphone cases.

Another method is worth addressing. Courts occasionally examine third-party registrations to determine descriptiveness. See McCarthy on Trademarks § 11:20. We decline to do so here. The Office granted Vatra's initial registration for "capsule," as it did Uncommon's. But more often than not, and more recently, it has decided that "capsule" is "merely descriptive," at least according to the record here. Spigen encourages us to heed those latter decisions, but given the internal inconsistency, we find little that can be reasonably gleaned from the Office's registration decisions.

Yet the remainder of the markers-the extent of use, the imagination test, and capsule's definition-show no issue of fact as to whether the mark is descriptive. Against the weight of that evidence, Uncommon offers little.

By way of argument, Uncommon presses that the mark is suggestive because connecting it to a cellphone case requires some imagination. It offers only its own insistence for that proposition, and we have already explained why it is incorrect. Uncommon further argues that because "capsule" could describe an array of products, it cannot be descriptive. We have rejected that thinking before. A term need not "bring to mind the product in question" to be descriptive, nor must it "depict the [product] itself." Quaker Oats , 978 F.2d at 952 . It is enough that the mark in fact "refers to a characteristic of the" product. Id.

By way of evidence, Uncommon provides even less. It points to the Martensen survey, which shows that consumers do not often associate "capsule" with cellphone cases (only 14 percent of the time), and Uncommon submits that this evidence can show that "capsule" is not descriptive. This argument treads the same mistaken path we just corrected. A descriptive mark need not immediately invoke in the consumer's mind the product itself; it must directly convey a characteristic of the product. Id. In Custom Vehicles , for example, we explained why "Work-N-Play" as a name for vehicles with convertible interiors was descriptive: not because consumers quickly associate "Work-N-Play" with interior-convertible vans, but because the vans were in fact "usable both for work and for play." 476 F.3d at 483 . The same is true here. Consumers may not immediately think "cellphone cases" when they hear "capsule," but "capsule" describes a quality of cellphone cases all the same.

Spigen has carried its burden to defeat Uncommon's presumption of inherent distinctiveness and demonstrated that there is no issue of material fact regarding the descriptiveness of the "capsule" mark. No *424 reasonable finder of fact could conclude otherwise based on this record.

b. Secondary Meaning?

Because "capsule" is descriptive, it must have a secondary meaning to be validly registered. A secondary meaning is "a link in the minds of consumers between the marked item and its source." Jay Franco & Sons, Inc. v. Franek , 615 F.3d 855 , 857 (7th Cir. 2010). Put less abstractly, a secondary meaning exists when consumers have come to "uniquely associat[e]" the mark with a single maker. Two Pesos , 505 U.S. at 766 n.4, 112 S.Ct. 2753 ; H-D Mich. , 496 F.3d at 759-60 . Secondary meaning, like descriptiveness, is a question of fact, and thus summary judgment is appropriate only if no reasonable factfinder could decide for Uncommon. See Thomas & Betts Corp. v. Panduit Corp. , 138 F.3d 277 , 292 (7th Cir. 1998).

As an initial matter, the parties and the district court appear to believe that Uncommon's presumption of validity extends to secondary meaning. The Lanham Act extends "one of two" presumptions-either that the mark is inherently distinctive or that it has acquired distinctiveness. Packman , 267 F.3d at 638 . Registration owners do not get both. See, e.g. , Falls Media , 602 F.3d at 1114 n.7 ; Borinquen Biscuit Corp. v. M.V. Trading Corp. , 443 F.3d 112 , 117 n.2 (1st Cir. 2006) ; Arrow Fastener Co. v. Stanley Works , 59 F.3d 384 , 392-93 & n.6 (2d Cir. 1995) ; Aromatique, Inc. , 28 F.3d at 870. So where, as here, the Office registers a mark without requiring evidence of secondary meaning, "it is presumed to be inherently distinctive." McCarthy on Trademarks § 32:134. But when the Office registers a mark "by acceptance of proof of secondary meaning under § 2(f)," then a secondary-meaning presumption applies. Id. The latter did not happen here, and so Uncommon is without the presumption at this stage. See, e.g. , PaperCutter, Inc. v. Fay's Drug Co. , 900 F.2d 558 , 564 (2d Cir. 1990) ("absent evidence that the Office registered the mark upon finding that it had acquired secondary meaning," the presumption does not extend to secondary meaning).

To the merits, there are many ways to prove a secondary meaning or lack thereof: amount of advertising, sales volume, length and manner of use, or consumer testimony, for example. Platinum Home Mortg. , 149 F.3d at 728 . In this case, Spigen marshals one of the more "helpful" means-a consumer survey. Burford v. Accounting Practice Sales, Inc. , 786 F.3d 582 , 590 (7th Cir. 2015). The Martensen Survey shows that very few customers (six percent) were familiar with "capsule" as a brand name; only slightly more (14 percent) could connect the mark to cellphone cases when prompted; and barely any (less than one percent) associated "capsule" with Uncommon. That evidence led the district court to conclude that "capsule" had no secondary meaning.

Uncommon challenges this conclusion on appeal. But it did not press the issue at the district court. Uncommon presented no affirmative evidence of secondary meaning, and it mentioned the issue just once in the course of its summary-judgment briefing-when it asserted, "evidence of secondary meaning is unnecessary because the mark is not descriptive." That amounts to waiver. See, e.g. , Puffer v. Allstate Ins. Co. , 675 F.3d 709 , 718 (7th Cir. 2012) ("It is a well-established rule that arguments not raised to the district court are waived on appeal."). Spigen makes this point on appeal, and Uncommon's only response is to call Spigen's claim "ironic" (given Spigen's Rule 26(a) troubles) and to assert baldly that "Uncommon has not waived arguments." That *425 also constitutes waiver. See id. (arguments that are "underdeveloped, conclusory, or unsupported by law" are waived).

Even forgiving waiver, though, the decision not to grapple with secondary meaning in the district court leads to another problem for Uncommon-it cannot show an issue of fact on the question. It is true, as the district court recognized, that there is not much to be found in this record regarding the amount of sales or advertising. We can also assume that the manner and length of use inures in Uncommon's favor. Still, on this record, no factfinder could reasonably decide that a substantial number of consumers think of "capsule" as the product of one cellphone-case maker. See Packman , 267 F.3d at 639 . The Martensen survey-the only direct evidence on this issue-demonstrates the opposite. Again, the survey shows that the vast majority of consumers do not associate "capsule" with cellphone cases at all.

Uncommon nevertheless tries to make do with the survey, arguing that it in fact supports a secondary meaning. It submits, for example, that because "capsule" had a similar amount of consumer recognition as a few other brands, it had "reached a level of distinctiveness." That conclusion does not follow, though, because none of the "capsule" comparators had much recognition either. See McCarthy on Trademarks § 32:190 (explaining that courts have recognized consumer recognition between 50 percent and 37 percent as sufficient to show secondary meaning, and lesser percentages are generally insufficient). Uncommon also submits that the survey wrongly lumps "capsule" into the list of other brand names (like Otterbox, Uncommon, and Spigen), when "capsule" is really a product. Yet the point remains: the survey demonstrates that a substantial number of consumers do not associate "capsule" with cellphone cases.

We therefore affirm the district court's decision to invalidate the contestable registration and to grant Spigen summary judgment. See Specht v. Google Inc. , 747 F.3d 929 , 936 (7th Cir. 2014).

2. Likelihood of Confusion

Alternatively, the district court granted Spigen summary judgment because it saw no likelihood of confusion in Spigen's use of the mark. This, too, is a question of fact, appropriate for summary-judgment resolution only if no reasonable factfinder could decide for Uncommon. CAE, Inc. v. Clean Air Eng'g, Inc. , 267 F.3d 660 , 677 (7th Cir. 2001) ; AHP Subsidiary Holding Co. v. Stuart Hale Co. , 1 F.3d 611 , 616 (7th Cir. 1993) ; Door Sys. , 83 F.3d at 171.

To decide if there is a likelihood of confusion, we ask whether consumers who might use either product would likely attribute them to a single source. Bd. of Regents of Univ. of Wis. Sys. v. Phx. Int'l Software, Inc. , 653 F.3d 448 , 455 (7th Cir. 2011). We use seven factors in making this decision: (1) the similarity between the marks; (2) the similarity of the products; (3) the area and manner of concurrent use; (4) the degree of care consumers are likely to use; (5) the strength of plaintiff's mark; (6) actual consumer confusion; and (7) the defendant's intent to "palm off" its product as that of another. Sorensen v. WD-40 Co. , 792 F.3d 712 , 726 (7th Cir. 2015). No one factor is dispositive. Fortres Grand Corp. v. Warner Bros. Entm't Inc. , 763 F.3d 696 , 704 (7th Cir. 2014). Usually, however, "the similarity of the marks, the defendant's intent, and actual confusion" are most important. Packman , 267 F.3d at 643.

Applying these factors, we agree with the district court that there is no issue of material fact as to the unlikelihood of confusion. To start, marks are troublingly *426 similar if a consumer viewing one party's mark would likely associate the product with the other party's product. AutoZone, Inc. v. Strick , 543 F.3d 923 , 929 (7th Cir. 2008) ; Nike, Inc. v. "Just Did It" Enters. , 6 F.3d 1225 , 1228-29 (7th Cir. 1993). We examine the entirety of the labels, not just the mark in isolation. Sorensen , 792 F.3d at 726 ; Henri's Food Prods. Co., Inc. v. Kraft, Inc. , 717 F.2d 352 , 355 (7th Cir. 1983).

The parties here use "capsule" in somewhat similar ways. Both generally use dark-colored, sans-serif font, and both use the mark alone and in combination with other modifiers. See AutoZone , 543 F.3d at 930 (similar font and use indicates similarity of marks). Spigen is quick to point out that there are obvious differences, too. The labels, for example, generally appear on different parts of the packages, which are themselves rather distinct. We agree that those facts make consumer confusion less likely. See Sorensen , 792 F.3d at 727 (different packaging for marks that had some similarities led to the "overall commercial impression" being "quite distinct"); Packman , 267 F.3d at 643 ("appearance and placement" of similar marks can make them distinct). Yet viewing the record in Uncommon's favor, as we must, the similarity of the marks favors Uncommon, albeit slightly.

The next few factors also slightly favor Uncommon. As to product similarity, Spigen concedes that the products are similar-both are cellphone cases for smart phones. The area and manner of use are also similar, although barely. Uncommon emphasizes that both parties sell nationwide through online retailers, like Amazon and eBay. We accept that as true. But the inference to which Uncommon is entitled is not a strong one. Selling through Amazon and eBay is ubiquitous, and it is hard to devise a more generalized customer base than nationwide cellphone owners.

Regarding consumer care, if a product is widely accessible and inexpensive it follows that consumers will exercise little care in discriminating among product makers. Sorensen , 792 F.3d at 730 ; CAE , 267 F.3d at 683. The cases here are easily available and inexpensive, between $ 15 and $ 40. Spigen submits that the gap in prices-$ 25-suggests that consumers will exercise more care in shopping for cases. That seems a stretch, but regardless, Uncommon is entitled at this stage to the other (and better recognized) inference-that the low cost and mass availability of cellphone cases means that consumers will not exercise much care in differentiating products. So this consideration also favors Uncommon.

The remaining factors, however, decisively favor Spigen. The "strength" of a mark usually refers to its distinctiveness. E.g. , Sorensen , 792 F.3d at 731 ; Quaker Oats , 978 F.2d at 959 . For reasons we have already explained, Uncommon's mark is neither inherently nor actually distinctive, and thus, it is a weak mark. See, e.g. , Fortres Grand , 763 F.3d at 704 (descriptive words are weak marks); Packman , 267 F.3d at 646 (same). There are other reasons the mark is weak, too. There is no evidence, for example, that Uncommon possesses robust "economic and marketing" power, a sign of mark strength. See AutoZone , 543 F.3d at 933 . The Martensen survey, in fact, demonstrates the opposite, given consumers' general nonrecognition of both Uncommon and "capsule." What is more, it is undisputable that neither party's cases feature the "capsule" mark prominently. The packaging often highlights the different and distinctive cases themselves; the "capsule" marks usually appear at the bottom or on the back of the packaging, in muted, nondescript typeface. We fail to see how there is a likelihood of confusion caused by a mark that is hardly featured.

*427 Uncommon also has presented no evidence of actual consumer confusion, although, as Uncommon correctly notes, such evidence is not a hard-and-fast requirement. CAE , 267 F.3d at 686. Additionally, there is no evidence that Spigen uses "capsule" in an effort to pass off its cases as Uncommon's. Spigen has used the mark for almost as long as Uncommon has, and Uncommon concedes that Spigen was unaware of Uncommon's trademark before this lawsuit.

The gist of Uncommon's counterarguments is simple: the district court wrongly used the summary-judgment motions to weigh evidence and make factual findings. That position misconstrues what the district court did. The court, as we have instructed, asked whether a reasonable factfinder could decide that there was a likelihood of confusion in light of all the record evidence. In doing so, it weighed the seven factors-not the evidence. See Sorensen , 792 F.3d at 726-32 (affirming summary judgment despite several factors weighing in appellant's favor); CAE , 267 F.3d at 678-687 (similar).

We agree with the district court's ultimate conclusion. The marks are, at most, barely similar, and they are too weak to confuse any purchaser, especially in the absence of evidence of any intentional freeloading. On this record, summary judgment was therefore appropriate.

III. Conclusion

For these reasons, we AFFIRM the district court's judgment.