Appeal from the United States District Court for the Western District of Wisconsin. No. 07 C 665-Barbara B. Crabb, Judge.
The opinion of the court was delivered by: Tinder, Circuit Judge.
Before FLAUM, WOOD, and TINDER, Circuit Judges.
Phoenix International Software created software that it called Condor and registered the CONDOR mark with the Patent and Trademark Office (PTO) for "computer software for on-line programming development, library management and system utilities functioning on mainframe systems." Phoenix used this mark since June 1978 and registered it in January 1997.
The delay in registration was due to a prior, separate registered CONDOR mark not at issue here. (We are capitalizing CONDOR when we refer to the mark, as the parties have in their briefs.)
The Board of Regents of the University of Wisconsin System (to whom we will refer simply as Wisconsin, because it is an arm of the state) registered its own CONDOR mark with the PTO in 2001 for "computer network operating system software, downloadable from a global computer network, that delivers large amounts of computational power by utilizing idle computing resources in a network of individual computer workstations . . . ." The question in this case is whether Wisconsin's CONDOR mark is likely to be confused with Phoenix's.
Phoenix filed a petition before the Trademark Trial and Appeal Board (TTAB) to cancel Wisconsin's registration on the grounds that Wisconsin's registration would create confusion in trade. See 15 U.S.C. § 1064 ("A petition to cancel a registration of a mark . . . may . . . be filed . . . by any person who believes that he is or will be damaged . . . by the registration of a mark on the principal register . . . . "). The confusion in trade allegation refers to 15 U.S.C. § 1052(d), which forbids the registration of a trademark that "[c]onsists of or comprises a mark which so resembles a mark registered in the [PTO] . . . as to be likely, when used on or in connection with the goods of the applicant, to cause confusion, or to cause mistake, or to deceive . . . ."
The TTAB granted the petition and cancelled Wisconsin's registration, finding that the marks are identical in every aspect. In such cases, even when goods or services are not competitive or intrinsically related, the use of identical marks can lead to an assumption that there is a common source . . . . In addition, the parties' software performs similar functions and, therefore, we cannot find that they are used in unrelated fields. Even sophisticated purchasers would likely believe that there is some relationship or association between the sources of the goods under these circumstances.
Phoenix Software Int'l v. Bd. of Regents of the Univ. of Wis. Sys., Cancellation No. 92042881, at 19 (T.T.A.B. Sept. 26, 2007) (quotations and citations omitted).
Wisconsin brought a civil action in federal district court pursuant to 15 U.S.C. § 1071(b) challenging the TTAB's decision. Phoenix counterclaimed, seeking dam-ages from Wisconsin for trademark infringement and false designation of origin under the Lanham Act (15 U.S.C. §§ 1114, 1125(a)); it also raised state law claims that it later voluntarily dismissed with prejudice. The district court dismissed Phoenix's federal counterclaims on sovereign immunity grounds and granted Wisconsin's motion for summary judgment, reversing the TTAB's determination. Phoenix appeals.
The standard of review in this case is complicated by its initial status as a matter before the TTAB. After losing there, Wisconsin had two options: take an appeal directly to the Court of Appeals for the Federal Circuit or institute an action in a district court. See 15 U.S.C. § 1071. The procedure for appeal to the Federal Circuit conforms to the familiar standard for administrative appeals. The parties present their case based on the closed record developed before the TTAB and the circuit court determines whether substantial evidence before the TTAB supported the decision. CAE, Inc. v. Clean Air Eng'g, Inc., 267 F.3d 660, 673, 675 & n.9 (7th Cir. 2001).
The choice to institute an action in the district court allows Wisconsin the benefit of expanding the record by offering new evidence to fend off Phoenix's cancellation claim. We have described the district court option as "both an appeal and a new action, which allows the parties to request additional relief and to submit new evidence." Id. at 673; see also Aktieselskabet AF 21. November 2001 v. Fame Jeans Inc., 525 F.3d 8, 12-13 (D.C. Cir. 2008) ("[T]he court may consider both new issues and new evidence that were not before the TTAB."). But a court's posture when considering an appeal and new issues is different, particularly in terms of issues of fact, which we will see are the key issues in this case. There is tension between the level of deference an appellate court pays to the fact-finder (in this case, an administrative body) and the parties' opportunity to present new evidence. See CAE, 267 F.3dat 674 ("[T]he district court is an appellate reviewer of facts found by the TTAB and is also a fact-finder based on new evidence introduced to the court.").
The district court, relying on CAE, adopted a defer-ential substantial evidence standard to review the TTAB's findings. See id. at 676. The court described its role as one that "affords deference to the findings of fact made by the board but considers the board's decision de novo to the extent the parties present new evidence. The board's findings of fact are properly reviewed under the standard set forth in [the Administrative Procedure Act (APA)] which requires the court to set aside findings and conclusions 'unsupported by substantial evidence.' " Bd. of Regents of the Univ. of Wis. Sys. v. Phoenix Software Int'l, Inc., No. 07 C 665, 2008 WL 4950016, at *8 (W.D. Wis. Nov. 18, 2008) (citing CAE, 267 F.3d at 674, 675-76). The district court's formulation of the interplay between both sets of evidence matches ours. "Although the district court's review of the TTAB's decision is considered de novo when the parties present new evidence and assert additional claims, the district court also must afford deference to the fact findings of the TTAB." CAE, 267 F.3d at 674. (We note, however, that the D.C. Circuit has recently held that because the Lanham Act, 15 U.S.C. § 1071(b), provides for judicial review of TTAB decisions, the APA is not directly applicable. See Aktieselskabet, 525 F.3d at 14. The D.C. Circuit did not address how this would affect its deference to the TTAB's fact-finding, except to note that courts have interpreted Dickinson v. Zurko, 527 U.S. 150 (1999), as requiring the application of the APA's substantial evidence standard to TTAB decisions. Aktieselskabet, 525 F.3d at 14 & n.2. We are one such court, CAE, 267 F.3d at 675, and neither party has made an issue of our standard of review.)
So an important part of this case should be to delineate the specific factual findings of the TTAB to which we owe deference and the new evidence, which we view in favor of the nonmoving party on a motion for summary judgment, and assess the impact of each on the summary judgment standard. Given that Phoenix was the prevailing party before the TTAB and that it was the nonmoving party on the summary judgment motion that was granted, we must say at the outset that the state of the facts presents a real obstacle to summary judgment in Wisconsin's favor. A party is entitled to summary judgment only if there exists "no genuine issue of material fact" and the moving party "is entitled to judgment as a matter of law." Fed. R. Civ. P. 56(c)(2). The entire issue in this case-the likelihood of confusion-is an issue of fact. AutoZone, Inc. v. Strick, 543 F.3d 923, 929 (7th Cir. 2008); CAE, 267 F.3d at 677; Barbecue Marx, Inc. v. 551 Ogden, Inc., 235 F.3d 1041, 1044 (7th Cir. 2000); Reed-Union Corp. v. Turtle Wax, Inc., 77 F.3d 909, 912 (7th Cir. 1996); McGraw-Edison Co. v. Walt Disney Prods., 787 F.2d 1163, 1167 (7th Cir. 1986).
To overturn the TTAB's decision, therefore, the district court (and this court, on de novo review) must find that substantial evidence did not support the TTAB's determination, or that a legal error clouded its understanding of the likelihood of confusion issue. If the TTAB's determination is supported by substantial evidence and there is no undermining legal error, Wisconsin must show that there is a killer set of facts, that Wisconsin neglected to bring to the TTAB's attention, that resolves the case in its favor even if we credit all the facts the TTAB found during its proceedings. To win on summary judgment, Wisconsin must show that the TTAB was not merely wrong, but wrong as a matter of law, and that the evidence was not only not substantial enough to support its decision but that the evidence compelled, without a new hearing by a fact-finder, the conclusion opposite to the TTAB's determination. (We ordinarily do not allow a party to hold evidence in reserve for appeal, but our standard of review is prescribed by the Lanham Act, 15 U.S.C. § 1071(b)(1). As the D.C. Circuit noted, the Lanham Act "does not require exhaustion of the administrative procedures itself" and therefore does not impose the traditional waiver rules on parties appealing TTAB decisions to the district court. Aktieselskabet, 525 F.3d at 14.)
We now proceed to the analysis. Only if the record, as credited by the TTAB and supplemented by the parties in the district court, reveals no genuine issue of material fact may we affirm the district court's decision.
A. Trademark Trial and Appeal Board
The TTAB found that Wisconsin's software "involve[s] using individual workstations in a network to better utilize idle computing resources," while Phoenix's soft-ware "functions on a mainframe system and . . . provides online programming development, library management and systems development." Phoenix Software Int'l, Cancellation No. 92042881, at 10. Before the TTAB was testi- mony from Wisconsin's mainframe coordinator that "an organization that did not have a mainframe or is not involved in developing software application[s] for main-frame computers" would have no use for Phoenix's software and testimony from Wisconsin's software creator that established that its software was not used on mainframe computers. ("[N]one of the [Wisconsin] Condor users that came back to us and asked any question told us that it's installed on a mainframe.")
But Phoenix presented evidence that its software, at least, was usable beyond the mainframe environment. Its sole shareholder, Fred Hoschett, testified in a deposition that "effectively we can run our software, unchanged, unaltered on a workstation, on someone's desktop, as if it were on a mainframe" and that the software "often" operates on a network of workstations, which he defined as a "LAN, WAN or some other network that allows the interconnection of these workstations." (LAN is an acronym for Local Area Network; WAN is an acronym for Wide Area Network.) He also testified that he has "many customers that use Condor that do not have mainframes."
In his deposition, Hoschett also read Wisconsin's description of its software from the University's web site:
Condor is a specialized workload management system for computer-intensive jobs. Like other full-featured batch systems, Condor provides a job queuing mechanism, scheduling policy, priority scheme, resource monitoring, and resource management. Users submit their serial and parallel jobs to Condor. Condor places them into a queue, chooses when and where to run the jobs based upon a policy, carefully monitors their progress, and ultimately informs the user upon completion.
Hoschett testified that this language "very much" concerned him because when he first read Wisconsin's description of its software he initially thought it was describing his product. Before the TTAB, Phoenix described its software as "a toolbox of functionality to be used essentially by anyone who uses a computer to assist them in doing their jobs, whether it be programming software, submitted batch jobs and queuing batch jobs, or managing the environment or managing the resources."
Based on this evidence, the TTAB found that there was "at least some evidence in the record that the parties' respective software performs the same general functions and the evidence does not demonstrate the goods are used in distinctly different fields," and that "there is no clear division between the parties' software that would cause us to conclude that these products are not related." The biggest difference between the products the TTAB found was that Phoenix's software was "used in a mainframe environment while [Wisconsin's] goods are used in a network of individual computer workstations." That distinction was "not necessarily" significant, because as one of Wisconsin's witnesses conceded, "there might be some incentive" to operate "in both environments."
The TTAB further found that both programs were "downloadable" and nothing in Phoenix's registration indicated otherwise, that consumers of either program were not ordinary consumers and had "some level of skill and sophistication to the extent that they are programming mainframe computers or networking computer workstations to increase computational resources," and that their "purchases would be made with some care." (The TTAB classified consumers of both programs as purchasers, even though Wisconsin distributes its software under an open source, free license. This difference is irrelevant to our analysis.) The TTAB also found that there was no evidence that any consumer was actually confused as to the source of either product despite evidence that Phoenix offered to support its claim to the contrary (including Internet searches for "Condor Software" that returned results for both products and a Wikipedia entry for Wisconsin's product, with no entry for Phoenix's).
The TTAB found that both parties' marketing practices were "relatively limited." A Wisconsin witness testified that "[w]e don't do any advertising with Condor" and Phoenix's controller estimated that Phoenix spent approximately $65,000 in marketing, focused on attending trade shows and printing brochures and other marketing materials. But Wisconsin told the TTAB that it was expanding its operations, which the TTAB found made the chances of confusion more likely.
Based on these factual findings and inferences drawn therefrom, and applying relevant law, the TTAB found that Phoenix had met its burden of proving the likelihood of confusion and granted Phoenix's petition to cancel Wisconsin's registration of the CONDOR mark. It noted that "the marks are identical in every aspect" and that both products perform similar functions. The TTAB concluded that "we cannot find that they are used in unrelated fields" and that "[e]ven sophisticated purchasers would likely believe that there is some relationship or association between the sources of the goods under these circumstances."
At the district court, Phoenix attempted to further supplement the record with evidence that bolstered its case that the Condor software products performed over-lapping functions. Wisconsin, according to Phoenix's sole shareholder Fred Hoschett, had struck deals with IBM to make Wisconsin's version of Condor available on PC-based mainframes. Hoschett also provided a list of customers operating Phoenix's version of Condor on PC-based mainframes. But the district court rejected the proffered evidence on the grounds that it was filed with Phoenix's reply brief instead of in its initial proposed findings of fact, in violation of a local rule. Phoenix does not argue that it did, in fact, comply with the local rule.
Based on the evidence it did consider, the district court found that it was undisputed that Phoenix's soft-ware cannot run on a network of workstations that are not connected to a mainframe system, but that the software may be run on non-mainframe computers through the use of emulation software. The court found that the Phoenix software allows "its users to submit batch jobs to local and remote computers through a network of computers to more effectively utilize and balance the available computing cycles," that Phoenix had 100 or fewer active licensees of its software, but that licenses cost anywhere from $30,000 to $300,000 a year (so that 100 licenses can be very profitable), and that Phoenix advertises at trade shows, on the Internet, and through brochures.
Phoenix's customers, the court found, must be specialized because mainframe computer systems "are generally expensive computing systems that are extremely reliable and secure and capable of enormous throughput," they are "centrally managed and maintained," and a choice of software for use on a mainframe "requires careful consideration." In other words, customers don't buy mainframe software on a lark. The end-users of software like Phoenix's are "mainframe systems administrators and mainframe systems application developers." The community of mainframe administrators and developers is a tight-knit one (in the words of the district court, a "niche") that learns about products through word-of-mouth advertising, mainframe trade shows and conferences, and the advice of consulting firms. Phoenix spent approximately $65,000 on marketing in 2000; that number was virtually unchanged in 2003. Phoenix's competitors are IBM and Computer Associates.
As for distribution and customer overlap, the district court found that Wisconsin distributes its software under an open software license and links together "a network of individually owned computer resources" to create a system wherein those computers trade operating capacities. The court determined that Wisconsin's software does not run in mainframe environments, but did not resolve whether a mainframe could be part of the grid on which Wisconsin's software operates. The court found that users of Wisconsin's software are generally systems operators of scientific research groups, but since the software is free and available for download others may use it. Because of this, the court found that users of Wisconsin's software are tough to identify; Wisconsin's estimates place the total number of users in the tens of thousands. They generally must have a "systems-level understanding" of computers in order to make Wisconsin's Condor program work. Users include the "high energy physics community, the DOE [Department of Energy] National labs, biology and computer science departments, and industrial groups." According to the district court, 3738 copies of Wisconsin's software were downloaded in 2000; by 2004, the number of downloads grew to 15,155, an increase of more than 400 percent. A promotional program offered by the University of Wisconsin, "Condor Week," showed similar growth from a one-day event attracting twenty participants (presumably it was then called "Condor Day") to a four-day event with more than 150 participants.
III. Likelihood of Confusion
The question here is whether consumers were likely to be confused by Wisconsin's and Phoenix's concurrent use of the CONDOR marks. As the district court framed it, "the only question is whether the identical marks used in the general field of computing create a likelihood of confusion for consumers . . . or whether the differences in the computer products for which the software is sold, in the trade channels, in the conditions under which sales of products are made and other factors eliminate the possibility of confusion . . . ." Bd. of Regents of the Univ. of Wis. Sys., 2008 WL 4950016, at *1. The district court found that the TTAB "erred when it considered the actual nature of the parties' goods and misapplied the burden of proof to its determination of a likelihood of confusion." Id. Because it determined that the TTAB opinion was erroneous, and Phoenix's evidence before the district court was relevant only to the analysis adopted by the TTAB, the district court found for Wisconsin.
We agree that the key issue in this case is the likelihood of confusion between the products described by the two marks. We are not considering whether the products themselves perform the same functions, but whether consumers, and specifically consumers who would use either product, would be likely to attribute them to a single source. AutoZone, 543 F.3d at 931. "[D]issimilarity is not dispositive of the likelihood of confusion inquiry. A likelihood of confusion may exist even if the parties are not in direct competition or their products and services are not identical. Rather, because the rights of an owner of a registered trademark extend to any goods that might be, in the minds of consumers, 'related,' i.e., put out by a single producer, the more accurate inquiry is whether the public is likely to attribute the products to a single source." CAE, 267 F.3d at 679 (citations omitted). This, as we shall see, is a key point of divergence between our analysis and that of the district court.
There are a series of multiple-factor tests that are used across the circuits to determine the likelihood of confusion. We use this one:
1. Similarity between the marks in appearance and suggestion.
2. Similarity of the products.
3. The area and manner of concurrent use.
4. The degree of care likely to be exercised by consumers.
5. The strength of the plaintiff's mark.
6. Whether actual confusion exists.
7. Whether the defendant intended to "palm off" his product as that of the plaintiff.
See AutoZone, 543 F.3d at 929. The Federal Circuit uses a different set of factors (referred to by the TTAB in its opinion). See In re E. I. DuPont DeNemours & Co., 476
F.2d 1357, 1361 (C.C.P.A. 1973). They essentially cover the same ground as our factors, and neither party suggests that the differences between the factors affect the outcome of the case. As noted above, likelihood of confusion is a question of fact for the jury to determine. AutoZone, 543 F.3d at 929. "The question of fact may be resolved on summary judgment only if the evidence is so one-sided that there can be no doubt about how the question should be answered." Id. (quotations omitted).
The district court overruled the TTAB in two key respects. First, it found that the TTAB erred when it considered the "actual nature of the parties' goods" rather than the goods as they were described in their respective registrations. This error, the court found, combined with the lack of evidence that the products were sufficiently related to cause confusion, removed a key rationale that supported the TTAB's decision. The district court also found that the TTAB erred in placing the burden on Wisconsin to prove that the parties' goods are distinct when it considered the way the products were used or sold. The district court found that the burden should have been placed on Phoenix, who sought to cancel a presumptively valid registered mark, and found that Phoenix did not present sufficient evidence to meet this burden (because the district court rejected all the TTAB's findings on the products' similarities as irrelevant).
The district court's analysis addressed several factors outlined in our test, but focused mainly on the similarity of the products and the area and manner of their use. Wisconsin does not (and cannot) dispute that the marks are identical and Phoenix has not been able to press the argument that actual confusion existed between the two marks-the TTAB found otherwise and evidence to the contrary was excluded by the district court. Funda-mentally, then, the district court's rejection of the TTAB's decision rested on a disagreement with its analysis of the products' similarity and their manner of use.
The district court examined the manner in which both products were described in their registrations, disregarded most evidence of their actual use, and focused on whether a sophisticated consumer would be likely to confuse the product described as "computer software for on-line programming development, library management and system utilities functioning on mainframe systems" with a product described as "computer network operating system software, downloadable from a global computer network, that delivers large amounts of computational power by utilizing idle computing resources in a network of individual computer workstations." It appears that the district court mistakenly assumed that the similarity of the products' functions was the dispositive issue in the case; this misapprehension was magnified by the district court's error in confining its examination to the registration of the parties' products. Instead of a focus on the description of the goods in the trademark registry, the proper analytical framework, according to our cases, is "whether the parties' products are the kind the public might very well attribute to a single source." AutoZone, 543 F.3d at 931 (citations omitted); McGraw-Edison Co., 787 F.2d at 1169. "The rights of an owner of a registered trademark extend to any goods or services that, in the minds of consumers, might be put out by a single producer. Thus, a likelihood of confusion may exist even if the parties are not in direct competition, or their products and services are not identical." AutoZone, 543 F.3d at 931 (quotation and citation omitted).
This is a proposition we have repeatedly laid out in the many cases in which we've considered the likelihood of confusion between goods described by similar or identical marks. See id. at 931; CAE, 267 F.3d at 679; Eli Lilly & Co. v. Natural Answers, Inc., 233 F.3d 456, 463 (7th Cir. 2000); McGraw-Edison Co., 787 F.2d at 1169. It was also recognized by the TTAB. "[I]t is not necessary that the goods and/or services be similar or competitive, or even that they move in the same channels of trade to support a holding of likelihood of confusion." Phoenix Software Int'l, Cancellation No. 92042881, at 9 (citing Hilson Research Inc. v. Soc'y for Human Res. Mgmt., 27 U.S.P.Q. 2d 1423, 1432 (T.T.A.B. 1993)). "It is sufficient that the respective goods and/or services are related in some manner, and/or that the conditions and activities surrounding the marketing of the goods and/or services are such that they would or could be encountered by the same persons under circumstances that could, because of the similarities of the marks used therewith, give rise to the mistaken belief that they originate from or are in some way associated with the same producer." Hilson Research Inc., 27 U.S.P.Q. 2d at 1432.
Once we accept this proposition, we see that Wisconsin's focus on the descriptions in the registry, adopted by the district court, is a bit of a red herring. Our cases make clear that products don't even have to perform similar functions, much less be described identically, for a likelihood of confusion to exist. In McGraw-Edison we found that a likelihood of confusion could exist between a mark for electrical fuses and the mark on Disney's merchandise for the movie Tron. McGraw-Edison Co., 787 F.2d at 1169. In CAE, we found that a likelihood of con-fusion could exist between a mark registered by a company that designed and manufactured sophisticated measuring equipment and a mark registered by a company that tested facilities for compliance with pollution laws. CAE, 267 F.3d at 679. In AutoZone, we found that a likelihood of confusion could exist between an auto-parts retailer's mark and the mark of a oil change and carwash operator. AutoZone, 543 F.3d at 931. All of these cases featured products with a far more tenuous similarity than that at issue here. We note that in these cases, the products had identical marks (with the exception of AutoZone, which featured two marks with identical "zone" suffixes). It is possible that we, as a circuit, have historically assigned too much weight to the fact that marks are identical. Cf. M2 Software, Inc. v. M2 Commc'ns, Inc. 450 F.3d 1378, 1385 (Fed. Cir. 2006). But, perhaps that is because the issue is fact-bound and identical marks often create triable issues of fact regarding the various ways a product is marketed. Regardless, neither party attacks our trademark standard as inconsistent with the Lanham Act. Furthermore, the identical nature of the marks weighed heavily in the TTAB's analysis. Phoenix Software Int'l, Cancellation No. 92042881, at 19 (citing In re Shell Oil Co., 992 F.2d 1204 (Fed. Cir. 1993)). The TTAB also cited Amcor, Inc. v. Amcor Indus., Inc., 210 U.S.P.Q. 70, 78 (T.T.A.B. 1981), for the proposition that when considering identical marks, "the relationship between the goods on which the parties use their marks need not be as great or as close as in the situation where the marks are not identical or strikingly similar."
The approach adopted by the district court limiting its consideration to the products as they are described is too formalistic and ignores the requirement that the products are to be examined as they appear to the consumer. "[O]ur inquiry in comparing the two products is not whether they are interchangeable, but whether the parties' products are the kind the public might very well attribute to a single source." Eli Lilly & Co., 233 F.3d at 463 (emphasis added) (quotation omitted); see also AutoZone, 543 F.3d at 931; McGraw-Edison, 787 F.2d at 1169 ("In finding the parties' product lines to be 'entirely unrelated' the district court apparently ignored the question of whether the purchasing public might believe a single source could produce both [products].").
Octocom Systems, Inc. v. Houston Computer Services, Inc., 918 F.2d 937, 942 (Fed. Cir. 1990), the Federal Circuit case that the district court relied on, is not to the contrary. In that case, the registrant whose mark was challenged tried to supplement the registration by showing that the registrant really intended its mark to cover a narrow class of goods, rather than the unlimited coverage that the actual mark sought. The Federal Circuit found that the subsequent attempt to limit the registration was improper and the court was required to consider only the goods as described in the registration. The court then found that there was a likelihood of confusion between the products because the registrant's original application "encompassed modems and computer programs" and thus conflicted with the petitioner's registration of a similar mark for computer programs. Furthermore, "the record supports no other factual findings but that modems and computer programs are commonly used together in networking, could come from a single source, and be identified with the same mark. Thus, [the attempted registrant's] elimination of 'computer programs' from its application leaving only 'modems' was pointless maneuvering." Octocom Sys., 918 F.2d at 943.
The likelihood of confusion in Octocom stemmed not from the fact that the registrations were identical, but from the registrations' coverage of similar products. In fact, in Octocom, the Federal Circuit recognized that the similarity of the products described in the registration may be "expressly or inherently" reflected. Id. at 942. The Federal Circuit noted that "[e]vidence that the goods of the applicant and opposer, as identified in the respective application and registration, are the types of goods that would be expected to move in different trade channels or be sold to different classes of purchasers may be material and relevant." Id. at 943. So, the relevant question for us is whether a product that runs on a mainframe and a product that runs on a network of computers are the types of goods that are similar enough to be attributed to a single source by consumers likely to use one of those products. While we must consider the marks as they are described, it would be a mistake to bar any evidence of their actual use as irrelevant. After all, one of the factors in the test is the "area and manner of concurrent use." See AutoZone, 543 F.3d at 929. And, as noted, the actual use of the product is relevant to explain the meaning of the terms used in the registration. See Octocom Sys., 918 F.2d at 943; see also CAE, 267 F.3d at 681-82; Forum Corp. of N. Am. v. Forum, Ltd., 903 F.2d 434, 442 (7th Cir. 1990). Indeed, the TTAB explicitly declared that it was considering the nature of the parties' goods "[t]o the extent that these facts provide some information about the market and purchasers of these goods." Phoenix Software Int'l, Cancellation No. 92042881, at 9.
Rejecting the district court's approach, however, does not dispose of the case. "Because confusion is a factual matter, the plaintiff must produce proof; a theory about how consumers might be confused will not do, unless evidence supports the theory." Reed-Union Corp., 77 F.3d at 912. Even as we consider the products' actual uses to determine the meaning of the marks' registrations, or more specifically to determine whether there is a possibility of confusion between the two products as described in their registrations, Wisconsin would still prevail if there is no evidence to support the idea that a supplier of software to a mainframe system (the manner of use specified in Phoenix's registration) can be confused with a supplier of software to a network of individual computer workstations (the manner of use specified in Wisconsin's registration).
The TTAB relied on three key factors in its order to cancel Wisconsin's registration. First, and most importantly, it relied on the fact that the marks are identical. Wisconsin does not challenge this finding (and as noted, this seems to be a compelling factor in our circuit). Second, the TTAB found that the parties' software performs similar functions, which precluded a finding that they were used in unrelated fields. Third, it found that "sophisticated purchasers would likely believe that there is some relationship or association between the sources of the goods under these circumstances." Phoenix Software Int'l, Cancellation No. 92042881, at 19. Ultimately, the TTAB concluded that sophisticated purchasers would associate the two products. Mapping the TTAB's findings onto our factors, we see that the TTAB's finding of a likelihood of confusion was based on the identical nature of the marks, the manner in which both parties' products are used, and the similarity of the products.
The district court dismissed these findings on the basis that all analysis of the products should have been confined to the terms in the registries. As we've explained, this was an error. Moreover, because the TTAB was using the correct factor-based balancing test, the district court was incorrect to dismiss the TTAB's findings on the area and manner of use factor by declaring that the TTAB had misallocated the burden of proof. Because the TTAB found that the similarity of the marks and products weighed in Phoenix's favor, the area and manner of use tipped the balance in favor of Phoenix. The TTAB noted specifically in its holding that "a presumption of validity attaches to a service mark registration, and the party seeking cancellation must rebut this presumption by a preponderance of evidence." Id. at 18. "We hold that petitioner has met its burden." Id. at 19.
Once we reinstate the TTAB's findings, we see that Phoenix has offered sufficient evidence to survive sum-mary judgment on the issue of confusion. The TTAB credited Phoenix's evidence that its mainframe software can operate unaltered on a network of workstations. The TTAB found that both programs perform similar functions to the extent that Phoenix's representative was himself confused by Wisconsin's public description of its product; he thought it was Phoenix's own. The TTAB further found that, as conceded by Wisconsin's witness, there was "some incentive" to operate in both the mainframe and network environments. The TTAB also found that both products were delivered in the same manner, and that the same customers are likely to encounter ...