Surveys in patent cases: watch your step for these 4 pitfalls

Surveys are often a part of a trademark or trade dress case, whether relating to secondary meaning (association of a mark with a particular source) or likelihood of confusion. See, e.g., Converse, Inc. v. ITC, 909 F.3d 1110, 1120 (Fed. Cir. 2018) (noting that trade dress association with source "typically measured by customer surveys"); see also Braun Inc. v. Dynamics Corp. of Am., 975 F.2d 815, 828 (Fed. Cir. 1992) (surveys and other empirical evidence of likelihood of confusion have "considerable significance" in determining trademark and trade dress infringement). Although some commentators calculate that surveys may not be quite as ever-present in trademark cases as many assume, they are certainly a common component. See Katie Brown Ph.D., Natasha T. Brison, and Paul J. Batista, An Empirical Examination of Consumer Survey Use in Trademark Litigation, 39 Loy. L.A. Ent. L. Rev. 237 (2019).

By contrast, surveys are a rarer sight in patent cases. Even when it comes to design patents (which are themselves far less common), courts have acknowledged that the different quantum of proof means there is less need for empirical evidence compared to trademark and trade dress cases. Braun, 975 F.2d at 828. But there is growing recognition of the utility of surveys to support damages and injunction arguments in patent matters. See, e.g., Estech Sys. v. Target Corp., 2021 U.S. Dist. LEXIS 135438, *31 (E.D. Tex. July 21, 2021) ("on the issue of damages survey results can be used as relevant evidence of consumer demand").

Patent litigators embarking into this less familiar area would be wise to tread carefully. A few common pitfalls to watch out for follow.

1. Strongly consider control groups.

If a survey is being offered to show causation (e.g., "Does this fact or statement about the product's capabilities make you more or less likely to purchase this product?"), it should have a control group. A Southern District of New York court recently called the use of a control group the "gold standard" for eliminating background noise. Edmondson v. RCI Hosp. Holdings, Inc., 2020 U.S. Dist. LEXIS 55421, at *22 (S.D.N.Y. Mar. 30, 2020) (internal citation omitted). The plaintiffs in that case argued that their survey did not test causation, therefore concluding that a control group was not required. The court rejected that argument, however, and excluded the survey entirely for that reason, among others. Another court similarly found last year that the lack of a control group may render an inference of causation impossible and, therefore, the survey inadmissible. Colangelo v. Champion Petfoods USA, Inc., 2022 U.S. Dist. LEXIS 60489, at *32 (N.D.N.Y. Mar. 31, 2022). A control group is not always needed, but be cautious about skipping one, and make sure it is an intentional and educated choice.

2. Make sure to test the correct population.

Another easy way to trip up is to target either too broad a population or the wrong population. Courts will sometimes find that this goes to the weight rather than the admissibility of the survey, but if the problem is severe enough, it can lead to a court excluding a survey outright. For example, a District of Delaware court excluded a survey expert's opinion as unreliable in a patent case based on, among other things, a failure to identify and account for the correct population. Parallel Networks Licensing, LLC v. Microsoft Corp., 2017 U.S. Dist. LEXIS 29613, at *10-11 (D. Del. Feb. 22, 2017). In that case, the survey did not provide any details on the panel used, so it was "impossible to know whether the participants of the underlying panel deviate[d] in some meaningful way from the target population." Id. Similarly, a Southern District of New York court excluded a "wholly unlimited survey" that made "not even a modest attempt to narrow the universe of U.S. adults" to the appropriate consumers. Marksman Sec. Corp. v. P.G. Sec., Inc., 2021 U.S. Dist. LEXIS 259366, at *8 (S.D.N.Y. Dec. 15, 2021). Make sure you work with your survey expert to focus on the correct universe.

3. Engage the right expert.

Avoid the temptation to simply have a technical expert wade into the survey world, even if only to rebut the other party's survey. A litigant took that approach in the Northern District of California. The court found that while the expert was qualified as a technical expert, he was not qualified to opine as to survey design or survey respondent perception--for example, what a "proper survey would have shown" or what the subjects may have believed. Droplets, Inc. v. YAHOO! Inc., 2021 U.S. Dist. LEXIS 259664, at *18 (N.D. Cal. July 1, 2021). Take the extra time to engage the right expert for the task.

4. Tie the survey to the asserted claims.

Survey questions necessarily require some conversion into plain English, but the description cannot be broadened or changed so much that it is no longer tied to what the asserted patent claims. The same District of Delaware court discussed above noted this problem in the survey it excluded, explaining that while the survey "generally relate[d] to the accused products," it did not support the assertion that survey respondents used the products in an infringing manner because it "failed to address even a single limitation of the asserted claims." Parallel Networks Licensing, 2017 U.S. Dist. LEXIS 29613, at *4-5. Similarly, an Eastern District of Texas court excluded an expert's apportionment theory (and therefore his entire reasonable royalty opinion) because it relied on a survey without explaining how it accounted for the value of the patented feature. Spike v. Huawei Techs. Co., 2016 U.S. Dist. LEXIS 199913, at *10 (E.D. Tex. Oct. 14, 2016). Be careful to ensure that the language in your survey accurately reflects the patented invention, not just the general area in which the patented invention operates.

Patent litigators may be wary of surveys, because where something is less familiar, it is easier to make an unforced error. But when used correctly, surveys can be a powerful tool to support a patent case.