Trademark confusion surveys depend on a sound connection between the consumers whose perceptions matter to a claim and the respondents whose answers are actually analyzed. This article develops Universe Correspondence as a way of organizing the evaluation of that connection, and it does so around two related questions. The first is whether the universe has been defined correctly, meaning whether the population chosen for study reflects the consumers whose perceptions are relevant to the type of confusion alleged. The second concerns sampling frame error, that is, whether the frame from which respondents are drawn adequately captures the population that was defined. The analysis presented here considers underinclusion and overinclusion along with practical matters such as the limited information available about how online panels are assembled, the screening performed by panel providers, and the screening questions within the survey itself. Perfect correspondence is unlikely in applied research. The pertinent question is whether departures from the intended universe could affect estimates of confusion, not simply whether departures exist.
Universe Correspondence
A trademark confusion survey is useful to the extent that its design supports inferences about consumers whose perceptions are relevant to the dispute. A survey can be executed carefully, contain clear questions, and produce precise estimates, yet still address an inappropriate universe. Those familiar with market segmentation and marketing research recognize that a sample need not perfectly replicate every characteristic of a specified population to provide useful information. However, recognizing that samples are rarely perfect does not mean that all samples are acceptable or that bias of any size or form must be accepted as an inevitable part of the process. Evaluating the universe therefore requires both establishing legal relevance and using appropriate sampling methods. The concept of Universe Correspondence offers a structured way to examine these issues. It asks whether the legally relevant population has been translated into appropriate eligibility criteria and whether the recruitment and data-collection process yields respondents who correspond sufficiently to that population for the proposed inference.
Defining the Legally Relevant Population
The legally relevant population hinges on the type of confusion being considered. The corresponding populations are summarized in Exhibit 1. Although the appropriate universe may vary with the nature of the alleged infringement, two principles hold.
First, the legally relevant population is derived from the underlying theory of confusion, not merely from general purchase behaviors within a category. For example, in traditional point-of-sale confusion, the concern is that consumers targeted by the junior mark may purchase the product believing it to be manufactured, sponsored, or affiliated with the senior mark. Accordingly, the relevant population centers on consumers targeted by the junior mark. Hence, if this related to running shoes, it would be critical to survey current and potential purchasers of the junior mark, not merely all people who have (or intend to) purchase running shoes. For example, if there is a luxury brand that is concerned that a junior mark is able to benefit from the goodwill established by them (i.e., the senior mark), the inquiry should consider confusion that may arise when a consumer targeted by the junior mark makes a purchase decision. Market segmentation recognizes that brands do not necessarily appeal to all consumers, and the legally relevant universe should reflect this distinction.
Second, the universe should be defined in a manner that accurately reflects the relevant marketplace, avoiding both over-inclusion and under-inclusion. Over-inclusion occurs when the universe encompasses individuals whose perceptions are not germane to the alleged infringement. Returning to the running shoe example, suppose the junior brand markets premium performance running shoes, typically priced above $150, to consumers seeking footwear for competitive running and serious athletic training. Defining the universe as all current or prospective purchasers of running shoes would potentially include large numbers of casual runners who purchase inexpensive footwear primarily for walking, recreation, or everyday use. These consumers may differ substantially from the junior brand’s target market in terms of product knowledge, purchasing motivations, brand familiarity, and attention to product attributes. Their inclusion could distort estimates of confusion among the consumers whose perceptions are legally relevant.
Conversely, under-inclusion occurs when the universe excludes consumers who are legitimately part of the junior brand’s market. For example, restricting the universe to individuals who have competed in a marathon within the past year would exclude serious recreational runners, half-marathon participants, and other performance-oriented consumers who regularly purchase the junior brand’s products. Such a restriction could produce a sample that is disproportionately experienced or knowledgeable, potentially understating confusion among other relevant purchasers. In both instances, the problem is not simply that the sample differs from the broader population, but that the resulting estimates may fail to represent confusion within the legally relevant marketplace.
Forward and Reverse Confusion
Exhibit 1 presents the relevant universe for different forms of confusion. For forward and reverse confusion, the nature of the universe is reasonably well settled. In traditional forward confusion, the relevant population generally consists of current and prospective purchasers of the junior user’s goods or services, whereas in reverse confusion, the relevant population shifts to current and prospective purchasers of the senior user’s goods or services (Barber & Yaquinto, 2022; Upadhye, 1997). Nevertheless, identifying the appropriate party’s customer base is only the starting point for defining the universe, as the sample must also reflect the characteristics of consumers reasonably likely to encounter the allegedly infringing mark under relevant marketplace conditions (Diamond and Franklyn, 2014).
Initial Interest Confusion
The identification of the legally relevant population for initial-interest confusion requires particular attention to the consumer’s information-search and evaluation process. From a consumer behavior perspective, initial-interest confusion arises when exposure to an allegedly infringing mark produces mistaken beliefs about source, sponsorship, or affiliation during the early stages of consumer decision-making, potentially influencing attention, consideration, or subsequent search behavior, even if the confusion is resolved before purchase. Scott (2013) examines this phenomenon in the context of online search environments, demonstrating how firms can exploit established trademarks to attract consumers who might otherwise have directed their attention toward a competing brand. He further argues that such practices have implications for customer equity and customer lifetime value, as firms may lose opportunities to acquire or retain customers through the appropriation of their established brand associations. Grynberg (2004) similarly emphasizes the relationship between initial-interest confusion and consumer search costs, while Goldman (2005) questions the assumption that particular search terms necessarily reveal consumers’ purchasing intentions. McKenna (2012) extends this discussion by distinguishing confusion that meaningfully interferes with consumer decision-making from mere exposure to competing alternatives.
These perspectives suggest that the relevant population for an initial-interest confusion study should be defined with reference to the consumers who participate in the particular search or evaluation activities in which the alleged confusion occurs, rather than simply by identifying the target market of either party. Importantly, initial-interest confusion is not inherently limited to forward or reverse confusion. Whereas forward and reverse confusion describe the direction of mistaken source attribution, initial-interest confusion concerns the stage of the decision-making process at which that mistaken attribution occurs. Thus, forward initial-interest confusion may arise when consumers encounter the junior mark and mistakenly associate it with the senior user, while reverse initial-interest confusion may arise when consumers encounter the senior mark and mistakenly associate it with the junior user. Although the direction of confusion remains relevant to identifying the appropriate population, neither direction necessarily dictates that the universe be restricted to one party’s existing or prospective customers. Rather, the appropriate universe should reflect the intersection of the alleged direction of confusion, the relevant marketplace exposure, and the consumers whose search or evaluation activities create a reasonable opportunity for confusion to occur.
Post-Sale Confusion
The identification of the legally relevant population becomes particularly challenging in the context of post-sale confusion, where the individuals whose perceptions are relevant may extend beyond purchasers and prospective purchasers of the junior mark to include third parties who encounter the allegedly infringing product after its purchase. Unlike traditional point-of-sale confusion, which generally focuses on consumers making purchase decisions, post-sale confusion may arise when individuals observe a product being worn, displayed, or used in ordinary marketplace settings and mistakenly attribute its source, sponsorship, or affiliation to the senior mark. One view (Barber & Yaquinto, 2022, based on General Motors Corp. v. Urban Gorilla LLC) would limit the universe to naïve observers who might ultimately purchase either the plaintiff’s or the defendant’s product. Nonetheless, courts have recognized that such confusion may be legally relevant even when the observer is not a prospective purchaser of the junior product (Insty Bit, Inc. v. Poly-Tech Industries, Inc., 1996). McCarthy (1999) similarly argues that the general public should not be categorically excluded from consideration in post-sale confusion claims. Even if the universe were limited to current and potential purchasers of either the junior or senior mark, the question of what proportion of each group should be included would remain. Should representation be equal, based on market share, or determined by another criterion?
As Raustiala and Sprigman (2018) explain, post-sale confusion involves a set of circumstances in which observers encounter a product, recognize potentially source-identifying characteristics, and form judgments about its origin. At a minimum, then, the defined universe and its operationalization should account for the likelihood that an individual would be an observer. As such, a more defensible approach would begin by identifying the circumstances in which the allegedly infringing product is ordinarily encountered and determining which individuals have a meaningful opportunity for such exposure.
Exhibit 1
Types of Consumer Confusion and Appropriate Survey Universe
| Type | Definition | Appropriate Survey Universe |
| Forward (Traditional) Confusion | Occurs when consumers are likely to believe, at the time of purchase, that the junior mark’s goods originate with, are sponsored by, or are affiliated with the senior mark (i.e., the actual trademark owner). This is also referred to as “traditional confusion.” | Actual or prospective purchasers of the goods or services bearing the junior mark within the relevant market, including consumers likely to encounter the challenged mark at the point of purchase. Eligibility should reflect the pertinent product category, channels, geography, and purchasing period. |
| Reverse Confusion | Occurs when consumers mistakenly believe that the senior user’s goods originate with, are sponsored by, or are affiliated with a well-known/established junior user. The alleged harm is that the junior user’s commercial dominance overwhelms the senior user’s trademark identity, diminishing its goodwill and ability to control its reputation. | Actual or prospective purchasers of the senior user’s goods or services in the relevant market who may encounter the senior mark and mistakenly associate it with the commercially dominant junior user. The universe is anchored to the senior user’s market, not merely the junior user’s customer base. |
| Initial-Interest Confusion | Occurs when a similar trademark captures consumer attention through mistaken source, sponsorship, or affiliation. Even if the consumer is not confused at the time of purchase, the alleged harm is the diversion of consumer attention during the search process. | Consumers engaged in, or reasonably likely to engage in, the relevant product-category search or selection process who could encounter the junior mark before purchase. The universe should reflect the search context in which the alleged initial diversion occurs. |
| Post-Sale Confusion | Occurs when an observer (not the purchaser) sees the junior mark in public and believes the source of manufacture, affiliation, or sponsorship is the senior mark. The alleged harm may involve a reduced likelihood of purchase of the senior mark or a reduction in prestige. | Relevant nonpurchasing observers who are reasonably likely to encounter the junior-marked product in the alleged post-sale setting, such as bystanders or other members of the relevant consuming public. Where the asserted harm involves future purchasing behavior, the universe should also reflect the relevant pool of actual or prospective category purchasers. |
Operationalizing the Universe: Defining The Sampling Frame
Even when the legally relevant universe has been appropriately defined, errors may arise in translating that universe into an accessible sampling frame. A sampling frame is the list, panel, or group from which a survey draws respondents, and in practice it rarely aligns perfectly with the reference universe it is meant to represent. Its adequacy depends on the extent to which it corresponds to the intended target population (Iacobucci & Churchill, 2023; Malhotra, 2022). Sampling frame error arises when the group from which respondents are actually drawn does not match the population the researcher set out to study (Malhotra, 2022). The consequence can be that people the survey is meant to represent are systematically left out, that people outside the universe are mistakenly brought in, or both. Accordingly, efforts should be made to align the principal characteristics of the defined universe with those of the sampling frame.
General Issues of Inclusion
Errors in this translation may result in underinclusion, whereby relevant segments of the universe are excluded from the frame, or overinclusion, whereby the frame includes individuals who do not satisfy the universe definition (Weisberg, 2005).[1] For example, suppose the legally relevant universe consists of current and prospective purchasers of premium running shoes, but the sampling frame is constructed using membership lists from competitive running clubs. Although these individuals may constitute a relevant segment of the universe, the frame would exclude other eligible consumers, such as recreational runners and fitness enthusiasts who purchase premium running shoes but do not belong to organized running clubs. Conversely, constructing the frame from a general athletic footwear customer database could introduce substantial overinclusion by including consumers whose purchasing behavior is limited to inexpensive casual footwear.
Importantly, these two forms of inclusion error differ in their potential consequences for inferential validity. Overinclusion does not necessarily introduce bias if ineligible individuals can be reliably identified and excluded through appropriate screening procedures. For example, a broad athletic footwear database may constitute a defensible starting frame if it provides adequate access to the legally relevant universe and screening procedures successfully distinguish eligible from ineligible consumers.
Underinclusion, by contrast, presents a more fundamental challenge because relevant consumers who are absent from the frame have no opportunity to be selected, regardless of the rigor with which subsequent sampling procedures are implemented. Nevertheless, underinclusion does not necessarily imply that resulting confusion estimates are biased. The magnitude and direction of any bias depend on the extent of the underinclusion and whether excluded consumers differ systematically from included consumers in their likelihood of experiencing confusion. For example, excluding recreational runners from a premium running shoe survey may have little effect on estimated confusion if their perceptions are comparable to those of competitive runners. However, if recreational runners possess less product knowledge or differ in their familiarity with the relevant brands, their exclusion could systematically distort the estimated level of confusion. Accordingly, the presence of a inclusion deficiency should be distinguished from its potential consequences for the validity of survey inferences.
The central methodological concern is not simply whether the sampling frame perfectly replicates the defined universe, but whether any discrepancies between the two materially compromise the ability to draw defensible conclusions about confusion within the legally relevant population. Even when the defined universe and the intended sample appear aligned, additional practical considerations remain, including the opacity of panel information, screening by the panel provider, and screening questions within the survey.
Uncertainties Due to Opacity of Panel Information
Online panels offered by reputable commercial vendors have considerably benefited investigators seeking timely, cost-effective responses from specified consumers. Even so, researchers should “check under the hood” to reduce sampling frame error stemming from the methods used to recruit, screen, and retain panelists. This includes understanding both the parameters the researcher supplied to the provider and the methods the provider used to generate the list of potential participants.
Randomization for participant selection may exist within a subsection of the panel, along with random assignment to condition, but the panel as a whole may not be randomly drawn from the target universe. Specifically, procedures used to recruit, screen, and retain respondents can introduce systematic bias (e.g., elevated literacy, heightened attentiveness). Of course, such validation efforts have real merit, providing respondents who are inclined to follow instructions and not rush through the survey. These limitations may be mitigated when the panel is sufficiently large and adequately represents key sociodemographic dimensions. Nonetheless, familiarity with the panel providers’ intake and retention methods is required to be confident in such assumptions.
Screening Questions Used by Panel Company
To develop a preliminary list of potential study participants, panel providers may conduct an initial screen based on existing participant information (e.g., demographics and/or purchase behaviors), applying an additional layer of pre-screening not directly visible in the researcher’s survey instrument. For example, as part of an intake survey involving running shoes, the panel provider may have information on gender and perhaps even running shoes in general. If the provider lacks information on who purchases running shoes, they may run a quick screening survey before admitting participants to the subset of the panel the researcher uses. This spares respondents the frustration of investing effort in a survey only to be dropped, uncompensated, partway through when a running shoe purchase screening question would otherwise have disqualified them.
Screening Questions Within the Survey
Screening questions typically appear at the beginning of a consumer confusion study and can distort Universe Correspondence in two ways. The first is straightforward misspecification, where the screening question internal to the survey itself fails to match the reference universe it is meant to operationalize. For example, if a survey regarding running shoes typically sold for $150, established only that respondents had purchased a running shoe in the preceding 12 months or intended to purchase one in the next 12 months, without screening for willingness to pay $150 or more, the resulting sample would be overinclusive.
The second mechanism occurs when a respondent is motivated to qualify for a survey in order to receive compensation. In such a case, it is conceivable that the person might answer in a manner that they believe will permit continuation (Diamond & Franklyn, 2014). This concern assumes both that a nontrivial number of respondents would misrepresent their eligibility and that they could identify the screening questions and responses needed to qualify. Screening questions are commonplace and often overt in practice (e.g., participants must be over 18, must wear corrective lenses, should not consult others during the study), so absent empirical validation, such concerns may not be well founded.
Implications for Litigation and Expert Evaluation
Evaluating the appropriateness of the population represented in a consumer confusion survey involves two fundamental considerations. The first concerns universe definition, specifically whether the population identified for study corresponds to the consumers whose perceptions are relevant to the alleged infringement. This requires consideration of the alleged nature of confusion, the characteristics of the relevant consumers, and the marketplace circumstances under which confusion is alleged to occur. The second concerns universe operationalization, specifically whether the sampling frame adequately encompasses the defined universe. Even when the universe has been appropriately specified, limitations in the sampling frame may exclude relevant consumer segments or introduce individuals who fall outside the intended population. The significance of these discrepancies depends not simply on their existence, but on their nature, magnitude, and potential relationship to the consumer perceptions being measured.These two considerations should also guide the interpretation and reporting of survey findings. Researchers should distinguish between deficiencies in defining the relevant consumer population and deficiencies in establishing a sampling frame capable of representing that population. Importantly, neither type of deficiency necessarily establishes that estimates of consumer confusion are biased. The inferential consequences depend on whether the resulting sample differs from the legally relevant population in characteristics associated with the likelihood of confusion. Accordingly, conclusions should reflect both the appropriateness of the defined universe and the extent to which the sampling frame supports generalization to that universe.
[1] Weisberg (2005) and other market researchers use the term “coverage” instead of “inclusion” but the meaning within the legal context is the same.