There is an assumption in US trademark law that the protection of consumer interests—a traditional normative pillar of trademark law--is best achieved by enjoining use by a defendant of a mark that creates a likelihood of confusion (with the plaintiff’s mark) for 15% or more (sometimes less) of relevant consumers. Courts often use survey evidence to determined existence of the likelihood of confusion. This article argues that the interests of all consumers are relevant in that determination. This means that determining the costs, if any, imposed on nonconfused consumers should also be part of the equation. This can be accomplished in part by making better use of verbatim answers and by taking a deeper look into the information conveyed by the survey. Likelihood of confusion analyses should be both quantitative and qualitative.