The authors develop a survey-based approach to determine the extent to which brands in a pre-specified product class compete with each other. The approach, called the prefer- ence structure measurement (PSM) procedure, uses a household survey to identify clusters of user/usage situations within a household and elicit preferences for brands and price- preference trade-offs for each cluster. The procedure specifies preference functions and estimates choice probabilities for each cluster within each household in the survey. Market level competition is modeled by aggregating over the survey sample. The authors illustrate their approach using the grocery coffee product category. A key output of the PSM approach is a cross-price elasticity matrix that can be used to assess item-by-item substitutability and potentially asymmetric brand strength. The procedure avoids many of the problems associated with estimating cross elasticities with secondary data. The PSM procedure also yields survey-based brand-switching matrices at both the cluster and household levels. A comparison of the two switching matrices for coffee reveals somewhat greater switching among types (e.g., instant and ground) at the household level than at the cluster level although the overall difference is small. A latent class decomposition of the cluster-level switching matrix for coffee reveals a product type-based structure with overlap across types and the presence of some brand name-based structure. The survey-based results were also found to be predictive of the actual purchase behavior of 2000 IRI scanner panel households residing in the same area, but who were not included in the survey sample.