Bestworst

Bestworst, commonly written as best-worst or max-diff, is a survey technique used to measure relative preferences among a set of items, attributes, or features. In a bestworst design, respondents are shown a subset of items and asked to indicate which item they consider the best and which they consider the worst within that subset. This paired judgment is repeated across many blocks to cover a larger universe of items.

The method provides more discrimination than single-rating scales because each question yields both a positive and a negative data point. Analysts convert the collected best and worst selections into scores or probabilities that reflect the relative importance or appeal of each item. Common approaches include simple win–loss counts and more formal max-diff models, such as hierarchical Bayes or multinomial logit frameworks, which estimate latent utilities for items.

Bestworst is widely used in market research, product development, brand attribute prioritization, and public policy to streamline decision making when dozens or hundreds of attributes must be evaluated. It is efficient for handling large item sets because each respondent contributes information about multiple items without requiring universal ratings on a long list.

Limitations include cognitive burden on respondents, especially with very large attribute sets, and the need for careful experimental design to avoid bias. Results can also depend on the chosen modeling approach, so transparency in methods is important for interpretation and comparability. See also best-worst scaling, max-diff, and conjoint analysis.