Comparing association grids and 'pick any' lists for measuring brand attributes

Using a split-ballot experiment with 940 respondents, this study compares the quality of data from an association grid with data gathered through a single 'pick any' list repeated for each brand on a new page in a web survey. The association grid is a multiple response matrix used to measure brand image associations for a number of brands at the same time. Attributes are usually presented as rows, and brands in columns, allowing respondents to select each association they perceive to be true (e.g. Coca Cola - Popular). Our results indicate that larger association grids are answered considerably faster, but are heavily prone to evasion bias and perform worse when it comes to drop-out, comprehension and attention to the task. Smaller association grids have no ill effect on the respondent experience, but are also devoid of material benefit in terms of field time or data quality. As a tool for measuring brand-image association, the association grid is therefore not recommended.