Projective techniques have their origins in personality research and have been utilized in marketing and consumer research for over fifty years. They are used to elicit deeper understanding of a brand, service, or other business entity. Projective techniques typically utilize questions that link the concept being studied to specific tasks, objects, or other people. For example, for a retailer, one may ask ``This retailer remains you of which animal?'' or ``If the retailer is a relative then which relative would this be?'' Projective techniques come under the banner of qualitative research, but the data can be quantified and analyzed using quantitative methods such as multidimensional scaling (MDS) and cluster analysis.

This talk looks at methods for exploring structure using projective techniques. A specific focus is given on how to optimally pre-process the discretized answer information. This can be structured using a bag-of-words approach, but with very uneven density. For example, when comparing a brand to an animal, some animals, such as lions, are used very often, which other animals, for example, walruses, are rarely used. This talk will describe methods of optimally weighting observations to maximize the information obtained relative to error and to improve the accuracy of brand mapping visualizations and clusterings . An example will show how a range of well known US retail brands are viewed and will give insights to consumer perceptions of brands relative to projective traits, including similarities with animals, cars, music artists, and American cities. An extension method is described where the attributes of the projective answers are crowdsourced from the web and used to improve the accuracy of the analysis.