Using MaxDiff Scaling for Message Bundle Optimization in Presence of Interaction Effects

• Activity Number: 180
• Type: Contributed
• Date/Time: Monday, August 4, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Marketing
• Author(s): Faina Shmulyian*+ and Dimitri Liakhovitski
• Companies: Markettools and GfK
• Keywords: MaxDiff ; Message Optimization ; Bundle Optimization ; interactions

Abstract

Maximum Difference Scaling (MaxDiff) is widely used for message optimization. It eliminates potential scale usage effects; its scores discriminate well between messages and between respondents. MaxDiff is efficient in differentiating among individual messages but what if one needs to determine the best combination, or "bundle", of messages? Some messages might resonate with each other; some might not work well together. Traditional MaxDiff does not test for this kind of interaction. Therefore, it is incorrect to conclude that a bundle of messages with the highest individual MaxDiff scores is always the optimal one. However, this is frequently done in market research practice. This simulation study directly models interaction for each pair of messages and each respondent. Based on "true" individual message utilities and interaction utilities, we simulate responses in a MaxDiff exercise and find winning message bundles. We systematically vary the size and direction of pairwise interactions and demonstrate how, as a result, the MaxDiff-based winning bundle changes; we further suggest possible extensions to MaxDiff to optimize message bundles.