Max-diff (also called best worst scaling) is a discrete choice model used to measure importance or preference in a list of several items (up to 30) such as service options, product features, marketing slogans, logos, etc. Like conjoint analysis, it is based on comparative judgments and not numerical ratings which allows for greater discrimination between the items and the ability to scale based on importance.
Respondents are presented with several screens and asked to pick the most important and least important item on each screen:
After selecting the most important and least important attribute, another scenario is presented with a similar mix of attributes:
Followed by another scenario:
The number of scenarios presented depends in part on the number of attributes being measured, but several combinations are typically presented.
Responses are typically analyzed using Hierarchical Bayes techniques to calculate importance scores.
One of the advantages of Max-diff scaling is that people are typically better at judging extremes as opposed to rating several items on scale. Additionally advantages of Max-diff are:
It produces better differentiation than ranking or scaling does
It is more discriminating
It is a more sensitive measure than ranking or scaling
It eliminates straight-lining by respondents
It is a better predictor of actual behavior
It is simple