Preferences aren’t always tangible, especially when they are asked directly. Not only can preferences be elusive, but they can be difficult to interpret without a standard method of comparison.
A MaxDiff experiment solves many of the problems inherent to traditional methods of asking consumer preferences by forcing a series of real-life tradeoffs. At Gradient, we harness the combined power of MaxDiffs and Bayesian statistics to produce not only a rank order of consumer preferences, but also the magnitude of preferences
A Max(imum) Diff(erences) analysis assesses a survey respondent’s preferences without directly asking them. Instead of ranking a long list of items which can be cumbersome and cognitively taxing, a MaxDiff has respondents choose the “best” and “worst” options out of a given set. Respondents are typically asked to designate the “best” and “worst” options across 10 choice tasks. For each choice task, a random subset of the total items are presented to reduce the cognitive burden on respondents.
Using Bayesian statistical analysis, we are able to calculate the probability of each item in a given set being chosen as the “best” in the list. Our output also allows us to pinpoint exactly how much one item is likely to be preferred over another.
The best part about a MaxDiff is its flexibility. Preferences can be rated in a number of ways, and are not limited to choosing the “best” and “worst” option. For example, if we wanted to know which qualities of a politician are most important when casting a vote, respondents would choose the “most important” and “least important” qualities in each choice task.
See how we’ve identified consumer preferences in the past.
The world's most forward-looking organizations trust Gradient