MaxDiff, also known as best-worst scaling, is a quantitative tool providing prioritisation of needs, attributes or messages. It is easy to design, easy for respondents to use and provides clients with an easy to interpret hierarchy of items. In this quick guide we outline what MaxDiff is and provide some tips for making the most out of this powerful methodology.
What can it be used for?
MaxDiff is most commonly used for message testing, needs assessment, brand preference and product feature preference/importance.
So, how does it work?
Respondents are shown a set of items/messages and asked to indicate which they consider to be the most important and which is the least important (or best and worst). The task is then repeated with different combinations of items/messages, forcing the respondent to make new trade-offs with each set.
The resulting scores are easy to interpret as they are typically placed on a 0 to 100 point scale, summing to 100. Essentially, if we gave one person 100 points how would they distribute these across the attributes, with the most points attributed to the best/ most important item.
Why is it better than standard Likert rating scales?
Unlike stated choice or Likert rating scales the MaxDiff method provides a clear and robust quantitative prioritization of needs/attributes/messages. The output shows not just the appeal of each individual item but how they interact with each other providing a hierarchy of items.
Any tips for making the most of the methodology?
- Include a pre-phase to conduct secondary research and a small number of qualitative interviews to hypothesis test the items to include and check that the language resonates as it should
- Keep it simple: keep the number of attributes/needs/messages to no more than 25. This will ensure that respondents are not fatigued by doing too many repetitive trade-off tasks
- Avoid having multiple versions: If you can design your MaxDiff to show all items to each participant, this will improve the robustness of the data and make segmentation easier.