Maximum Difference Scaling: Finding preferences made easy


Maximum Difference Scaling (MaxDiff) is a highly efficient method to determine preference differences between a relatively large – theoretically even unlimited – number of similar stimuli or attributes. It works similar to a simple paired comparison.




Questions for which MaxDiff is suitable are, for example, which product features have the highest relevance for the purchase decision, which features are best suited to a brand or a product, or which benefits or claims have the highest significance for communication or advertising.


Questions that can be addressed with the MaxDiff method are, for example, “Which product features have the highest relevance for purchase intention?” or, “Which features have the highest brand or product fit?” or, “Which benefits or claims have the strongest impact in communication or advertisement?”
Usually, the above questions are answered with scaled applicability or rankings (“fits best” etc.). These techniques do have some known weaknesses, however:


Beware of…


Scaled ratings are usually not selective enough (i.e. you get similar values for all stimuli). Moreover, they are prone to unidirectional answer tendencies like, for example, simplification, social desirability and cultural effects (in one country people prefer characteristic differences more than in another), to mention just the most important.


Rankings, are usually limited to a small number of stimuli.


Advantages of MaxDiff


On first view, MaxDiff is based on paired comparisons. However, the comparison is not based on the decision within a single pair of attributes or features (A versus B); instead the decision is based on a greater set of attributes or features (A, B, C, D …).


Respondents are asked to complete several sets of similar tasks, that each contains four different items/stimuli, for example. From this set the respondents needs to choose the best and least liked items, respectively.


This method offers several advantages:

  • Respondents need to make a discrete choice only
  • Theoretically, the number of items that can be researched is unlimited
  • For each item or attribute there are rationally scaled values that facilitate further analysis
  • With the help of multi-nominal logical regression utility values are determined
  • For an easier understanding and interpretation these values can be given in percentage or as indices
  • Moreover, a further analysis based on Hierarchical Bayes is possible
  • Method and analysis of MaxDiff are thus in accordance with the classical Choise-Based Conjoint-Analysis (CBC). As such, MaxDiff can be seen as a simpler version of CBC

Last, but not least this means: It’s a user-friendly questionnaire that’s easy and fun for respondents. At the same time it offers a wide spectrum of possible use and its results are easy to interpret!


MaxDiff should always be considered when it’s important to determine the differences of preference between a relatively big amount of similar stimuli or attributes – costs are low in comparison to the high benefit gained!