Preference measurement: Maximum Difference Scaling

Appinio Market Research

Introduction

Which product features are the most important to my customers? What services should I offer to satisfy my target group? Which claim best fits my brand? Questions like these determine the daily life of every marketer and decision-maker in a wide variety of companies. The fact is: not every seemingly important product feature or service can end up becoming part of a product.

Therefore, it is important to prioritise features and services. One method for this is the so-called Maximum Difference Scaling (MaxDiff), also known as "Best Worst Scaling". In this process, respondents are given different sets of product attributes from which they are to select the best and worst alternative in their view. Based on the results, concrete preference or importance rankings can be derived.

This ranking looks like this, for example:

How does a MaxDiff scaling work?

In order to carry out a MaxDiff scaling, a list of items is first created (e.g. characteristics of a product), the importance of which is to be determined from the customer's point of view.

These items are divided into randomly combined sets with 4-6 alternatives each, whereby each item is always combined equally often with every other item. For each set, respondents must then select the best or worst alternative from their point of view.

Example:

The difference between the best and the worst alternative is the "maximum difference", which is reflected in the term "MaxDiff".

Based on the respondents’ "MaxDiff scores", a preference or importance ranking can be created at the end. The ranking is based on an overall score, which is calculated from the difference between the proportionate positive and negative evaluations and can lie between -100% and +100%.

Advantages and disadvantages of MaxDiff scaling

Ranging from product features, to services, claims or associations, a MaxDiff scaling is possible in many areas and is therefore very versatile. Its main advantage is that a so-called claim inflation of the respondent can be avoided. This means that all attributes are considered important by the respondent and how important they are perceived in relation to each other cannot be determined. This happens quickly when respondents are given a list of attributes. For example, according to the respondents, vegetables should be particularly cheap, organic, tasty and presented in a smart design.

In a MaxDiff scaling, on the other hand, exactly this inflation of expectations does not happen, since it must be prioritised very precisely how important one attribute is in comparison to another. In a MaxDiff scaling, for example, it can be determined whether a favourable price or rather organic quality is a main argument for buying food and to what extent different target groups differ in this respect.

Further Advantages:

Surveys for MaxDiff Scaling are very simple and therefore very efficient. Each respondent is given only four items per question, from which he or she must choose the best or worst.
Especially compared to long lists, this makes a survey seem much easier for respondents.

Different attributes/characteristics can be directly contrasted and compared on the basis of their importance.
The simple presentation of the results facilitates analysis. Percentile rankings allow for an intuitive interpretation of the data.

Disadvantages:

MaxDiff Scaling only provides information about the relevance of a particular attribute (e.g. packaging design in general) but not about different variations of it (e.g. different alternatives of a packaging design). For this, a systematic testing of different variations of an attribute (e.g. several design alternatives) in the context of a monadic or semi-monadic survey is rather recommended.
For some aspects it is difficult to find out preferences on the basis of a MaxDiff scaling. For example, it is often difficult to use for pricing models, as the smallest price is usually preferred.
The percentages that can be found using this method indicate how important an attribute is compared to all other attributes, but not how they are rated in detail.

Conclusion

MaxDiff scaling is a form of preference measurement that shows which attributes (characteristics, services) respondents consider particularly important (good) or unimportant (bad). In doing so, they simply have to decide in different sets of two to four attributes which they think is most important/best and which is least important/worst.
The results are always in the form of a percentage value and show the relative importance of one attribute compared to others.

Accordingly, the data can be easily and intuitively interpreted and used for further purposes.