Purpose
It is often necessary to identify the best or most important item from a series of features. For example, a credit card provider may wish to know which ancillary offers or services are most appreciated by the customer. MaxDiff scaling provides the possibility of calculating a complete ranking of all features by means of a simple questioning technique.
Survey method
The respondent is shown a series of screens (or questionnaire pages), each of which shows a choice of the features under investigation. In each case he is asked to select which one he prefers the most, and which one the least. A value is calculated for each feature reflecting its importance in comparison to the others.
Advantages
Fig. 1: Results of a MaxDiff scaling
- Variants of different fields (e.g. prices, brands, image aspects, concrete product characteristics) can be directly compared with one another.
- Very large numbers of variants can be included. Up to 45 such features and levels are practicable within a survey. From the technical point of view, even as many as several hundred could be investigated at the same time.
- Completion of the survey tasks is very easy for the respondent, requiring little intellectual effort.
- Due to the type of questioning (no ratings or rankings, but only best/most vs. worst/least) the results are very robust. There are no cultural influences, such as is usually the case with numerical ratings
- Interpretation of the results is clear and easy. All the resulting values are between 0 and 100. The values sum to 100, so that the individual values can also be interpreted as percentages. The results can therefore be implemented with confidence even with no previous experience of this form of analysis and without extensive training.
- The method delivers individual item values for each respondent. The results can thus be used for the investigation of specific sub-segments of the target group.
- The MaxDiff questions can be implemented with practically all survey media (paper-and-pencil, CAPI, online, and for simple models even CATI).
Disadvantages
- Especially on account of the simplicity of the test structure and results presentation, the method holds the danger of producing non-valid results. The results are delivered at the highest level of measurement, i.e. a ratio scale measurement. In order for these values to yield meaningful interpretation, the content of the stimulus material must fulfill correspondingly demanding requirements.
- For many types of survey, this method cannot replace conjoint analysis. Where it is necessary to compare products as whole with one another, and to provide a basis for e.g. simulation calculations, a conjoint methodology must be used.
- There are aspects that cannot be meaningfully compared using MaxDiff questions. For example it makes little sense to ask respondents to compare price levels. Normally the lowest price will always be selected, so that the amount of useful information obtained is practically zero. It is possible to avoid this by imposing restrictions on the survey conditions, but only at the cost of reducing the statistical reliability of the results.
