Maximum Difference Scaling
Maximum Difference Scaling is a very effective method of establishing the relative priority attached by an audience to a large set of items (up to 30). These items might be:
- Potential marketing messages for a new product
- Features or benefits of a service
- Areas for potential investment of resources
- Products or Services used
- Interests and activities
- Future needs
Traditionally, for a large number of items (more than 10) this might have been addressed via a rating scale. For example, we might ask on a numeric scale of 1..10 where 1 is not important at all and 10 is Extremely important, how important is each item? Each item's average score can then be presented for any subgroup.
The main problem with this approach for a large number of items is that differentiation between items on this type of scale is typically poor. This is due to the following:
- Respondent fatigue – the respondent tends to give similar ratings to all items without giving much thought to individual items, just to get through the questionnaire
- In situations where many item are considered important then a large number can receive very similar ratings at the top end of the scale
- Items are not traded-off against each other and therefore many items that not at the extreme ends of the scale or that are considered similarly important are given a similar rating
- The use of the scale can be quite arbitrary for different individuals. In international studies there are also cultural biases where some countries prefer to use the top-end or the mid point more than the others
This is undesirable. Typically we are looking for high differentiation across the items and also high differentiation across different subgroups of the population. High differentiation across subgroups can be an extremely useful marketing tool, allowing the needs of distinct groups to be understood and targeted.
Maximum difference scaling meets this requirement. It allows a large number of items can be traded off against each other in an efficient manner, which is independent of any rating scale bias. As well as placing the items on a highly differentiating scale, the technique also produces a needs based segmentation, allowing priorities to be estimated for any subgroup.
Questions can be framed in a similar way to that shown on the slide below:
The items presented are chosen using a statistically optimal design which gains the maximum amount of information on item trade-offs using the minimum number of questions. These questions can be administered using any methodology. If telephone interviewing is to be used, then it is more practical to trade off no more than three items at a time.
The analysis uses a form of Conjoint modelling to assign utility values to each item. These can be scaled so that the items are assigned a relative importance which sums to 100% and/or indexed using some average importance score as zero as shown below.
This method has been shown to produce much stronger differentiation across both items and respondents.
Another benefit is that we can produce a scale for any sub-set of the total items which is much more realistic than that produced from a rating scale. For example, if we removed item 10, we might find that items 4 and 7 are promoted to higher importance, while the others remain largely the same.
This is more like real life. For example if a set of choices of TV channels included CNN, Sky News and BBC New 24 plus a whole range of entertainment channels, removing BBC News 24 (the preferred news channel) from the mix would mainly promote the importance of CNN and Sky News, if the audience were keen on news and current affairs programmes. In other words, all remaining channels do not benefit equally.