Maximum Difference Scaling, also known as Best-Worst Scaling, is an approach for understanding the preference and importance scores allowing researchers to analyze a higher number of items generating discriminating results as respondents are asked to choose the ‘Best’ and ‘Worst’ option which simulates real-world behavior. Max diff is a powerful tool used by Hamilton to further understand and identify which attributes in a product /service / offer are most important.
- Max Diff always generates discriminating results as respondents are asked to choose the BEST and WORST option which simulates real situations (in the real life people make choices and trade-offs no ordering or ranking, for example, on a purchase in a supermarket).
- Max Diff is a simple method for all the targets involved in the project: researchers, end user and respondents. The question is simple to understand, so respondents from children to adults with a variety of educational and cultural backgrounds can provide reliable data less monotonically. For researchers and end users is easy to use and applicable to a large variety of projects and market research situations.
- Since respondents make choices rather than expressing strength of preference using some numeric scale, there is no problems of scale use bias, so cultural differences are absent in the Max Diff scales. Comparisons between items are referenced against other attributes tested, rather than pre-defined points of a scale.
- In Max Diff scales more items can be included due to the question is simple to perform and understand providing to the analysts a preference value for each attribute reflecting its relative importance in comparison to others.
At a methodological level, the respondents see a list of items and they are asked to determine from that list what is the most important to them and what is the least important. The items are not shown all at one time. The technical teams determine how many items must be shown and how many sets of these items each person has to go through in order to move to next question.
Example of Max Diff Scale:
- Count Analysis: the simplest alternative, tallying of the number of times each item is chosen as ‘Best or ‘Worst’ important by respondents. A simple form of summarizing MaxDiff scores combines the two measures: percent of times each attribute has been selected as BEST less the percent of times each item has been selected as WORST.
- Logit Model: a more complex but fast alternative, using a Logit model to obtain the importance value of each attribute in percent-shared utility scale.
- Hierarchical Bayes or Latent Class: a more advanced statistical technique that provides respondent-level utilities and can be used in simulators or segments of respondents with similar needs / preferences.
- Brand preferences: to identify a brand market position, relative to its competitors.
- Advertising: to identify which messages are most preferred by key targets.
- Concept and / or product testing: to determine which variety of products has the greatest potential for success.
- Customer satisfaction: to identify the key strengths and enhancement opportunities to improve quality index.
- Needs-based studies: to determine which attributes are critical vs. those consumers are willing to sacrifice.