MaxDiff (Best-Worst) Scaling Apps

The MaxDiff task: Respondents are presented with a series of small subsets, typically with 2 to 4 items each, and asked to select the best and worst item in each. Simple!

The problem: MaxDiff presents challenges both in the design of the subsets and in the complexity of the data analysis.

The solution: Separate web-based design and analysis applications have been created by Qi to specifically help researchers overcome these challenges.

MaxDiff Design Application

Since the statistical quality of a design for a MaxDiff study can vary greatly, the application (‘app’) attempts to steer the user towards a good design:

• where each respondent sees each item approximately the same number of times and
• where the number of different versions of the basic design is selected optimally.

The ‘app’ can find designs extremely quickly, and so the user can change the basic parameters and see instantly how this impacts design quality.

Once the details are finalised the design is exported in a format compatible with the analysis application.

Key features

• Web based
• Simple to use
• Statistically ‘good quality’ designs
• Fast to calculate
• Interactive user assessment of design
• Designs exported simply

MaxDiff Analysis Application

The analysis application implements ‘discrete choice modelling’ with Multinomial Conditional Logistic Regression and presents tables and charts with utility values and choice probabilities.

Alongside the logistic analysis, the application also computes more traditional summary statistics, detailing the number of times each item was selected as best and worst, as well as a best-worst score.

To run and analysis 3 data files are required:

• a table with item names
• a table with the design
• a table with the respondent responses

Key features

• Web based
• Simple to use
• Complex stats method made easy
• Utility values and choice probabilities
• Simple summary statistics
• 3 clear data files input structure