Statistical Consulting

Choice modelling

Case study

Choice modelling, sometimes called “conjoint analysis”, is a scientific method used to model real-world decision maker purchasing behaviour and predict an individual’s probability of choosing among different outcomes.

The applications of this technique span across industries, from healthcare to transportation to retail, which means the choice outcomes can range in complexity, too. Choice modelling research can measure decision maker preferences for simple things, such as marketing messages, or more complex scenarios around services, product features or complete products described by their underlying features.

The survey data used in choice modelling is collected using a unique statistical survey design, allowing us to determine what drives the choice.

The main type of choice models are:

  • Maximum Difference Scaling (MaxDiff), when the trade-off is between simple options, such as additional services, marketing messages, or benefits. It is used for prioritising simple options in a discriminating, scale-free way.


  • Best-Worst Case 2 (BWC2) is halfway between a MaxDiff and a DCM. The trade-off is between specific product or service feature levels rather than simple options (e.g., particular specifications for different features). It is used when the main goal is to understand which features and feature levels are most influential in driving choices.


  • Discrete-choice modelling (DCM) is when complete concepts (usually products or services) are traded off against each other. This is a classic Choice-Based Conjoint. From the underlying design, we can infer which features and feature levels drive choice. It is the preferred method when one of the goals is to simulate market share for new products for hypothetical product profiles and specific competitor scenarios.


Our approach ensures there is always a lot of flexibility in how these models are adapted to account for current competitors, different contexts (e.g. patient types for medical studies) and the need to calibrate predictions using a ‘likelihood of buying’ question. And we’re always happy to customise where necessary.

All models produce a needs-based segmentation as part of the output, enabling subgroup differences to be reflected.

As part of our service we deliver smart, interactive, user friendly simulators for all of our choice models

Feels like we are a good match?
Let’s talk and see how we can help you.


Contact us