Predicting preference share between options,
features and products
Choice modelling is used to estimate the probability that an individual will choose a particular option from a set of alternative options.
This type of modelling is sometimes referred to as Conjoint Analysis or Choice-Based Conjoint (CBC).
Predicted probabilities can be averaged over a sample, to determine the “share of preference” for each available option under a fixed set of scenarios. Typical examples are:
- To forecast preference share for a set of products, each with its own uniquely configurable set of attributes and attribute levels. A “no-buy” option is optional on these.
- To forecast preferences for particular feature levels over other feature levels (for example, in a new medical treatment, average 10 year’s progression free survival vs only 5% chance of serious side effects).
- To get a sensitive trade off of a large number of fixed options in terms of relative preference.
- To forecast the likelihood of buying a product, given a particular product profile.
For all of these methods, data is collected using an experimental design which ensures we get multiple observations for different scenarios and enough information to determine what is driving choice. We are able to tailor the models to subgroup level by relaxing the constraint that the same choice process applies to everybody.
We are well known for providing very flexible discrete choice models which are tailored for a specific problem rather than just providing an off-the-shelf solution. Our portfolio includes the full range of Discrete Choice Models from simple Maximum Difference Scaling (Max-Diff) trade-offs, through Best Worst Feature level trade-off (Special Case of Max-Diff, sometimes known as Best-Worst Case 2) though to more complex multi-attribute problems with alternative specific designs, competitors and dual response question format. We are also constantly working on improving our models and developing new, innovative applications.
We deliver high quality interactive tools in excel which summarise the models and enable simulations to be run for different what-if scenarios.
We worked on a Max-Diff Analysis for a device used in a sensitive medical procedure…
We analysed the Best Worst Feature Level trade-off for a Digital TV provider to understand…
We have looked at a large number of Multi-attribute discrete choice (CBC) models trading off…
By evaluating the Best-Worst Feature trade-off among physicians, we were able to analyse the type…
By examining a Max-Diff Analysis among residents of a country council, we were able to…
We have undertaken numerous multi-attribute Conjoint studies for telephone and broadband providers. These determine which…
- From trading off simple options (MaxDiff) through to complex Multi-attribute choice sets
- A flexible and customised approach
- Pragmatic advice and consultancy throughout project
- Robust statistical models tailored to subgroup level and with need based segments
- State-of-the art beautifully presented simulators