Behavioral Data for Churn Calculations

In many “set and forget” markets, such as telecommunications, a user must first decide to switch and then make a choice. A choice-based conjoint question, by contrast, forces the person to make the choice. The result is that there is inevitably a poor congruence between the real-world choices and a standard choice-based conjoint model.

One alternative is to add a 'none of these' or use a dual-response format. However, this only makes the problem a little less serious, rather than solving it.

A more pragmatic solution is to:

• Collect data for each person on their historic propensity to switch.
• Use a 'none of these' and then at analysis time, calibrate the data based on their historic propensity.

Model the overall churn rate using historical data.