Issue

An insurer selling annuities through an online portal wanted to increase their conversion rate. The client requested Crowe to conduct a proof of concept exercise for using machine learning to identify likely sales targets.

Solution

Crowe UK developed a supervised learning model to identify dimensions within the insurer's data that were predictive of the propensity to buy an annuity. 

This went beyond typical measures already used by the sales team including time of day and day of the wek that a quotation was requested, the browser used to access the portal, as well as ruling out dead-ends by using IP address matching to identify quotations that were likely driven by benchmarking exercises conducted by consultancies.

Outcome

The model Crowe developed allowed a more refined targeting of customers who are likely to buy. Of a sample of the 100 customers who were identified as most likely to purchase an annuity from the previous month, the sales team was able to increase sales by 30 times the amount previously seen. 

The modelling has also enabled the insurer to improve the customer experience by pre-empting their requirements and indentifying stages of the quotation process at which customers drop out.