Trimming the Fat from Unprofitable Customers

​

A German insurance company serving the Southeast Asian market sought for PulseMetrics’ help to improve its existing customer segmentation strategy in 2015. Segmenting customers according to highest profitability or highest loss did not provide sufficient information. Our client therefore wanted to deepen its customer knowledge and profiling by also taking into consideration the number of insurance claims and losses incurred by each customer. To fully and fairly appraise customers’ risk and value, our consultants set out to build predictive and advanced segmentation models.

Drilling Deep into a Customers’ Profile

​

To start with, we designed a heuristic scoring model to predict each customers’ frequency of insurance claims and amount of losses incurred based on existing information on claims and losses for various scenarios such as whether the policyholder was at fault. This made it possible to tell if a customer would likely incur a sizeable loss or not in the future. Thereafter, we ranked and classified customers into three key segments using a decision tree: the ‘good’, ‘normal’, and ‘bad’.

Regression analysis was also done to predict the loss for each customer segment.

To understand the behaviour and attitudes of customers in each of these segments and what differentiated them, we cross-referenced their distribution with enriched demographic and vehicle attributes such as age, gender, occupation, region, vehicle type, vehicle size, and channel type. In this sense, we could discern which vehicle type attracted the ‘good’ customers and the ‘bad’ ones for example.

​

Knocking Three Birds with One Stone

​

More concretely, our results enabled our client to adopt the following three strategies. Firstly, the insurance company tweaked the pricing of its insurance products for various customer segments. Secondly, it also adjusted its sales strategy to acquire more ‘good’ customers, after we revealed that direct sales brought in the largest proportion of ‘bad’ customers compared to brokers and agencies. Finally, it also became more strategic with its partnerships with specific brokers that brought in a higher volume of ‘good’ customers. In the effort to be data-driven in managing risk, our client was therefore able to hack its customer journey—targeting and converting lower-risk customers, thereby minimising insurance claims losses.