Identifying commercial potential with segmentation in insurance

Context & Objectives

A leading European insurance company operates in a niche market where price margins are shallow. The insurer possesses a vast sales network, equally divided between direct sales and a brokerage network (+/- 4,500 brokers). They needed to know their brokers better and the intensity of their relationship with the end-client.

With this in mind, several "inspectors" regularly visited the various brokers and stimulate the relationship with the end-client. However, the insurer encountered two significant difficulties. First, inspectors made visits according to the brokers' availability, which leads to suboptimal use of resources. This made it very complicated for inspectors to maximize their number of visits over the year.

Second, it wasn’t always possible for the inspector visits to account for the actual sales potential of the brokers. Therefore, it was likely that their time (already limited) was spent with brokers with low/medium potential while ignoring larger brokers with high potential. Accurate segmentation of the sales network was crucial for the insurer to better allocate commercial resources.

Approach

To prioritize the inspectors' visits with a segmentation of the existing broker base, we worked through these steps:

First, Data Consolidation comprises data cleaning work and data linking with OPENDATA (Statbel and BCE/BNB). Data featuring then uncovered the median/average turnover a broker carries out annually and the ratio between the turnover of a broker at the insurer over its annual turnover. To round out the Consolidation, we prepared the data for segmentation.

Second, we performed a Data Analysis with the TIMi data interpretation tool to present data to the insurer in a clear, practical way.

Finally, we ran two segmentation models to identify active and inactive agents through cluster analysis, detect the most critical opportunities, and find the optimum number of clusters for the insurer. We assembled this information into easy-to-comprehend visualizations.

We delivered the project's output through a Powerpoint (to offer key data insights and visualizations), Excel Files (each covering broker segments of differing potential), Python Code, a TIMi Data Audit, and further explanatory documentation.

Results

At first sight, the insurer, thanks to the segmentation and data analysis, now can better interpret their brokerage network on a province-by-province basis. These insights led to a reorganization of the sales department to better coordinate geographic commercial efforts.

In total, for each segmentation model, we discovered four segments:

• Two highly location-targeted, requiring immediate attention given their potential during the next inspection.

• Two are divided into low potential and well-established brokers. The latter underlines the importance of brokers already doing business and showing high potential, for the insurer to retain in the future.

After completing the project in 8 days, the insurer quickly gained a clear vision of its subsequent actions to increase its commercial potential. In total, about 200 brokers have been identified as high potential or underdeveloped.