UK Building Society

The Society needed to understand the profile and profitability of customers who took up Home Insurance as well as significant trends. The questions were:

  1. Who takes insurance?
  2. Who drops Home insurance products, and after what time, typically?
  3. Who makes/loses the most money?
  4. Among those customers without insurance, who should be targeted for insurance marketing?

The challenge was not only to provide useful answers to these questions, but also to deliver well-documented technical material, both statistical and SAS® code, to allow the in-house Strategy Section to adapt and develop the analysis further. At the same time, the data to be analysed had to be assembled afresh from existing SAS® datasets, validated, corrected and documented.

The results and benefits

The first result was a completely validated and corrected SAS® dataset combining mortgage and insurance information, two areas seldom brought together in analysis within the society. This was documented and the generating code made available for further updating.

The customer base was segmented with the aid of classical data-analytic techniques such as logistic regression and cluster analysis. These identified features of customers who were most likely to drop insurance and were able to indicate an important trend in customer behaviour as well as portions of the customer base worth approaching with insurance products and cross-sales.

Customer profitability is directly connected to tenure of insurance, so here survival analysis takes over from classical risk assessment. This modern technique, emerging now strongly in risk assessment, estimates a customer’s tenure of house insurance on the basis of imperfect or biased data. The full power of SAS® is required for the analysis, together with a sophisticated appreciation of the ways in which bias and censorship influence the data.

The dataset construction and analysis advanced in close daily consultation with the Insurance and Strategy Sections; the needs of these departments were paramount to the project’s success. Presentation to executive and senior managers prompted enthusiastic interest and positive action. To gain market advantage an immediate decision was made that in future data collected would be used for statistical analysis.