AI-Enabled Actuarial - Customer 360 Applications

Background: The increasing use of data analytics, artificial intelligence and machine learning in actuarial work has opened a wide range of “actuarial data science” use cases. In the context of insurance, Customer-360 is a suite of analytical and actuarial models used to score customers based on complexity, competitiveness, favourability, risk, and customer lifetime value. Customer-360 has tremendous potential to revamp pricing, ratemaking, reserving and actuarial modelling.

Current State: Currently, actuaries typically perform customer segmentation via predictive models for loss cost, loss ratio, frequency, and/or severity. However, the concept of Customer-360 has typically been confined to marketing exercises, limiting actuarial involvement in modelling broader customer behaviors like price sensitivity, sales-funnel conversion, cross-selling and retention. The increasing adoption of AI-ML techniques in actuarial work has now massively increased the breadth of actuarial insights available from Customer-360 datasets.

Business Need: There is tremendous need for a holistic, actuarially sound views of customer segmentation. Actuarial and underwriting teams typically focus on risk-based aspects of customer segmentation, while marketing and distribution focus on competitiveness, cross-selling, retention, and customer lifetime value. Customer-360 models help bring these varying dimensions of customer behavior under a single, comprehensive solution that provides a unified view of customer segmentation to inform strategic decision making. For example, we developed AI-enabled actuarial prospecting models that create interactions between quote-bind models, Bayesian-risk models, and customer lifetime value models, to identify actuarially favourable new business segments.

Solution: The paper illustrates how a combination of GLMs, clustering, ensemble models, and neural networks can be used in conjunction with actuarial techniques to develop customer-360 actuarial insights, and thereby inform actuarial modelling. This paper also demonstrates the real-life application of Customer-360 models, leveraging a multi-line insurance dataset to build complexity, favourability, risk, and customer lifetime value models. Model results are demonstrated by scoring a massive, anonymized customer database. The paper then proposes a way forward to integrate these models into actuarial work (pricing, reserving, capital, etc.). The concept introduced in this paper has extensive applications for actuarial modelling, ratemaking, reserving, pricing, underwriting, capital modelling distribution, sales and marketing across non-life, life, and health insurers.

Find the Q&A here: Q&A on 'Artificial Intelligence and Machine Learning'