While there are numerous use cases of successful isolated implementations of Machine Learning applications (e.g. digital sales and policy issuing or automated claim handling), a holistic approach often lacks and more often than not insurers encounter issues at the back-office.
In this article we give our view on capturing the value of the data by transforming the actuarial modelling landscape and by making use of new concepts and techniques to connect the dots bottom-up. We introduce the concept of Not Incurred, Not Reported (NINR) reserve to link the models in the cycle and present three use cases to illustrate the added value of machine learning in our framework.
Introduction
We define the actuarial cycle as the link of the central actuarial activities of insurance: pricing, reserving and risk & capital management. These activities surround the core of the insurer’s business (see figure 1). In practice little communication exists between the models that are used in each of the individual elements. Models too often make use of different data sources, granularity, modelling techniques and frequency of updating. Therefore their output is difficult to align and to compare. This results in sub-optimal pricing,
inaccurate reserving, and imperfect risk quantification. All these lead to
ineffective steering of the business.
Read the full article under Download.