Practitioners’ Guide to Building Actuarial Reserving Workflows Using Chain-Ladder Python

Practicing actuaries are managing more complex system and workflows every day, especially in reserving. While reserving in of itself is a complex topic, many actuaries prefer routine workflows that are intuitive and easy to manage.

This paper discusses practical considerations on the analytical and workflow structure of setting an unpaid claim estimate. Literature abounds on actuarial methods and assumptions used to tackle specific reserving issues, but there is very little research into the best practices of managing actuarial reserving workflows. Management teams of companies increasingly desire

1. Faster speed in end-to-end analysis,
2. Lower cost in running a reserving shop, and
3. More detailed insight into the drivers of loss experience

These pressures continue to shape the evolution of reserving against a backdrop of technology advancements that are outpacing advancements in actuarial reserving processes.

To address these pressures, we propose a standard for data management of the loss triangle along with a standard for building actuarial models with an explicit declaration of model assumptions. This standard is manifested in the open-source package called chainladder-python, and we will demonstrate its use in core reserving as well as its integration into actuarial departments’ workflows from a practical perspective