Working with models is the quintessential activity for most actuaries. From projecting reserves to pricing, almost every core activity routinely performed by actuaries involves modeling balances and analyzing the results and outputs of models. Indeed, models provide the framework for analysis, allowing actuaries to better understand uncertain future events and, therefore, address the business needs of the organization.
Typical actuarial modeling environments incorporate many technical components beyond the calculation engines that perform projections. Such components transform data, store assumptions or build reports. Collectively, they work together to enable actuaries to do their work — such as producing and reporting actuarial reserves. The effectiveness of the actuarial function depends, to a large extent, on how those components are linked, or coupled, within the actuarial environment. For each specific actuarial task, the actuaries or their resident systems architect must decide how the component completing that task integrates with the other components. The actuarial IT architecture can be designed to be loosely coupled or tightly coupled.
Coupling considerations: loose vs. tight
The term “coupling” refers to the degree to which the components or applications depend on each other within an actuarial environment. To give an analogy, consider the cameras that are standard features in today’s smartphones. The camera is tightly coupled with the smartphone. If you want a better camera or need to fix it, you must buy a new phone. While many people have given up separate cameras for everyday use, professional photographers and passionate hobbyists still carry high-powered cameras with an assortment of professional accessories.
Similarly, many stereo systems include receiver, preamp, amplifier, input devices and even speakers all contained within one box. The audiophiles among us shudder at the thought and prefer our components that stack to the ceiling allowing us to achieve a custom sound with loosely coupled or decoupled components. For some, solid state audio components are not acceptable because they do not afford the user the flexibility of further tuning the sound to their liking as one can do with high-quality tube-based equipment where vacuum tubes within a component can be decoupled in order to achieve additional sound refinement.
With those pictures in mind, let’s return to the actuarial environment and the implications of coupling on architecture, design and the relationships between components. First, let’s consider several areas in the actuarial process that can be a challenge to an effective architectural design:
- Where will assumptions be stored — in an assumption repository (which feeds models) or directly inside the models?
- How will data transformation be handled? Will it take place before or after data is fed into the actuarial modeling environment? Who will own data cleansing and transformation — actuarial or IT groups?
- How will change management be performed? Does the modeling software offer change controls, or will separate controls be necessary?
- How will model runs be automated? Is job scheduling of the model run sufficient, or will end-to-end automation of the process from data collection to reporting be required?
- How often are new reports or changes to existing reports needed, and how responsive does the system need to be that produces these changes?
While there is no one-size-fits-all answer to the question of which is the right architectural approach, there are multiple factors that can considered to decide how tightly or loosely coupled the integration of components within an actuarial environment should be.
- The need for periodic changes to actuarial modeling tool is an important factor to consider when deciding on a tightly or loosely coupled architecture. A loosely coupled architecture helps an organization make changes with minimal impact on other components (assumption repository, scenario generator, etc.) within the actuarial environment. This modular approach decreases time to market and reduces cost and time in making the changes to the modeling software since some of the externalized components can be reused within the environment.
- The degree to which the actuarial organization relies on IT resources must also be taken into account. Actuarial shops that limit the involvement of IT resources in the actuarial process may prefer a tightly coupled architecture that can be installed and configured and operated by actuaries who have responsibility for the whole process end to end. A loosely coupled architecture requires a deeper set of IT skills to design and build and, ideally, returns this investment by allowing for more involvement of IT in selected sub-functions such as data preparation and report design and development.
The bottom line
While loose coupling vs. tight coupling of actuarial modeling components is only one of many factors to be considered in the design of an effective architecture, it is a central factor with influence on several critical goals. While improving or designing a new actuarial modeling environment, consideration of the appropriate interaction between key components and how those components are used is necessary in adopting an architecture that best serves the customers of the actuarial models. The key to designing the optimal actuarial modeling environment is to balance the big-picture, long-term strategic objectives with the many technical details that can lead to improvements in this quintessential actuarial activity.
To read this full article on EY’s Insurance website, please
