Selecting Actuarial Model Platform


Introduction


Choosing the right platform for actuarial models is extremely important as it can cost an insurance company a fortune if the platform is not suitable. Model users should consider the initial costs, maintenance costs, governance, flexibility and extensibility when they evaluate a platform.

Companies will have different preferences in selecting a model platform, based on their size and the business they are in. The usages of a model also affect the platform selection.

Initial Costs


The license fee may be just a fraction of the total costs of setting up the model. A comparison of only the license fee may not tell much. We also need to consider what features are included in the platform and what services the software vendor can provide. The table below includes the benefits of built-in features and typically included services for ALM model building.

Feature and Service Benefit
Consultant and training services Shorten learning curve, avoid errors and redundancies in model building.
Built-in modules for insurance products The more modules it has, the less work for model building. Model users are mainly paying for the built-in modules.
Integrated accounting reports Accounting reports consistent with the accounting framework that the platform provides.
Coverages of exotic assets Directly model exotic assets - less supports needed from other platforms, i.e. externally projected assets.
Supports for software installation and updates More self-sufficiency, fewer IT requests, easier to test the new release.
Separated user defined setting If user defined settings are stored in tables, plug-in functions are supported, the model building works from various modelers can be merged easily.
Debugger Efficient troubleshooting.
User manual A good user manual makes modelers pick up the software quickly.

Maintenance Costs of the Model


The maintenance costs can be split into the costs for production and for model updates. The production costs are mainly the machine costs, while the costs for model updates are largely associated with human resources. The maintenance costs may be reduced if the platform has the following characteristics.

Characteristics Cost-Effectiveness
Model runs fast, i.e. shorter per simulation run time. Substantial
Various dynamic periods in one projection, i.e. quarterly followed by annually. Substantial
Distribution of jobs, i.e. distribute stochastic runs by simulation to different machines. Moderate
Isolated system and library updates, i.e. updates don’t impact user defined settings. Substantial
Input Data transformation, i.e. can take Database or CSV File. Moderate
Customizable output format, i.e. accounting report with various granularities. Moderate
A database repository for results that can quickly generate any reports without having to rerun the models. Moderate

Model Governance


A closed system is easier to control than an open system as it doesn't allow users to alter its libraries. Additionally, certain features embedded in the system can help users have a better control of the model, including
1. Time and name stamp for any modifications to the tables and functions.
2. Version comparison, can compare the selected codes, tables and toggles between multiple versions.
3. Different levels of read and write permissions for libraries and data.
4. Tools and reports for model validation, which assist in back-testing, roll-forward analysis, step-by-step calculations.
5. Audit tracker for approving job submission and reviewing result log.
6. Input data validation.
7. Results comparison.

Flexibilities

An ALM model is often used in R&D projects. Model users like to add some customized settings or calculations for new products and Ad Hoc assumptions. An open system is usually more flexible regarding the customized settings and calculations. A closed system only provides a certain amount of flexibility. Hence, quick responses to the customized requests from the vendor are very important.

Extensibility

Model users are likely to extend model coverages for new business, new projection (such as, PBR) and new regulation (for example, 2017 tax reform). For a closed system to handle a new regulation or projection, a new release of the closed system is usually required. Therefore, timely support from the vendor is crucial.

Further Consideration

Unlike many years ago, there are now many vendors developing software for insurance companies to model assets and liabilities A one-size-fits-all platform is not yet available. Besides having different preferences regarding costs, model governance, and flexibility, the company may also need to consider the availability of ALM modelers for the platform within the organization and the availability of recruiting qualified individuals from outside.