Many insurers are in the process of modernizing their life and annuity policy administration systems to achieve faster time to market, greater flexibility, and digital capabilities improvements. The nature of life insurance is such that it can remain in force over protracted timescales. Simple products, e.g., term life, involve fairly straightforward processing once the insurer has issued the policy and set the rates. However, transactions that occur over time, e.g., those of complex products like universal life, variable annuities, can result in significant variations in policy values, automatic processing, and customer communications.
Modifications to policy administration systems can introduce errors over time—errors that may not surface until years later. It is, therefore, imperative that insurers have rigorous regression testing programs in place.
It is not enough to perform positive testing to confirm that any new changes work according to requirements. Systems that are 30-50 years old can have numerous errors festering in the environment before discovery; these errors can require fixes that do not align with normal system processing, i.e., hardcoding.
Insurers have developed and integrated formal regression test processes into their systems development lifecycles to address these errors. Central to these procedures is establishing a robust testbed that represents all insurance products and policies present in production at various durations. There also needs to be a broad set of transactions to process against these policies; the resultant data represents the baseline of expected results. Insurers rerun the test after a new release of code changes, comparing the results against the baseline. They then refer developers to any unexpected differences that the test identifies for correction.
The challenge for insurers performing regression testing is to strike a balance between the number of policies, the number of years to cycle forward, the transactions to include, and the data elements to compare. Too much of any element can result in unwieldy processes that take too long to run, distract from the development of new functionality, and stifle time to market and agility. Too few of any can allow errors to slip by and hide undiscovered until years later. Most insurers can point to a watershed moment that hardened regression testing requirements: a time when some significant error was communicated to agents or policyholders.
Unfortunately, strict regression test requirements can take weeks or months to run and review for many companies. These timescales can limit releases of code to twice or three times a year. They can be the primary obstacle for implementing an Agile methodology, especially if the architecture of the admin system is tightly tied to surrounding systems.
Companies contemplating PAS transformation strategies need to ensure that the new system considers the persistent and complex nature of life insurance and provides an efficient means of accommodating regression testing as part of its code certification process. Some vendors have begun to address this challenge as part of their solutions.
One method for cloud-based solutions developed using an Agile-first approach is incorporating regression test capabilities into the architecture of the system. Doing so allows developers to run through test scenarios earlier in the development process. These scenarios relate directly to the functionality they are modifying, which avoids hitting a Big Bang Test brick wall at the last minute. Engineering testing capability into the core system also makes it possible to set up regression test monitoring as part of normal production processing. Systems could use artificial intelligence as part of an early alert system when internal system values do not align with expected results.
Insurers will need to bake regression testing into their transformation requirements rather than treat it as something to design and develop after selecting a system. Otherwise, insurers risk repeating the same quality issues that plagued legacy PAS systems.
For more information on quality control best practices for insurance systems, refer to Novarica’s recent executive brief QA and Testing in Insurance: Expansion and Key Issues.
Add new comment