While actuarial science has been leveraged for decades in insurance pricing, the industry stands to benefit greatly by both expanding the methods being used, and extending those methods to claims operations and underwriting. For example, predictive modelling can help with early identification of people whose injuries are likely to become more severe so that proactive action can be taken. However, using predictive analytics effectively across an entire business can be complicated.
Typically, actuarial teams develop various models and test them against a data set to identify an analytical approach that performs best. Deploying models into the operation often consist of emailed spreadsheets and triage lists of potentially problematic claims or reviewing difficult policies amid policy administration. This offline approach to rollout limits the probability of success.
How this happens is not surprising. Actuaries are not experts at core system integration. Even for IT experts, there is a lot to consider beforehand. Who will be using it? How can the data be made usable? How can insurers oversee the operation and impact of the models on an ongoing basis?
To learn about the four-step process for predictive analytics success, please read my Actuarial Post article.