Renewal seasons are a time for the company’s executives and underwriting teams to boost their portfolios’ performance. During this time, they review deals and consider how they might improve these. Since people in business do not have entire weeks or months to spend on strategising, they have to assess their plans rapidly. They must know which changes align the best with their strategic goals.

When putting together a renewing portfolio, there is a trade-off between risk and return, with many possibilities for changes. For example, suppose you would like to evaluate how you can maximise returns and minimise Value at Risk on a portfolio with just ten treaties. In that case, you could end up with 1,010 combinations or ten billion new portfolios. Even if you could evaluate 1,000 portfolios per second, it will still take three months to process everything. An increase of five, meaning 15 treaties, will need 31,000 years to complete!

Insurance Software: A Way Forward for Renewals

With machine learning, financial services companies can make renewals more efficient. Multi-objective genetic algorithms, in particular, take an approach that allows them to sample and iterate combinations to the most feasible ones. Machine learning can augment—not replace—the underwriting process, allowing service providers to process more in a shorter period. Note that the algorithm does not choose the portfolios; that still falls on the portfolio manager. What the machine does is narrow the choices to the most optimal settings.

Machine Learning for Portfolio Optimisation

One way to illustrate how machine learning works is to draw parallels with finding a car in a parking structure. Suppose I ask you to find my slot, but I don’t give you the floor number or even the parking building’s name. Instead, you have to find my car in a general area or neighbourhood. There will easily be hundreds, even thousands, of parking slots in that set. If you check each slot yourself, it will take you months to finish. That is what happens in manual portfolio optimisation.

When you use machine learning, you don’t need to go to every single parking space. Iterative resampling allows you to “collect” locations and ask a series of questions about them, narrowing your focus each time. For example, you can ask, “Is parking slot x in this neighbourhood?” Then, you can go down to the block level, then the street level. Portfolio optimisation chooses only the best set of solutions to produce the next set, so you are sure that the latest iteration would be the one that maximises profits and minimises risks in the best possible way.

The Future of Financial Software Development

Machine learning is becoming a standard in today’s financial services sector. In the future, the industry might see AI or machine-learning-based chatbots that can produce responses to complex questions about risks and rewards for specific actions. It sounds like science fiction at present; as with autonomous cars and augmented-reality eyewear, there is a lot to be done before this becomes the norm. However, with underwriters using machines to boost their efficiency, it might come sooner than we think.

Conclusion

Reinsurance is a traditionally highly involved process, and financial service providers need to be quick in completing portfolio renewals. Machine learning keeps managers efficient and enables them to provide the best recommendations for their clients.

Always at the forefront of technology, Informatics provides customised insurance software solutions. We help insurers and financial service providers re-engineer their processes and offer continuous support to their clients. Contact us today to learn more.

Written by Daniele Paoletti