When predictive analytics meets an unpredictable world
No big data sets, predictive analytics or machine learning could have predicted the current state of affairs. COVID-19, of course, captures the imagination. This makes it easy to forget the enormous insured losses from natural disasters this year as well. So with the world in such a state of flux, what does this mean for predictive analytics used by insurers?
In the pre-pandemic world, insurers were making huge investments in their predictive analytics capabilities. For example, a report by Willis Towers Watson last year found that 70% of life insurers surveyed in North America were deploying predictive analytics.
This technology is already in use in several business areas and across a wide range of use cases. Some of the big-ticket deployments include customer churn prevention, optimizing loss reserves and claims payouts, assessing optimal litigation management strategies, and of course, providing a more accurate assessment of risk when pricing policies.
However, with governmental responses to the pandemic still changing drastically within short time frames (case in point, European nations being plunged back into lockdowns), the efficacy of existing predictive analytics models are undoubtedly being called into question. As an industry, it is right that we’re now having this conversation, as we must ensure our models evolve in line with this new risk environment. This is in order to be more resilient to not just the immediate challenge of COVID-19, but also to the much longer-tail impact of climate change.
Nothing is sacred: Re-examine everything, continuously
The first place to begin is by assessing and auditing existing models. Here are some critical areas that must be included in any audit:
- Claims frequencies: Personal lines policies have been particular impacted here due to COVID-19. Homes occupied all day due to home-working will likely suffer fewer burglaries, but incidences of accidental damage are bound to increase. Less congested roads could result in a sustained drop in minor traffic accidents, but could perhaps increase fatal accidents as people are driving faster on empty roads.
- Personal data points: How robust are the assumptions behind data points used for premium calculations, such as credit scores? Millions of policyholders could see their credit scores deteriorate due to the economic impact of COVID-19, so do existing assumptions on risk modeling with this data still hold true in this context?
- Policy exclusions: This is a continually evolving area as we await further significant rulings in jurisdictions around the world concerning COVID-19 coverage.
These models will then need to be reassessed and recalibrated on a far more regular basis, to better manage the constant changes we’re witnessing within the new risk environment.
Ensuring quality and relevancy with third-party data
Accessing valuable third-party data sources can be expensive and time consuming for insurers. This, inevitably, can lead to a set-it-and-forget-it mentality. But as predictive analytics is only as effective as the quality and relevancy of the data being analyzed, insurers must pay incredibly close attention here.
Much of this will come down to the granularity of the data. For example, if you are a workers’ comp insurer, does your COVID-19 infection data provide infection rates within workplaces, and if so, does it split it by different workplace settings?
As well as ensuring quality and relevancy, insurers must also make sure they have the expertise to appropriately analyze and interpret this data. Meteorological data for modeling the impact of climate change against insured assets, for instance, is complex, to say the least. Therefore, creating teams such as a climate risk board, that oversees climate risk and interprets predictive analytics data, can help form more informed strategies, from underwriting to planning for claims surges.
Insurers also must not overlook the power of collaboration. Working with science communities can help insurers understand how to make better decisions based on broader data points and detailed risk assessments.
Overall, it is the insurer’s responsibility to verify the quality of their data sources. Cataloging data and continuously updating information management processes might be time-consuming for insurers but can reduce the risk of using irrelevant or poor quality third-party data.
Creating optimal vertical integration
Many insurers approach predictive analytics as primarily an underwriting and actuary tool. However, siloing this technology into a single business function creates a huge amount of missed opportunities.
Within claims handling, analytics should be deployed to provide more accurate reserves at the FNOL stage and throughout the process. This is particularly useful during black swan events such as COVID-19, as well as with dynamic climate change-related events such as wildfires. It should also be deployed with the objective of reducing pressure on internal resources during claims surges, such as through part-automating claims payout calculations within certain criteria.
Another area of deployment within the industry is to better manage the litigation stage. This is of course very pertinent in the current business climate, given the deluge of litigation cases insurers have faced from policyholders excluded from cover for COVID-19-related losses. In the U.S. alone, there are currently more than 1,000 cases working their way through the courts. Analytics deployed in this setting can help insurers make more informed decisions on when to settle or not.
Ultimately, with the on-going impact of COVID-19 and the ever-increasing incidences of extreme weather events, implementing effective predictive analytics solutions across business divisions will help insurers face this brave new world. The effectiveness of this, however, depends on a large number of variables that insurers must intelligently control.
Finally, unlike with Covid-19, the insurance industry has the chance to prepare for what climate change will bring. Effective preparation will require forward-thinking leaders who can embrace the evolution of predictive analytics to future-proof their businesses and adequately protect their policyholders.
Read More …