How data-driven insurers are succeeding with advanced analytics

“Advanced analytics” can be a subjective term. What really qualifies as the “advanced” use of analytics in P&C today? Is it a matter of technique or mindset?

While the answer is both, a data-driven culture inevitably determines a company’s proficiency in the use of advanced analytics. Data is nothing without analytics, and “advanced” analytics is sophisticated math — the ability to reveal and find correlations and patterns in data. Put together, though, they are a company’s secret sauce and can yield a resounding competitive edge.

One of the biggest obstacles insurers face today is trusting their analytics. Building models, especially predictive and prescriptive models, can be difficult for risk professionals to accept and trust. We tend to look for a corner case, an example where a model isn’t “right” 100% of the time. However, we live in insurance, and insurance is all about the law of large numbers.

Leading insurers today prioritize predictive accuracy over “explainability.” According to a 2021 survey report from Willis Towers Watson, more insurers are making use of AI and machine learning. Around 30% reported the use of AI to better understand risk drivers, up from 20% in 2019. However, Insurity’s 2021 Analytics Outlook Report found that while the adoption of predictive analytics has gained traction, there’s still work to be done on the trust-the-model front, with more than 70% of insurers surveyed expressing moderate to high concern over underwriter adoption of predictive analytics.

Every company has a different appetite for data and analytics. For some, it’s difficult to give up a little bit of “gut instinct” and go with the data. But having the discipline to follow the data leads to more sophistication and automation, including predictive and prescriptive models, straight-through-processing, no-touch claims and product innovation.

Consider Progressive. The commercial auto industry has struggled for years, yet Progressive leads the industry by a wide margin — two and a half times more direct written premium than its closest competitor and three times the market share.

How did Progressive do this? By investing in its existing data and expanding on that investment by purchasing telematics devices. With these devices, they have accumulated billions of miles of driving data, providing precise predictions around good driving behavior and enabling them to discount and incentivize commercial haulers for safe driving habits.

On the property side, one independent reinsurance brokerage has invested in real-time automated weather event notifications. As climate risks increase, the technology-forward broker needed a way to keep its carrier clients informed 24/7 while providing them with analytics around business impact.

4 tips for taking advanced analytics to the next level

As insurers become more sophisticated in their use of analytics, here are four critical components of a data-driven culture:

1. Prioritize predictive accuracy over explainability: Leading carriers have moved into sophisticated machine learning, allowing them to find correlations in data that can’t be done with traditional predictive methods. Machines provide a more holistic view by aggregating all the data and providing a final answer that tends to be more accurate and predictive. The use of machine learning in auto underwriting and claim frequency estimates, for instance, has added a dramatic lift in predicting the likelihood of future accidents. In those models, several predictors are intuitive, but many are not.

For example, using telematics data, the number of hard stops and instances of speeding at time of day are intuitive predictors. However, machine learning variables that interact driver’s age/experience with wheelbase or curb weight along with G-force turns may not be straightforward variables everyone understands. In the end, machine learning and AI are being leveraged to find nonintuitive interactions that provide additional predictive power, and therefore competitive advantage.

2. Go beyond predictive to prescriptive: In the next phase of advanced analytics, the industry will shift from predicting the future to prescribing actions or next steps.

Using auto physical damage claims as an example, predictive analytics will quickly estimate the total incurred loss for damage to a vehicle. Prescriptive analytics, however, goes further to recommend a repair shop along with the exact repairs to be made.

Likewise, in property underwriting, not only can models predict the likelihood of water losses, but they can also recommend where water sensors are placed. In this way, prescriptive analytics can improve the customer experience, prevent losses, and create competitive differentiation.

3. Integrate real-time data and analytics: Speed of insight is yet another differentiating factor among leading insurers. Making data-driven decisions in the moment and taking advantage of third-party data is increasingly important.

For example, with climate change and catastrophe risk, the past is becoming less representative of the future. Hazard models and historical data are being replaced with real-time data and innovative technology to better predict and manage catastrophes like hurricanes and wildfires.

It’s a 24/7 effort for carriers to process the latest datasets from multiple hazard providers, including increasing amounts of unstructured data.

Being a week, a day or even hours behind can create a competitive disadvantage. To keep pace, leading insurers leverage partners to streamline data ingestion and combine it with the latest snapshot of their portfolio data. With the data immediately consumable, they can focus on applying analytics to inform business impact, deploy claims response and set moratoriums.

Likewise, on the predictive modeling side, a shift to real-time models that incorporate new data immediately into the modeling environment allows for the early identification of trends and keeps models from going stale.

4. Modernize data platforms: Before insurers can do any of the above with proficiency, a solid data infrastructure must be in place in order to fully leverage an organization’s own data, ingest third-party data and apply all the analytics.

For many insurers, data management, handling and warehousing is still a key factor in slowing progress on the advanced analytics front. Leading insurers are modernizing their data platforms to accelerate progress and maximize the potential of data and automation.

Being a leader in advanced analytics today isn’t just about technique, it’s also a matter of mindset. Leaders today maximize the potential of data and automation in underwriting and claims. This is a conscious and constant endeavor — data doesn’t just fall into their laps — they are willing to pay for it and test it.

Beyond that, companies who are sophisticated in their use of analytics have figured out how to take the output of models and real-time analytics and apply them to the business for more efficient claims handling, lower loss ratios and improved underwriting profitability. This is a competency that is rooted in a data-driven culture. It is the marriage of data with advanced analytics that is a company’s defining differentiator and driver of profitability.

But it’s not always intuitive. AI and machine learning aggregate massive amounts of data and find correlations beyond a human’s ability to grasp. If you build a model and prove that it works, that becomes a pattern. Every time that model says it’s a good risk and you have better performance, well, maybe it’s time to trust that magic eight ball?

Kirstin Marr is Insurity’s chief analytics officer. She is a recognized thought leader, specializing in data and predictive analytics in the insurance market. Prior to this role, Kirstin served as president of Valen Analytics, which was acquired by Insurity in 2017, and previously served as the company’s head of marketing, leading market strategy, brand awareness and innovative thought leadership.

Opinions expressed here are the author’s own.

Related:

• Ford Motor Co. expands U.S. user-based data services
• Leveraging data to better anticipate & plan for NAT CAT events
• Using location data to maximize efficiency & cut costs