12.31.20
Phil Edmundson

Property Insurance Meets Big Data

Corvus uses new data sources and machine learning to better predict and prevent commercial insurance claims.

We are a team of innovators, equally derived from the tech and insurance worlds on a mission to reimagine commercial insurance. We aspire to make the world a safer place by changing the insurance buying decisions of organizations. 

Utilizing Data to Make Informed Property Insurance Decisions

As “big data” and data science techniques have made inroads into the insurance industry, some of the most obvious applications were the first to be successfully productized: cyber liability underwriting and parametric personal auto insurance, for two examples. But none of these touches on the broadest category of traditional insurance markets that can be impacted by these technology and process advances: Property Insurance. 

The earliest form of commercial insurance -- dating back to the ancient Phoenicians -- Property Insurance provides payment for economic loss as a result of damage to property. It has been grown to include resulting business interruption, additional expenses, and other related financial losses as it has expanded over the centuries. 

Using data for underwriting this type of risk is not a new convention. News of shipping ports, storms, and wars influenced early Cargo and Ship Insurance. Today, of course, data sources and forms are exploding. The volume of recorded electronic data in Internet of Things (IoT) devices alone is estimated to be 14 zettabytes (ZB). That’s an astronomical amount of data: the entirety of the world’s digital data only exceeded one zettabyte in 2012. Projections from Statista show that rising to five times that amount in the next five years.

[QUOTE] "It's exciting to imagine what kind of surprises may be lurking in the data of property insurance."

New data sets that can better inform property underwriting are easy to identify in our everyday lives - satellite imagery, Google street views, weather reports, social media, mobile phones, building system sensors, and temperature sensors just begin to scratch the surface of new sources to enable. The data can lead to individual company insights - we score Cyber Risk for a Business Interruption caused by hackers of logistics software, for example. The data can also lead to macro insights. Our analysis of tens of millions of temperature data points showed us that the risk of spoilage on some shorter routes was greater than for longer routes - a counterintuitive result. It’s exciting to imagine what kinds of similar surprises may be lurking in the data for Property Insurance. 

Much of the data needed to explore properties is freely available from various government agencies. Other data access requires commercial agreements. However, some of the unique and revealing data can only be obtained by creating the data source - requiring that each location mount a sensor for data collection, say a water leak detection device at the base of a home water heater. 

Brokers Remain the Linchpin

There are several challenges to tech-enabled underwriting and buying behavior of commercial Property Insurance. First, the data frequently needs to be structured. Google Street Views may give us a visual view of the distance of the first floor from grade but building the software that moves from address to underwriting input is a challenge for internally built technology. This is where Corvus comes in. We sort, clean, and structure the data. Then we use machine learning tools to analyze it for underwriting insights. 

We also analyze market opportunities by evaluating not just the source of the data, the friction involved to obtain it, and the cost to structure it in a meaningful way-- but also through a comprehensive understanding of the insurance brokerage distribution changes required to ensure the data brings value to all of our stakeholders.

The Corvus playbook is optimal when we can extract data that informs our underwriting but also bring value to all of our stakeholders, including both policyholders and brokers. So, the same data insights that can inform our underwriting decisions in order to allow us to better price and choose risk (and be a more profitable insurance underwriter) must also be converted to our Dynamic Loss Prevention reports that inform policyholders of ways to manage their risk and provide data analytics to insurance brokers as a motivation to promote our insurance solutions over those of incumbent insurers. 

The challenge then is to understand the insurance ecosystem and the decision points in insurance buying behavior. Most commercial insurers have no direct visibility into buyer behavior. In fact, the last major direct seller of commercial insurance, Liberty Mutual, abandoned most direct sales a decade ago and moved to the brokerage channel. 

At Corvus, we think that the recommendation of commercial insurance brokers determines insurance buying decisions.

[QUOTE] "Broker recommendations determine insurance buying decisions at Corvus."

We build our software to leverage new data that produces not just our policyholder-facing Dynamic Loss Prevention reports but also to bring advantages to brokers. Speed, ease of use, and data analytics top the list of attributes that brokers value. 

So we share our risk insights with our brokers, and in turn, brokers use our data to better serve their clients creating a virtuous “flywheel” of reduction of risk and fewer claims. 

Data Leads to a Broader View

Most insurers count their own claims data as the primary source of developing rate plans. They study SERFF filings of competitors for additional clues. All of these are based on data that is incomplete (no insurer has a large market share of most lines of insurance) and out of date, with longer history seen as a virtue rather than a recipe to repeat prior mistakes. Corvus views larger sets of data as more valid both in scope and in time. More recent data, if complete, can offer better predictions of the future. So, we look for very large data sets like those available from government sources or from third parties like Google. 

New categories of Property Insurance risk await our next generation of Smart Insurance products. Some will be targeted at specific industries while others will focus more on a peril, such as flood or earthquake. We exist to make the world a safer place and if we can activate more data then we will continue to succeed.

[RELATED POST] Tech Companies: Beyond Cyber Risk, the Cost of Downstream Impact

Tech Companies: Beyond Cyber Risk, the Cost of Downstream Impact

The rise of remote work and growing concerns over ransomware acted as partners-in-crime to get organizations to hone in on risk mitigation efforts over the past couple years. Through compiling our Risk Insights Index, we found that with certain initiatives —  safer or reduced usage of RDP, growing use of email security tools, and other measures taken to limit the impact of threat actors — businesses are more prepared than a year before and ready to play defense. Those efforts are borne out in our finding that the rate of companies who pay a ransom when attacked with ransomware fell by half within a year. 

[RELATED POST] Tips from Top Brokers: How to Play Offense in a Cyber Hard Market

Tips from Top Brokers: How to Play Offense in a Cyber Hard Market

The whisperings of “firming rates” start first, quietly in business meetings, then published in industry reports. Soon to follow, rumblings of a “hard market” are brought to the conversation. It’s cyclical in nature, and we see it across all insurance lines at one point or another. For years, Cyber Insurance stretched far and wide with “soft” market conditions, remaining highly profitable. Now that period of growth, with exceedingly available coverage and inviting terms, has stalled in the face of a hard market.