Data Points

 View Only
  • 1.  What is weather data's legacy in the insurance market?

    Posted 10-02-2019 23:31
    While visiting our partners at Baron Weather yesterday, I was asked a really good question: "How has weather data been used in the insurance industry in the past and present, and how do you think it will be used in the future?"​

    What came out was long and complicated, so I decided to write about it.

    Long story short: Weather data has been a part of the insurance industry for decades, but it has never played as significant of a role as it should, because, so far, only static weather data has been captured in an easily "deployable"​ format. Read on if you care to hear this statement explained:

    • "as significant of a role as it should" – Most property and casualty insurers stack rank perils by level of importance and then assign each peril a weight when building underwriting algorithms. Importance will obviously change as the geography changes, but flood and wildfire data are usually prioritized towards the top while weather data is found towards the bottom. This is true even though wind and hail events are responsible for ~50% of all claims in convective storm heavy years (that's a lot of billions). The reason for this is because the market has not provided insurers a great way of understanding weather data - at scale - when and where they need it… more on this below

    • "easily deployable format" – Insurers need to understand weather impact as it relates to property locations. They don't care about wind magnitudes in an empty field nor do they care about wind magnitudes aggregated to incoherent boundaries like counties or states. Weather APIs (like Baron Weather's) are great for developing fit for purpose applications that address this challenge, but it's still difficult to assign weather impacts to property level locations accompanied with a date and time stamp. First, the raster image must be converted into a more "manipulation friendly" format, like vector, to achieve data extraction at scale. Once this is complete, the data can then be organized into a self-descriptive schema and delivered in a database agnostic format such as .txt. Most vendors understand that .txt files are much easier to work with than raster files, but converting the format of live weather APIs is a clunky, costly data processing exercise. If you figure out how to streamline that process and deliver in .txt, then you are removing some serious complexity for your customers and enabling them to "operationalize" their weather data more seamlessly (btw, we've figured that out).

    • "static weather data" – Understanding weather information immediately before and after damaging events is where most insurance use cases fall. Proactive use cases rely on accurate forecast data to predict damages at property locations before they actually happen. This allows the insurer to take action on reducing damages (and ensuing claims) beforehand. These actions could include alerting, implementing moratoriums, educating their customers, or all the above. In general, the longer the timeline - the more valuable the forecast data. Reactive use cases rely on readily available historical data that describes weather impact at property locations minutes, days, or even months ago. The quicker the insurer can access this data - the more valuable it becomes. There are some underwriting/actuarial use cases that look at a long-tail of historical data for several years into the past, and the "readily available" requirement becomes less important. This aside, high value insurance use cases hug the bookends of a damaging event. Right now, there are plenty of weather data products on the market which deliver in an "easily deployable format" (see bullet 2), but they are very static and miss the window of opportunity around the damaging event. A monthly dump of weather data cannot aid in claims validation, adjuster coordination, alerting, and several other time sensitive activities that are key to an insurer's bottom line.

    All 3 points above will be required of weather data in the immediate future.



    ------------------------------
    Bryan Bonack
    Data Product Manager
    Pitney Bowes
    Boulder, CO
    ------------------------------


  • 2.  RE: What is weather data's legacy in the insurance market?

    Moderator
    Posted 10-03-2019 14:23
    I actually wrote about this a few years back as I thought about the problem insurers have with customer satisfaction and loyalty. Most people resist using insurance since having claims leads to rate hikes, and in some cases policyholders being dropped by carriers. Neither is good for business because of the churn it creates.

    Marketers in their wisdom tout 'claims paying satisfaction' because they don't have too many other good metrics or value props to promote. Start using this data to help clients avoid claims and it immediately benefits both parties: churn and cost go down, and policyholders start to think about insurers as delivering value whether there is a event or not.   

    As a customer guy I get it and think peers in my marketing would too. Actuaries see the numbers and create the algorithms for rating which also takes into account churn. Consumers are conditioned to expect/use real time alerts, so leveraging this kind of data to help them prepare for situations as best the can when an event is bearing down on them or their loved ones seems to be a no brainer. Can't wait!

    ------------------------------
    Dave Andrews
    Pitney Bowes Inc
    Stamford CT
    ------------------------------