The Center for Data Innovation spoke with Bill Pardue, CEO of Maryland-based data provider Weather Analytics. Pardue discussed the shortcomings of the National Oceanic and Atmospheric Association’s (NOAA) data, what kinds of data are hardest to collect, and how Weather Analytics could help fix the insurance market in Cape Cod, Massachusetts.
This interview has been lightly edited.
Travis Korte: What are the main problems with NOAA data, and how does Weather Analytics help solve them?
Bill Pardue: NOAA’s mission focuses on protecting life and property, informing the public, and conserving natural resources. At Weather Analytics, our primary goal is very different: delivering actionable information to businesses to make better risk-and-reward decisions where the weather conditions can have a big impact.
So, there is a mission mismatch between NOAA’s objectives and business needs.
NOAA data is not cleansed, rationalized, stored, nor made accessible in formats well-suited for business applications. In the publicly available NOAA archives, there are troublesome gaps in time, place and weather types that don’t matter when people are trying to decide where to vacation. But when the choices are where to underwrite property insurance, or how to optimize transportation routes, or understanding weather patterns and impacts on commodity trading prices, those gaps matter.
TK: What are some things you’re currently working on adding to the database?
BP: Downscaling is our most exciting and meaningful initiative. Using proprietary algorithms—and some of the best meteorology talent in the U.S.—we are increasing our database’s resolution from 35-by-35 km squares to 1-by-1 km.
We know of no other organization in the world—either government, academic or private sector—that is undertaking creation of such micro-climate reporting on an operational, large-scale basis.
This program means we can access historic weather information on a hyperlocal basis. We are moving into micro-climate forecasting as well. Weather data that is reported property-by-property, or neighborhood-by-neighborhood, is a new frontier for weather companies and business applications.
TK: You mentioned that certain types of weather data, like lightning, are harder to collect than others. What are some more examples, and what makes them so hard to collect?
BP: Information on hail also can be highly challenging to collect on a consistent and widespread basis. The issue is understanding the complexity of weather conditions that must take place to create hail. In meteorological modeling there is no single observed variable for hail, but instead a culmination of several variables that must all take place for hail to occur. Modeling is improving and as Weather Analytics continues to downscale this also includes parsing through micro events such as hail to create a more granular and unique picture of the atmosphere.
TK: You gave a fascinating use case for Weather Analytics data during your talk at the Big Data for Decision Makers meetup last week. It centered around property insurance in Cape Cod. Can you repeat it for our readers and elaborate a little on how Weather Analytics could help?
BP: More than 40 percent of homes on Cape Cod are insured by the state of Massachusetts, not by private carriers. A key reason is that the current information on weather data and storm and risk exposure is so poor and broad-brush that most insurance companies either won’t insure Cape Cod homes, or otherwise offer prohibitively expensive premiums.
What ends up happening is that owners of the most expensive homes—most often those on the coast and of the highest value (and also with the highest severe weather exposure)—pay for coverage as though the risk were the same as for homeowners more distant from the coast—the latter often having much less risk exposure.
Effectively, this practice means middle-class property owners inland are subsidizing the luxury of living on the coast. Using hyper-local weather data, Weather Analytics can run historic models to determine the wind or storm profile of any given location, using downscaled data that can determine the difference between highly exposed houses near the shores and much less exposed properties only a couple miles inland.
This knowledge in the different levels of risk opens up opportunities for private insurance companies to offer premiums that are market rate to the inland properties.
TK: Your current major client is In-Q-Tel, but what do you see as the long-term distribution of industries Weather Analytics will be working in? Does this resemble the landscape of areas where weather data could be useful, or are you after a particular subset of the market?
BP: Weather intelligence of the breadth and depth we offer presents opportunities in dozens of vertical markets. Achieving the right laser focus on a target market is a paramount task for any early-stage company.
Delivering for the U.S. intelligence community is enormously important: a given. Next up, the insurance, energy, and engineering sectors.
But one of the most exciting pieces of being a technology company—sitting atop a huge data asset such as ours—is the ability to be nimble and scale quickly. As significant opportunities develop in 2015 and beyond, we’re ready to take them on.