Progressive Insurance houses an impressive amount of information, said Pawan Divakarla, the company’s data and analytics business leader, during a keynote at INN’s Insurance Analytics Symposium, held this week in Boston.
Its Snapshot car-driving monitoring systems alone has collected data from cars that have driven 14 billion miles, according to Divakarla. “We’ve gone further than [the] Voyager” space probe, he said, outlining some of the ways the insurer is using big data and analytics now -- and some of the ways the insurer might tap into the technology in the future.
Data is nothing without analytics, Divakarla said. To assist its information analysis, the company has built a data warehouse and a framework that consist of data access and modeling tools such as SAS, R and new tools like H20; algorithms such as the generalized linear model (GLM),gradient boosted machine,(GBM), Random Forest and some machine learning tools; and a presentation layer featuring Tableau and D3 JavaScript interactive visualization.
The company is using its big data for fraud mitigation, marketing, finance, claims management and telematics, he added. And, Divakarla said, discoveries will be coming faster. Whereas Progressive might have needed a month to analyze some data in the past, it can now do it in nine hours, allowing the company to do two or more different models in a given day.
One of the company’s goals with its analytics is to segment its customers. Another is to more accurately price its products. He said analytics allow insurers to keep improving pricing algorithms, even if it’s in increments. Incremental gains using big data, over time, he said, can give an insurer improvements that end up not being incremental.
Analytics is crucial to helping Progressive figure out what’s “really happening” in certain situations, Divakarla says. For example, if a Snapshot user is stuck in snow, they might be revving the engine and spinning the wheels – indicating erratic behavior. Progressive is using analytics to try and determine bad driving from spinning tires.
Another insight came from the device’s voltage readings. Fluctuations in those readings can be analyzed to spot possible problems with a car’s alternator. The company then has the ability to alert owners that there might be a mechanical problem with their cars.
“The future is here,” he said.
