What Big Data Means For Infrastructure Costs
As data-hungry insurers collect more data from new and different sources to get to know their customers more intimately, they're finding that big data efforts require massive upgrades in their storage capabilities, and many are turning to the cloud for help.
It's so early in the game right now that you're not sure what data is valuable and what isn't, not to mention how long to keep it, says Jim Ditmore, senior VP at Northbrook, Ill.-based Allstate ($28 billion in premium written). "Because it's generated in huge quantities, how do you ensure you're storing the right data for the right length of time when in reality you don't know which is the most valuable?"
New sources of data that insurers are looking to tap into at a higher level include proprietary telematics as well as site behavior and third-party data such as search engine information, Ditmore adds. And an increasing amount of this data is unstructured, such as the dozens of iterations of a search a potential customer does on an insurer's website when looking for car insurance. Unstructured data requires more work on the analytics side -- another IT expense.
"You now have a vast sea of data you can apply analytics to, as well as other sources of unstructured or mobile Internet data, all of which are multiplying at a high rate," he says. "And you have the big data analysis models, which are largely based on Hadoop and other analytics." If you store all of it yourself, you're going to quickly overwhelm reasonable storage capacities, he says. You have to filter and select the best and most appropriate data and decide how long to keep it.
From a pure storage standpoint, Ditmore believes storage vendors are doing a good job keeping up with the demand. If that changes, insurers may have to turn to public cloud providers for some big data needs, he says.
For now, data center hardware costs are flat or dropping, Ditmore says. But the amount of data being collected is growing at such a high rate that companies are looking to third parties to do data analysis.
Dave Shively, insurance consultant for BPO service provider Mphasis, says he's seeing clients doing just that. This approach lets them focus on storing and structuring the data that gives them the greatest competitive advantage, while delegating responsibility for hosting the analysis of that data to the third-party provider.
"We entered into a relationship with a cloud vendor and took the nondata-related assets -- the Web servers and application servers, for example -- and put them on the cloud," Shively says. "Thus we can leverage the public cloud for what it's good at, which is providing high-volume on-demand services, while at the same time, you've got the data stored in a way that's not open to the public domain."
That's not to say that data stored off-site is open to cyber attacks, Shively says. Cloud vendors have made strong advances in data security. Insurers simply want to go the "extra mile" with their proprietary, sensitive data.
Insurers are lagging in their uptake of cloud-based computing, Shively says. Some of that is because of all the regulation they face, but the amount of data subject to some sort of regulatory constraint "is fairly finite and can be segmented," he says.
It's for this reason that Allstate's Ditmore says he doesn't expect insurance companies to upload much more data to the cloud in the next five to 10 years. "Leveraging the third-party stuff, which is really very ancillary, that definitely will occur in the short term," he explains, noting that much of the third-party data is anonymous, lacking personally identifiable information.
Nathan Golia is senior editor of Insurance & Technology. He joined the publication in 2010 as associate editor and covers all aspects of the nexus between insurance and information technology, including mobility, distribution, core systems, customer interaction, and risk ... View Full Bio
Tweet This