Big data” seems to be all therage these days, but just collecting it isn't enough. If the dataisn't helping your organization be nimble, it's time to startlearning about high-performance analytics.

The amount of data available for analysis is expandingexponentially. Sentiment data, sensor data, transactional data,third-party data and other big-data sources are streaming in.Insurers looking at scenarios need to make decisions in minutes orat most, hours—not in days or weeks.

High-performance analytics (HPA) allows modelers to work fasterwith full sets of data—not sample sets. This can provide the typeof nuanced analytics that allow an insurer to be more successful inhigh-risk situations, create innovative and profitable new means ofinsuring customers, and respond to large-scale disasters in arelevant and cost-effective way.

AN END TO SAMPLING

Today, many insurers are using advanced analytical techniquessuch as generalized linear modeling for rate-making and productpricing, and a recent survey byTowers Watson showed that 70 percent of U.S. insurers are usingpredictive modeling for Personal Auto insurance.

However, actuaries often rely onsubsets of historical data to run pricing models since it is tootime-consuming to prepare large sets and run the models.

But sampling has its limitations. If you are pricing claims withone-million records versus 10,000, you will be much less likely toencounter outliers that throw the models off.

By using HPA, insurers can now get more precise analysis from all their data. They can also incorporateexternal data (such as Google maps, GPS, credit scoring, socialmedia, etc.) to supplement the results.

Take the example of segmentation and pricing. Insurers currentlyuse a handful of variables to support those two processes. Becauseit operates much faster, HPA helps insurers increase the number ofvariables used in “what-if” analysis to find the ones with thebiggest impact on profitability.

Here are three more areas where HPA can improve insurerprofitability:

• Telematics: According to a study by ABI Research, the number of telematics users willincrease from fewer than two million in 2010 to nearly 90 millionby 2017, inundating insurers with data from these in-car recorders.Many insurers already are struggling to analyze existing telematicsdata.
  As it increases, HPA is the only option to analyze billions of datarecords in a fraction of the time required by traditional computingenvironments. With that ability, insurers can better understand howto use telematics to create innovative rate plans and learn moreabout how driving habits influence claims. 
• Customer Intelligence: As customerinteractions in insurance move from in-person to digital channels,insurers must react faster and better predict futurebehavior.
  Insurers can also improve customer experiences and make relevant,real-time offers with higher acceptance probabilities. Fasteranalytics delivers predictive-modeling results more quickly andidentifies the best future actions to take while considering bothfinancial and organizational constraints.
• Catastrophe Modeling: Insurance companies arewell-equipped to manage the potential losses associated with claimsfrom individual fires and automobile accidents because of a wealthof data. Actuaries can determine future losses with a high degreeof confidence.

However, since catastrophic events are relatively infrequent andhistorical data is limited, it's virtually impossible to reliablyestimate potential future catastrophe losses using standardactuarial techniques.

With HPA, actuaries can take all the data points available fromprevious events and begin to create more robust models. Inparticular, modelers can incorporate information available after anevent from weather and geological services that denote storm paths,wind speeds and Richter-scale readings and then overlay it ontodata for the properties they insure in other parts of the countrythat might experience similar catastrophes. This can give actuariesa better idea of how to price and who to cover.

WHAT VALUE ANALYTICS?

When I talk with insurers about HPA, I often run acrossexecutives who understand they have access to all this data butask, “So what? How does running analytics faster affect the bottomline?” HPA not only provides enhanced analytical-model performanceby eliminating sampling, but it also gives you back somethingprecious: time.

Hours spent prepping and loading data decline to minutes; weeksto days. One of my clients talks about how it takes six days toanalyze one day's worth of data. What would happen if you could doit in six minutes?