It is estimated that the world generated 1,200 billion gigabyteslast year alone. How does one draw insights from such a vast amountof data to make actionable decisions? Can insurers use thisoverwhelming amount of data to improve their overall profitabilityand efficiency?

Availability of large volumes of consumer behavioral data,advances in device and sensor technologies, improvements inpredictive analytics and simulation techniques, and methods forvisualizing large volumes of data are revolutionizing how P&C insurers consume large sources ofinternal and external data to and make business decisions.

Let's look at the drivers of predictive analytics and howinsurers can use certain techniques to maximize the benefits ofthis technology.

Why Now?

Focusing on future projections is at the heart of everyinsurer's business. Whether they are P&C or life and health(L&H), they take cash from their customers today and make apromise to pay for certain eventualities or future conditions. The"future" varies from a couple of years for personal lines insuranceto 30 years or more for life insurance and annuities. Insurers'actuarial and underwriting divisions have always used predictivemodeling in one form or another. However, predictive analytics haverecently gained ground in claims, as well as marketing, sales anddistribution.

A number of advances over the past decade have made it not onlypossible but essential for insurers to explore broader applicationsof predictive analytics:

• Accelerating technology and consumer adoption.Advances in sensor, computing, and communication technologies areadding to the volume of data and enabling the analysis,interpretation, and visualization of it. The rapid growth ofmobility and mobile data use, coupled with social networking, hasresulted in exponential growth of behavioral data. The "Internet ofthings," which makes devices from automobiles and plant machineryto dishwashers and washing machines relay real-time data,substantially increases the amount of information generated. Smartsensors that cost just a few dollars can measure a singleattribute, such as temperature or moisture, and communicate thereadings on a real-time basis. As these technologies improve andtheir use accelerates, they will increase the amount of availableinformation.

• Increased data availability. Consumer adoptionof rapidly evolving technology and automation has resulted in thegeneration of billions of gigabytes of data by insurers, governmentorganizations, regulators, non-profit organizations, ratingagencies, and independent data aggregators. By some estimates, theproduction of data has increased from 150 billion gigabytes in 2005to 1,200 billion gigabytes in 2010. As insurers have startedintegrating their legacy systems, externalizing data intoenterprise data warehouses, and developing a single view of thecustomer, their internal company data sources have significantlyimproved in terms of quality and quantity. Moreover, external datasources have become more standardized, allowing for greater sharingof data. Non-profit data aggregators and for-profit data providershave facilitated the flow of information between differentparties.

• Sophisticated analytical techniques. The needto generate faster and better insights from increasing multi-mediadata is resulting in fresh techniques for analyzing text, speech,video, and sentiments. Analysis of online interactions,unstructured text, speech, video, and social data mining have allemerged as distinct areas of focus.

Value to Insurers

Professionals are using predictive analytics in underwriting,pricing, marketing, sales, distribution, customer service, claims,reserving, and hedging. While many personal lines carriers haveseen lower profitability as a result of price-based competition,early adopters of predictive analytics in the personal lines sectorhave managed to price risk better and attract more profitablecustomers. As a result, companies that are not consideringimplementing some kind of predictive analytics will be at adisadvantage.

Generally speaking, there are four application areas forpredictive analytics:

1. Claims management. According to the InsuranceInformation Institute, fraud—most commonly, staged accidents andclaims padding—costs P&C insurers more than $30 billionannually. By analyzing historical claims information anddemographic profiles, predictive models can identify potentialfraudulent cases for further investigation. This allows claimsadjusters to focus on suspicious cases and conduct more detailedinvestigations. Predictive analytics can also reduce losses. Byanalyzing the types of claims, predictive models can flag casesthat might be subject to litigation. Routing such claims throughspecialist adjusters and streamlining the process can helpadjusters reduce litigation costs. As a result, predictiveanalytics can contribute to reduced fraud costs, reduced lossadjustment expenses, improved adjuster productivity, and reduce theoverall claims ratio.  
2. Demand management. Insurers use a multitude ofdistribution channels to sell their products, and the sellingprocess takes place over multiple channels, including in person,over the phone, and online. Because insurers face increasingpressure to produce better returns on their marketing investments,they are now using predictive analytics to analyze consumerbehavior, which helps calculate their propensity to purchasespecific products. Insurers can collect policyholders' data overtime to determine individual policyholder receptiveness tocross-selling other products and when it is appropriate. In doingso, they can see increases in conversion, cross-product, andretention ratios.
3. Producer acquisition and value management. Theaverage age of an insurance agent is 57. As retirements reduce thenumber of agents and advisors and economic growth remains low,acquiring, retaining, and enhancing producer productivity hasbecome an even greater priority. Predictive modeling can combineinternal insurer data with external socio-demographic data todetermine the market potential for specific products. Theseinsights can help the head office sales force improve produceracquisition, retention, and productivity ratios.
4. Underwriting and pricing. Actuaries andunderwriters have typically used predictive modeling to computerisk scores based on things like an individual'ssocio-demographics, driving record and behavior, and credit score.They use these predictive scores to determine pricing, as well asautomate the underwriting process by setting rates andautomatically approving customers for coverage beyond that line.Similarly, they can automatically reject customers who fall below acertain threshold, leaving underwriters to manually evaluate asmaller set of customers. Auto insurers have long used suchtechniques, but now property, commercial, and life insurers are aswell. Predictive modeling plays a critical role in reducingunderwriting cycle time, enhancing the ease of business for agents,increasing underwriting consistency, and reducing underwritingcosts, all of which lead to better risk pricing. The positiveresults for insurers include reduced expense ratios, betterunderwriting results, and enhanced customer and agentsatisfaction.

Getting Started

Implementing a predictive analytics program can be daunting,considering the variety of possible applications, the volumes ofdata that need analysis, and the sophistication of availabletechniques and tools. Accordingly, a four-stage approach can beemployed to achieve results relatively quickly:

1. Select a business problem and map the decisionprocess. Insurers often undertake large data warehousingor customer data integration initiatives to justify a predictiveanalytics program. However, it is better to begin by identifying atangible goal and desired metrics, as well as the criticaldecisions it has to make to realize them. This top-down approach,which starts by determining the data and insights the company needsand what it has to do to get them, is better than the typicalbottom-up approach of agglomerating data in one place beforedetermining what to do with it.
2. Identify, collect, and analyze data (datamining). Once the insurer has identified its goals and thenecessary steps to achieving them, the company should identify theinternal and external data sources that can be used to generatenecessary insight. 
3. Build and test the predictive model. Once thecompany collects and analyzes internal and external data, it canbuild a predictive model that uses a number of differenttechniques, including regression, simulation modeling, neuralnetworks, and evolutionary computation.
4. Institutionalize predictive analytics and aninsight-driven culture. By demonstrating the value ofpredictive modeling in a tangible business application, an insurercan undertake the more difficult task of changing how managementand staff make decisions by promoting predictive analyticsthroughout the company.

Predictive analytics will be increasingly important forinsurers, and their claims organizations specifically, who want tostreamline their operations and use more insightful data to makefaster and better decisions. Insurers that consistently base theiracquisition, retention, and management of their customers andagents on predictive models will have an advantage over their peers in targeting, pricing, service, andclaims handling.

Anand S. Rao is principal with PwC Diamond AdvisoryServices.