For those of us who have been in the claim business a decade ortwo, we have witnessed the introduction of technologies that havehad a profound impact on managing claims. Every few years, apromising technology packaged in a practical claim applicationcomes along, and rapidly becomes an industry-adopted “bestpractice.”

These new waves of technology not only alter how claims aremanaged, but permanently impact insurers' bottom-line results.Telephonic loss reporting, medical bill processing, and bodilyinjury evaluation tools are just a few of many examples.

We are now witnessing predictive analytic technologies becomingthe latest breakthrough technology. Can insurance companies harnessits power? Furthermore, what do claim executives need to know inorder to fully benefit from its potential?

The Power of Predicting

An advanced form of “data mining,” predictive analytics appliessophisticated analysis techniques to enterprise data resulting inthe discovery of meaningful patterns and relationships in data.Identification of those data patterns leads to actionabledecision-making information and, in turn, become a self-learningset of rules that impact how a company conducts business.

Industries best served by predictive analytics have largequantities of data that hold the key to future customer behavior.Unlocking that information can help a company target the rightproduct at the right price using the right business channel. Bankswere early adopters, as in the case of Capital One. In themid-1990s, the company revolutionized the credit-card industry bybuilding its entire business around intelligence that was containedand mined from customer data. Using predictive modeling, thecompany was able to target distinct segments of the market withunique products that significantly helped increase Capital One'srevenues.

The gaming industry also has discovered the value of predictiveanalytics. Ten years ago, Harrah's Casinos discovered that their“high rollers” accounted for about one-fifth of corporate revenues,yet incentive programs skewed highly towards these high-spendingcustomers. As a result of implementing analytics, Harrah's alteredtheir reward programs to non-high-roller clientele; a clearmajority of their revenue base.

The insurance industry has collected and accumulated massiveamounts of data on customers, policies, and claims. But it has beenslow to embrace the use of predictive analytics throughout theenterprise. Actuaries have used various forms of predictive modelsfor years for pricing purposes. Underwriters at most companies arerapidly implementing predictive models based on credit scores orother quantitative methodologies. But the most wide-ranging andpowerful application of analytics in an insurance company is inclaims, especially given some of the historical challenges ofworking with claim data.

Legacy Issues

In the 1990s, insurance companies began to realize that therewas enormous untapped value embedded in their massive stores ofdata. In order to make some use of this data, insurers first had tobe reassemble it into a usable format. As a result, datawarehousing projects were instituted throughout the industry ascompanies attempted to consolidate decades of policy, claim, andcustomer information into a single physical location and format.Many of these projects collapsed under the enormity and complexityof the scope. For those who did create a usable data warehouse,there remained the problem of extracting actionable businessintelligence it.

Despite these challenges, the urgency to find ways to leveragetechnology for better business decisions has increased. In 2006, aGartner study found that chief executives expected their chiefinformation officers to move beyond concerns focused solely on theIT department and to increase their emphasis on helping thebusiness units grow revenue, reduce costs, and improveprofitability. With that directive in mind, a recent survey foundthat the top technology priority of CIOs is “business processimprovement through business intelligence applications.”Translation: CIOs are looking to improve a company's competitiveadvantage through the strategic and innovative use of information,business processes, and intelligence in products and services. Thechallenge is how to apply this in practical ways, and to do that,you must understand the different categories of claim data.

Structured vs. Unstructured

Structured data is standardized, easily entered and handledinformation, such as numbers or company-designated codes forfinancial information, lines-of-business abbreviations, causes oflosses, and the like. Structured data fields have a defined rangeof values and a fixed length. This type of data rarely includesdetails of the claim and what was discovered during theinvestigative process.

Semi-structured data has some structured characteristics, butthe data may be irregular or incomplete. It doesn't necessarilyconform to a fixed schema. Web-related information, such as HTMLformatting and markup, and fields where a brief description of lossor a street address may be entered, are examples of semi-structuredmaterial. Traditional predictive analytic programs usually provideinconsistent interpretations of this data, at best.

Unstructured data is the non-standardized, freeform, explanatoryinformation. It exists in a majority of corporate data repositoriestoday. Examples of unstructured data in the insurance industryinclude adjuster notes, imaged documents, and e-mail messages. Thistype of data is a goldmine of rich, but hidden, information.Unlocking it and extracting its actionable business value has beena daunting necessity. Historically, companies have tried to gainaccess to this data in manual-intensive ways, extracting hard copyfiles from departments and having teams of individuals review themfor specific information. This is an extremely time-consuming andexpensive task that leads to minimal actionable results.

Unique Characteristics of Claim Data

Claim data is unique in many ways when compared to enterprisedata gathered elsewhere in an insurance company. What is thebiggest differentiator? The proportion of unstructured data storedin claim systems. Research groups have estimated that up to 80percent of a company's data is unstructured, and therefore,inaccessible or extremely difficult to gain access to.

As staggering as the 80 percent figure seems, other independentresearch on P&C claim data found that more than 97 percent ofdata in claim systems is unstructured. It's not only the percentagethat is stunning, it's also the fact that the richest informationresides in the unique set of words, acronyms, and abbreviationsthat comprise the adjuster's notes. But adjuster notes are just apart of a claim department's unstructured data inventory. Therealso are e-mails with attachments, imaged documents, case managernotes, web-based information, recorded statement transcriptions (ordigital audio files), and digital photos. All of this valuable datacontent currently exists within insurance companies' datainventories. Yet it's safe to say that MIS reports available toclaim executives are extracted exclusively from the structured datafields that only comprise a sliver of the availableinformation.

Figuring out how to use technology to read and interpretunstructured data has always been the Achilles' heel of analyticsefforts in the claim area. The technical and scientific skillneeded for any individual insurance company to accomplish this hasbeen overwhelming. Some traditional text-mining technologies usegrammatical parsing as a main technique, but it requires correctspelling, no abbreviations or acronyms, proper punctuation, andcomplete sentence structure, all of which are rarely excluded inmost claim files.

Applications for the Claim Executive

There is an explosion in the number of applications that arebeing developed and implemented by adopters of analytics technologyin claims. The predictive nature of analytics is potent. Theseapplications can improve the management of claims and loss costs inthree general categories: mitigation, avoidance, and recovery. Byunderstanding the underlying patterns in large and troublesomeclaims and identifying these types of claims weeks or monthsearlier, insurers can dictate a more effective triage of the claim,and consider more case strategies to influence the outcome.

Which claims are most likely to evolve into large, complexcases? Which claims are likely to show a reserve stair-steppingpattern over the life of the claim? Which claims will involvelitigation, and what will be the nature of the litigation? Whatmedical-only claims have the greatest propensity to eventuallyconvert to lost-time claims? What claims are demonstrating enoughsuspicious characteristics to initiate an SIU referral? Thesequestions are just the tip of the iceberg of solutions provided bypredictive analytics technologies when properly and expertlyapplied. The number of new applications is increasing at anincredible rate.

Not only can analytic applications be predictive in nature, butthey can have powerful discovery and forensic analyticapplications, too. Simply put, discovery-based analytics can helpshed more light on what happened, in ways that structured dataanalyses cannot provide. For instance, how many claims involved anincidence of road rage? How many claims have been handled where aparticular auto body shop was utilized? How many claims have beenpaid that may have been caused by a laptop battery fire? There areinfinite queries of this type that can be solved in real-time.

Forensic-based applications clarify why something happened. Thefocus is on seeing trends and deviations that are difficult todetect. For instance, is there an uptick in claims that resultedfrom a flaw of a particular brand of household appliance,indicating a previously undetected manufacturer's defect? Is therea disproportionate spike in claims where a specific partnering ofproviders has occurred (such as a particular chiropractor andlawyer frequently working together)? What will be the next asbestosor lead paint type of exposure that could disrupt an insurer'sprofitability or viability? Which insurers will utilize forensicanalytics to notice this trend in their claims first, and tightentheir coverage exclusions before their competitors?

Total Loss Cost Management

Predictive analytics technology is a potent fit within the claimfunction and particularly effective in managing loss costs. Toharness the full power of analytics, companies must ensure that allof their claim data — structured, semi-structured, and unstructured— can be accessed, read, and analyzed. For too long, marketpressures have short-changed claim executives by not providing themthe right kind of actionable business intelligence. Now, predictivetechnologies have been adapted for the claim environment. As thetechnology is being rapidly deployed, claim executives areachieving unprecedented new levels of total loss cost management,similar to other waves of groundbreaking technologies deployed overthe past several years.

Stephen Holcomb is the founder, president, and CEO of FullCapture Solutions and can be contacted [email protected].