Over the past decade, the use ofpredictive modeling in commercial lines has gone from a powerfultool leveraged by a few strategic early adopters, to table stakesfor most insurance companies. The industry has seen a tremendousevolution in the use of advanced analytics and predictive models inthe underwriting process. On the claim predictive modeling side, weappear to be moving rapidly out of the early adopter stage,partially because of the advent of Software as a Service (SaaS) andcloud computing.

Unlike the early days of underwriting predictive models, claimpredictive models can now leverage the cloud as a way of bringinganalytics to life faster for insurance companies and self-insuredentities that don't have the large scale or resources of somebigger companies. As a result of the changes in technology, manyclaim organizations are wrestling with the “how” and “when”elements of the deployment equation, versus the historical “if”side.

The Evolution of Claim Predictive Modeling

Deloitte Consulting LLP has been engaged with the development ofadvanced analytics capabilities for more than 15 years, beginningwith insurance underwriting, and moving on to claim management in2006. Deloitte's advanced analytics efforts in claims started withthe workers' compensation line of business, developing separatefirst notice of loss (FNOL) models and subsequent monthly modelsfor lost time indemnity, lost time medical, and medical onlyclaims.

The firm's claim modeling then expanded to personal lines bodilyinjury, commercial auto bodily injury, and general liability. Foreach of these lines, there are a variety of ways to turn modelinginsights into actionable business results. For example, usebusiness rules to drive resource allocation based on factors suchas the potential claim severity and duration, propensity for fraudor litigation, subrogation opportunity, or other businessapplications.

Similar to the early adopters on the underwriting side, theearly claim adopters have primarily chosen to develop their ownin-house customized models and scoring engine capabilities.Although a considerable investment of time, internal resources, andcapital is needed, the ultimate financial benefits of claim predictive modeling have been impressive.Early adopters of claim predictive analytics are achieving four toeight percent annual reductions in their loss ratios. These savingshave been achieved in part by the redeployment of supervisoryresources and claim management improvements by nurses and SIUreferrals.

The Analytics Discussion

The conversation surrounding claim predictive analytics hasquickly moved beyond if it is a viable solution, to how anorganization can rapidly deploy these types of solutions in orderto reap the savings benefits.

One significant lesson for organizations who have alreadyimplemented claim predictive models is that much of the data needed todevelop significant segmentation power is actually available veryearly in the life cycle of a claim.

Even at FNOL, when much of the claim information is stillunknown, the segmentation power of the models is noteworthy. Forthe loss time indemnity models and loss time medical models that wehave developed at FNOL for workers' comp clients, the average claimseverity for the best ten percent of claims is often over 35percent better than average. For the worst ten percent of claims,the average claim severity is often over 40-percent worse thanaverage. For medical-only models at FNOL, the average claimseverity for the best and worst risks tighten by about ten to 15points on each end. At one month, with additional information, thesegmentation power of the models more than doubles for the worstten percent of claims, with loss time indemnity showing averageclaim severity over 100 percent worse than average.

For commercial auto and personal auto bodily injury models thatDeloitte has developed for its clients at FNOL, the average claimseverity for the best ten percent of claims is usually in the rangeof 50 percent better than average. For the worst ten percent ofclaims, the average claim severity is frequently over 200 percentworse than average at FNOL.

Although it takes longer to retrieve information fromthird-party claimants who may not have much incentive to divulgeadditional information in a timely fashion, the segmentation powerof the models is still significant when external data is leveragedto supplement the traditional internal data available at FNOL. Soagain, it is no longer a matter of if the models work, buthow an organization can leverage the model's segmentationpower to develop a competitive advantage in the marketplace.

Claims on a Cloud

In the current economic situation in which top line growth iscompressed and claim organizations are under the microscope to runas efficiently as possible, it takes a very compelling businesscase to launch a major analytics or predictive modeling effort. Theimprovements in claim predictive modeling have also come at a timewhen many claim organizations are already looking to upgrade orreplace their core claim management capability. As a result ofthese efforts, claim organizations have allocated a large part oftheir budgets and IT resources to these types of initiatives.

So how can claim organizations take advantage of advanced claimanalytics opportunities when their IT resources and budgets areconstrained? How can smaller companies reap the benefits and pursueleading capabilities being derived from advanced analyticsinitiatives? How can claim organizations on a smaller scale andwith smaller claim volumes stay in the game? Now there are viablealternatives, no matter the size of the organization.

Deloitte worked with Experian to develop a set of claimpredictive analytics offerings that include a hosted solution forinsurance companies and self-insureds. The firm says this solutionis designed to bring the best of two worlds together—a tested,production-based suite of claim predictive models hosted in asecure, technical environment. This private cloud setting can offermany of the same benefits of a custom-built, stand-alone solution,with five distinct potential advantages:

• Deployment to results cycle time is reduced from years tomonths. This can in turn translate to a faster ROI for claimorganizations.
• Costs can be transformed from a heavy, front-end investment toa more predictable, per-transaction basis or a 'pay as you go'model.
• Technical resource and integration requirements are virtuallyeliminated.
• Implementation, data, and deployment risks are reduced andshared without compromising security.
• The available data universe may be much larger and more diversethan that of a stand-alone, custom-built implementation.

The basic mechanics of how claim predictive models function inthis hosted environment begin with collecting a limited ,but highlyprescriptive set of initial data at FNOL (and/or designated pointsthereafter). This information typically consists of data capturedat FNOL, and from other basic sources such aspolicy/insured/claimant data and external third-party sources.

This captured data is sent in real-time to the hostedenvironment for analyses, comparison, and scoring. The scoreincorporates knowledge from a much larger and more comprehensivedatabase than most organizations have access to, thereby increasingcredibility. A score and the associated reason codes explaining thedrivers behind the score are produced and returned quickly to theinsurance company or the self-insured's claim environment forimmediate use by the claim handler.

The model output received from the hosted environment can eitherbe referenced independently, or integrated into the organization'score claim business processes and work flow. Re-submission andre-scoring of claims based on pre-defined claim events, scoringtriggers, and as the claim lifecycle evolves can also be designedinto the iterative process.

Clearly, there are both functional and technical optionsavailable for insurance companies and self-insureds to enable claimpredictive modeling capabilities without burdening or relyingsolely on their IT departments. This is excellent news for claimexecutives who have been struggling for years to find the rightbalance between resource allocation, straight-through processingopportunities, and identifying the claims with the largest losspotential before they actually become large losses.

A Competitive Advantage

Organizations willing to implement claim advanced analytics “onthe cloud” are aiming to develop a competitive advantage over thosewho may not have the information technology bandwidth or financialresources to turn competitive modeling insight into actionablebusiness results. As Gartner shared at its 2010 GartnerSymposium/ITxpo, their analysts believe that cloud computing andadvanced analytics are the top two strategic technologies andtrends for most organizations this year.

As insurance companies and self-insured organizationssuccessfully leverage both strategies to improve the claimmanagement process from assignment to resolution, the savings interms of improved loss costs and reduced expenses, fraudmitigation, and resource optimization can be tremendous. Moreimportantly, providing analytics-based insights into the claimresolution process is ultimately the better solution for injuredworkers and claimants. By identifying the right claim and the rightresource at the right time, companies will be better able to reachthe ultimate goal of helping an injured person get back on his orher feet as soon as possible.

Jim Kinzie CPCU, ARM, is a senior manager at DeloitteConsulting LLP in Irving, Texas and a leader in Deloitte's claimand risk management practice.

Kevin M. Bingham is a principal at Deloitte Consulting LLPin Hartford, Conn. He is a leader of Deloitte's claim predictivemodeling practice and chairperson of the American Academy ofActuaries Medical Malpractice Subcommittee.