Looking further at industry profitability quickly shows that claims, including loss adjusting expense (LAE), represent roughly 80 percent of an insurance company's costs. Expense reductions are unlikely to provide enough savings, likely will impact service, and represent an often irreplaceable loss of expertise.

Rate increases are occurring and will continue, but given intense competition, are likely to be managed rigorously. That leaves claims, a tremendous offset to revenue where the potential for improvement is rapidly gaining attention. Consider the following, many fueled by the difficult economic conditions faced by consumers as well:

• The National Insurance Crime Bureau estimates insurance fraud has escalated by 19 percent since 2009.

• The Coalition Against Insurance Fraud estimates insurance fraud costs Americans $30 billion a year.

• Insurance industry studies indicate 10 percent or more of property & casualty insurance claims are fraudulent.

• One out of every three bodily injury claims from car accidents involves fraud.

• U.S. insurance fraud runs an estimated 10 to 20 percent of premiums; in other markets like Brazil 25 to 30 percent.

• One in 10 Americans says they would commit insurance fraud if they knew they could get away with it.

• Fraud is costing the average household an estimated additional $300 per year

• Missed subrogation opportunities could represent up to $15 billion annually just in the U.S.

Cost of Business

Historically, fraud has been difficult to identify and catch amidst the hundreds of thousands of submitted claims, particularly with the pressures for shorter cycle times, higher customer satisfaction, and lower staff expenses. As a result, it is often considered a cost of doing business to be pursued responsibly.

Given the industry's current financial situation, the best solution to the ROE problem would seem to rest in claims, combining aggressive pursuit of fraud with close operational management. That said, the focus cannot be reducing the number of legitimate claims paid. Claims payments represent the delivery on the promise, an insurer's most important opportunity to achieve customer satisfaction and competitive differentiation.

Fast, easy and efficient should be the mantra of the forward-thinking claims operation. This is especially true in today's world of easily accessible choices and the amplified messaging provided by social media.

Technological advances including predictive analytics and text mining have contributed to an insurer's ability to more efficiently and effectively process claims. More specifically, whether dealing with dishonest body shops, unscrupulous providers, ethically challenged consumers, construction scams driven by the many catastrophic events or well-coordinated, cross-country schemes, the use of predictive analytics to detect and escalate potential fraud is providing significant improvement.

The results include fewer false positives as well. As a contrast, consider how traditional rules-based systems would tag claims that met set criteria for review. Once a “hole” in the rules was found by payment of a fraudulent claim, that “hole” would get aggressively exploited until a new rule was put in place.

Predictive analytics operate much more intelligently. Instead of static rules, claims are evaluated based on all available history for all involved parties and service providers along with the stored outcomes, and then enhanced by correlating millions of similar claims. As the number of claims decisions, involved parties, service providers and characteristics increase, predictive analytics learn and adjust future outcomes.

Another advantage to this automatic adaptation is the ability to retrospectively revalidate claims, potentially identifying previously undetected cases of fraud based on links to the new case. Fraud schemes involving “cash for crash” and intentional collisions are more quickly surfaced via this type of link analysis. Some of the obvious immediate benefits include:

• A more efficient and wider “net” that quickly identifies more fraudulent claims with fewer false positives.

• Faster escalation of questionable claims, reducing often unrecoverable interim payments.

• Dynamic adjustment to new fraud schemes with the ability to retrospectively filter approved claims.

• Focused and more efficient use of SIU expertise, measured by higher savings per salary dollar.

• Lower unrecoverable payments by prioritizing investigations based on probability and financial impact.

• Improved customer satisfaction resulting from significantly faster payment of “clean” claims.

Real Benefits

According to estimates from FICO, the functionality represented by these new tools could find up to 50 percent more fraud than the traditional rules-based tools. Based on the cost of fraud mentioned earlier, and the fact that claims and LAE represent 80 percent of revenue, these kinds of results are clearly a material improvement.

As an added benefit, insurers that have successfully integrated predictive analytics into their claims operation are also seeing an increase in the consistency of how claims are handled—always a challenge in large and/or geographically dispersed centers—and even more notable is an increase in accuracy of loss reserves. Given both the impact of error and the resulting attention given to loss reserving, improved accuracy as a side benefit to claims analytics further strengthens the importance to rapid adoption.

Claims subrogation and recovery represents another area where predictive analytics has been successfully deployed with meaningful positive results. Often, these opportunities are buried under mounds of data and claims payment activities. Equally as likely is when the indicators of opportunity are in handwritten notes in the file.

These challenges are compounded by the complexity and fluid nature of the rules and regulations that guide these efforts. The result is usually either not enough resources to pursue, a missed identification of the opportunity, or a long, drawn-out manual process that ends up taking up more time and effort than the recovery brings in.

By combining predictive analytics with text mining tools, these issues can be greatly reduced and the value of recovery efforts significantly increased. By outlining, and updating as necessary, the rules surrounding subrogation and recovery as well as the triggering conditions, cases fitting recovery parameters are identified early and efficiently for marking or discounting.

Late stage opportunities are found via text mining that searches the specific note fields for trigger words, bringing cases forward that fit the definitions. A feedback loop ensures that both mechanisms—predictive analytics and data mining—constantly update their search criteria based on successful discoveries.

Once discovery is profiled, predictive analytics can provide insights into which path and vendors will likely lead to the quickest recovery of the most funds. Much of the tedious manual labor expended in looking for the opportunities, browsing the notes, determining the applicable options based on regulatory parameters, and selecting the right vendor(s) is eliminated, leading to increased recoveries and reduced total claims costs.

Be Efficient

Improving operational efficiency represents an opportunity for leveraging predictive analytics. The size and cost of the claims department is typically a significant investment in extremely important intellectual capital. Optimizing the use of this expertise will not only increase employee satisfaction but will improve productivity, consistency, and customer satisfaction.

The application of analytics tools for resource scheduling and work allocation has been proven to have a positive measurable impact on cycle time, staff-to-claims ratios, and average staff cost. The predictive element comes into play in extrapolating not only the arrival and complexity pattern for FNOL, but for subsequent documentation, field and customer follow-up contacts, and escalation patterns to name the main key variables.

Staff skill sets, schedules, productivity “sweet spots”, and secondary expertise are then integrated with the projected work patterns to calculate the optimal allocation of individuals and skill sets by time-slice. Practical rules have to be incorporated to prevent every-other-hour schedules, while at the same time allowing for split-shifts and shared part-time as allowed by the company.

While there are a number of call center products that model and monitor call arrival patterns across skill-sets, the intention here is to take that concept and extend it across a broader range of mixed complexity, alternative source work activities. The more complex the model, the more efficient the eventual use of resources; and as an indirect benefit, the more satisfied those same resources, and their customers, are as a result of the closer pairing of time and need to availability and skill.

Additional noteworthy advances are being made in claims analytics tools beyond fraud detection, claims recovery, and schedule optimization. Text mining is one in particular that seems to be gaining the fastest traction. Simply put, text mining involves searching unstructured data like handwritten notes, translating findings into data that can be meaningfully manipulated, and then either operating against the results to generate predictive outcomes or incorporating them into the applicable database.

In March, Text Analytics World was held with an impressive range of topics and real-life applications for both text analytics and the expanding field of sentiment analysis. Specific to insurance claims was Accident Fund's presentation on the use of text analytics to extract information about claimant co-morbidities—particularly obesity and diabetes—from workers' compensation adjuster notes.

Here, the data did not exist in digital form, but it had been proven to correlate to claim outcomes and total costs. Given that the average combined ratio for the workers compensation industry is estimated to reach 120 this year—shored up by a continuing increase in medical claims costs—the ability to leverage text analytics in determining and better managing these claims could prove an extremely valuable application.

Top Challenges

According to a 2011 study by Bloomberg, the number one challenge preventing companies from more fully adopting analytics was executive concern over data quality, acquisition, and integration. Within the insurance industry, it is likely that this challenge is often expressed as the “legacy systems” barrier, that hauntingly familiar roadblock to innovation and transformation. While a valid concern for insurers who have yet to either start or complete their transition to new systems, it is worth noting that analytics tools can operate against data warehouses. And data warehouses can be fed by a multitude of old, new, external, and in transition systems.

Unlike many transformations that are utterly dependent upon a successful migration to a new system or modular replacement of a legacy system, analytics can be implemented across diverse platforms of varying age. A well planned roadmap can be designed to start realizing benefits in as little as six months, as experienced at Chartis in the case study presented at IBM's Financial Services Summit in September of last year. Despite having many business segments and a variety of systems, a repository was used to capture the relevant data, analytics tools were engaged for a very specific purpose to operate on that data, and the project's first of many stages was implemented on time with positive impact from day one. At that same conference similar examples were provided by AIG, Oppenheimer, and Chubb.

The three key criteria in common across the all-senior-executive panel, as well as several other panels and presentations since, are as follows:

Have a focused roadmap that clearly outlines what resources will be involved, specifically where analytics will be used and when, what tools will be used, and how will success be measured. The roadmap may change with time and knowledge, but requires from day one executive sponsorship and business buy-in on who, what, where, when and how.

Use clean data that is comprehensive, accurate, and current; this does not mean that 100 percent of the data has to meet these criteria, as some have used as little as 70 percent of available data. The bottom line is that the data that is used should be representative and should meet these criteria. This is where starting small and building big has the greatest advantage as representative subsets of data can be used with proper thought.

Staff with talented and engaged people who completely understand the business problem and are proficient with analytics. Note that one person does not have to have both qualifications; it can be a team where some are experts on the business problem and others experts in analytics. Both skill sets have to be represented.

According to an IDC study, the average ROI for a predictive analytics project was 145 percent and a nonpredictive analytics project was 89 percent.

According to research by the Aberdeen Group, insurers integrating predictive analytics into their key decision processes achieved a one percent improvement in profit margin and a six percent improvement in year-on-year customer retention versus a two percent drop in profit margins and a one percent decrease in year-on-year customer retention for insurers who did not adopt predictive analytics.

No matter how it is measured or what study is reviewed, the bottom line is that the use of predictive analytics, particularly in the claims area for p&c companies, can provide near-term positive impact despite the challenging economic times. Looking at the field of those already engaged in the process, it is clear that tomorrow's market share leaders are being defined by today's actions.