Predictive analytics for self-insureds

  • Print
  • Connect
  • Email
  • Facebook
  • Twitter
  • LinkedIn
  • Google+
By Elizabeth Bart | 04 November 2013

Predictive analytics is widely used in the insurance industry. Is it time for self-insureds to reap the benefits of predictive analytics and realize significant bottom-line improvements as well?

Self-insureds can use predictive analytics for employee cost benchmarking, early identification of late-developing claims, and budget and allocation decision-making tools. Currently, the key area of focus for self-insured risk managers is claim prevention. But even the best claim prevention methods are not enough to avoid all claims. Claims occur despite a companys and risk managers best efforts. If traditional claim prevention is the only defense against losses, these companies will lag behind their contemporaries who, to keep claim costs down, are already using predictive analytics.   

Primer on predictive analytics

Whether we know it or not, we have all encountered predictive analytics in our personal lives by simply browsing the Internet. Companies like Amazon leverage their immense amount of data to predict customers shopping preferences to drive additional sales. Likewise, the insurance industry, with its substantial volume of data, views predictive analytics as an essential capability. Forward-thinking self-insureds see the benefits of these tools and realize that they should be used to better understand their exposure and better control their losses. 

Predictive analytics can improve both customer satisfaction and company profits. Last week, I bought a webcam from Amazon. The product page displays the specific details on the webcam and shows an assortment of additional items frequently bought by other customers who purchased this webcam. Amazon uses its customer purchase data to predict that webcam purchasers also routinely buy extension cables, microphones, and speakers along with their webcam. Thus, Amazon has effectively applied predictive analytics to identify cross-sell opportunities that benefit its customers (I was glad to be reminded that I would, in fact, need an extension cable) and increase its revenue (I spent an additional $5.99).

Amazons product pages are an easily visible example of predictive analytics. Similar to Amazon, the insurance industry has adapted predictive analytics to not only increase premiums but also to improve risk selection. Risk selectionbeing able to identify the good/profitable risks and the bad/unprofitable risksis so important for insurance companies that the popularity of predictive analytics comes as no surprise. The range of ways insurance companies are using these tools includes evaluating the profitability of their accounts, assisting in effective underwriting and proper pricing, marketing to the appropriate client base, retaining their customers, and estimating the lifetime value of each customer. 

It is important to note that predictive analytics is not just a data summary. Predictive analytics sorts through vast amounts of data to find relationships among variables to predict future outcomes. This data can often be seemingly disparate, or even appear to be unrelated; it also often combines the use of both internal company and data from outside the organization. For example, a commonly cited and successful example of such a relationship involved insurers finding a direct correlation between two variables: credit scores and auto claims. Also, the more data fed into an analysis, the more robust it will be. Self-insureds have quite a bit of employee data that greatly aids analyses predictive abilities. In addition to loss-specific data, payroll, human resources information, and other available third-party data should be utilized to its fullest extent for optimal results. 

Applications for self-insureds

Self-insureds have numerous opportunities to benefit from predictive analytics. Three large workers compensation areas on which predictive analytics sheds light are: 1) identifying which employees cost more than industry and company averages, 2) predicting early on which claims are the most likely to have late-developing costs, and 3) constructing qualitative cost/benefit scenarios to help risk managers allocate their budgets effectively. 

Self-insureds applying predictive analytics to their workers compensation claims, for example, have a number of employee variables to work with such as: employee age, length of employment, state of residence, employment type (full time/part time), salary type (hourly/salary, low wage/high wage), and claim type (indemnity/medical only). Predictive analytics can find relationships that will affect future claim activity on current and future employees. 

Predictive analytics for self-insureds table 1

The real goal of predictive analytics for self-insureds is to help guide and support risk managers decisions. Predictive analytics can be applied to both pre-claim and post-claim loss prevention methods. To aid in claim prevention, pre-claim-focused analyses are used to highlight high-risk (high-cost) employees. The loss costs of various groups are compared to each other, a company average, and to average industry loss costs provided by the National Council on Compensation Insurance (NCCI). Loss categories higher than company or NCCI averages get closer examinations for loss drivers and mitigation strategies. A higher-than-average cost for newly hired employees may signal a need for more training. A higher-than-industry cost for claims in certain states may be noted. A particular type of injury may emerge as the most costly. The potential savings can be estimated as the difference between the current loss costs and the benchmark loss costs, times the percentage of employees or expected claims involved.

For example, a self-insured entity may know that its newly hired employees experience a larger proportion of losses than employees with longer tenure. If it conducts an analysis and discovers that low-wage employees working in Illinois with less than six months of experience have substantially higher costs than the average employee, claim prevention resources could be specifically aimed at that employee demographic to control costs. 

Savings opportunities

A notable benefit of predictive analytics is that it provides quantitative cost-saving information to risk managers. Continuing with the prior example, assume 2,500 employees are newly hired, low-wage employees in Illinois and their average costs have been shown to be three times higher than the company average of USD $1.50/$100 of payroll. We can estimate that a reduction from $4.50 to $1.50 could create $2.25 million in savings. Asking senior management for $100,000 for more new hire training in Illinois facilities will be much easier with the quantitative support provided by predictive analytics.

(2,500 employees with an average payroll of $30,000 save $3 = 2,500 x 30,000/100 x 3 = $2.25M)

Not only can predictive analytics assist with reducing cost pre-claim by focusing on exposure, it can also reduce costs once a claim has occurred. Knowing the easy-to-identify large claims will be second nature to risk managers, however, post-claim predictive analytics can look into claim development details to find characteristics that late-developing, problematic claims (and often not the obvious large ones) have in common. After a loss has occurred, one of the most effective ways to manage costs is to involve a very experienced claims handler as soon as possible. The results of effective post-claim predictive analyses will assist in implementing cost-saving claims triage. Because the best resource post-claim is good claim management, predictive analytics can get late-developing, problematic claims the timely attention they need to contain the ultimate costs or even settle the claim.

Loss savings based on predictive analyses extend beyond claim cost reduction. Being able to quantitatively show potential savings and concrete mitigation plans will make a positive impression on senior company management and excess insurance carriers. Demonstrating shrewd knowledge of the loss drivers and material plans to reduce the losses can aid in premium negotiations with excess carriers for all future policy years. And if the insurer or state is holding any collateral, the predictive analytics' results can be used by the self-insured in negotiating.

The key to unlocking further potential cost savings in your self-insured plan is readily available in your own data. Predictive analytics is the tool that will help risk managers make better claim reduction decisions and produce actionable items with real cost savings now and in the future. Risk managers and self-insured companies can look forward to possible benefits such as loss cost reductions along with reductions in excess premium and collateral, and quantitative information to help them with budgeting and allocation. As more self-insureds begin applying predictive analytics to control costs, companies that are not using these tools will be at a competitive disadvantage.