Measuring performance is vital to any management process. Since the ultimate costs and results of an insurance program may not be known for years, many risk managers and insurance professionals prefer to compare the performance of their program to industry or historical internal benchmarks. However, even answering the simple question of “How does my program rank compared to others and the past?” can involve a certain complexity that may not be immediately recognized.
A well-engineered benchmark will showcase the value that comes from a sound risk management program and detect areas where additional risk management contribution could enhance the position of the organization or lower insurance costs. Although each insurance program is unique, there is a time-tested approach to benchmarking that can help accomplish these goals.
Identifying the Purpose
Management may have any number of goals in conducting a benchmarking exercise. They could want to evaluate risk management performance or control losses. They may wish to understand how divisions or programs compare internally or to industry standards. They might hope to connect data that can be used to generate additional business intelligence. Or their main desire could be to compare savings from a new safety initiative as part of a cost-benefit analysis.
Regardless, by defining its purpose at the outset, the benchmark will be more focused and efficient. The purpose should be discussed among benchmark users but may also be shared with those being measured, such as plant or factory managers, for example.
It will also help if the benchmark is relatively easy to calculate and comprehend. While a more complex benchmark may provide better details and information, the extra time and costs may not be incrementally beneficial. Management may also benefit from the independence of an outside estimate depending on the purpose of the study and if potential conflicts exist.
Determining the Style
With all the resources available to risk managers today, it may be overwhelming to figure out where to start looking for information. Some popular sources respected for providing accurate industry information include the Insurance Services Office, the National Council on Compensation Insurance, the Workers Compensation Research Institute, the Reinsurance Association of America, the National Association of Insurance Commissioners, A.M. Best Company, state insurance commission studies, and various federal government agencies.
Published industry studies are also available. Most are by industry trade associations, academic institutions, and service providers—consulting firms, actuaries, brokers, third-party administrators (TPAs)—which often specialize in a particular niche. Service providers often aggregate data, which then can be used to compare an individual program’s results against the whole.
Users should be cognizant of the data sources underlying any study, however, recognizing that the study may not be fully credible because of sample size or perhaps may be inconsistent with the exposures of the user’s own program. For example, data may range from large insurance companies to small self-insureds, or may only contain certain coverages or types of losses. Some organizations require membership or may charge for information. Therefore, it may be difficult to navigate and obtain the information, especially for those unfamiliar with the source or when data is not in a user-friendly format.
Another route may be for a program to benchmark against itself. Management can compare current or expected results with historical results, thus tracking change over a period of time. If a program is large enough or has multiple divisions, management could compare internal divisions. Possible choices for an internal benchmark base include the program’s largest division, oldest division, main headquarters or maybe even the division where the safety record is the best. An internal benchmark has the advantage of data being measured on a consistent basis and may require fewer adjustments for comparing divisions on an apples-to-apples basis. Also, management already possesses the data, which eliminates any external collection costs or complications.
Sample Frequency and Severity Measurements
Two of the most commonly used benchmarking metrics are frequency and severity. Both offer an effective starting point to determine the true risk characteristics of a particular insurance program.
Sample Frequency Measurements
- Average frequency (number of claims/number of exposures)
- Percentage of closed known claims (number of closed claims/number of reported claims)
- Percentage of closed claims with payments (number of closed claims with payments/number of total closed claims)
- Percentage of closed claims without payments (number of closed claims without payments/number of total closed claims)
Sample Severity Measurements
- Average severity (losses/number of claims)
- Average case outstanding (case reserves/number of open claims)
- Average paid on closed claims with payment (paid on closed claims with payment/number of closed claims with payment)
- Average paid on closed claims (paid on closed claims /total number of closed claims)
- Average reported (reported losses/number of reported claims)
- Costs per exposure unit (losses/number of exposures)
- Percentage of paid losses on known claims (paid losses/reported losses)
Tailoring Statistics to the Distinct Exposure
When divisions are not equivalent or when benchmarking against an industry source, management should consider several modifications to the base statistics. Due to differences in state benefits, rules, loss development and settlement practices, adjustments should be made for industry type, class (e.g., factory vs. office worker), retention limits and jurisdiction. Statistics may also need to be adjusted for trends when comparing across years.
Other modifications may be based on the varying practices of service vendors such as the reserving and settlement patterns of a TPA, or even potential economic conditions that may drive the number of filed claims.
Management should verify that losses are coded and categorized on a consistent basis between sources. While it may be difficult to determine all data adjustments, the goal is to compare a similar insurance product consistent to the exposure of the program.
No matter which benchmark statistics management ultimately chooses to monitor, it will need to make decisions about the estimation of each. First, management will need to determine how detailed it would like the benchmarks to be, such as by division, line of business, cause or size of loss. Of course, breaking down data to extremely fine levels may result in a reduction of data points and loss of credibility.
Second, management will need to select the number of loss years to include for the right balance between responsiveness and stability. Typically, the more years included the more stable the results; however, if the goal is to be more responsive to current activity, then management may choose to only include more recent years. Additionally, management may also decide to cap the size of the claims to limit the influence of a large, random loss and to be more responsive to “working layer” losses.
Finally, management should decide how often the results will be updated. It may not be feasible to have all statistics calculated with each evaluation. Evaluations may be quarterly, semiannual, annual or even less frequent as influenced by claim activity or new risk management initiatives.
Selecting Benchmark Measurements
Selecting the right benchmark statistics and measurements to capture a program’s true risk characteristics and meet management’s goals is critical. Measurements may be completed in total, by division, line of coverage, cause of loss, claim status, size of loss or any combination thereof.
One of the first questions to be addressed within a sound risk management program concerns how often claims are occurring or being reported. Frequency may be measured purely as the number of reported claims by policy period but would be better if normalized on a per-exposure basis. In addition, it should be recognized that, depending on the coverage type and maturity of the claims, reports of future claims for a policy year may occur. Certain count-based measurements may reveal information about claim settlement practices.
Complementing frequency is severity, which answers the question about the size of an average loss. Similar to frequency, losses may be developed to measure an ultimate cost or may be monitored in the interim. Some dollar-based measurements may also reveal information about claim settlement and reserving practices. (See “Sample Frequency and Severity Measurements” above.)
While frequency and severity statistics may be deemed straightforward and likely to be calculated by management on its own, some information may require outside assistance. Advanced statistics such as closing lags, reporting and payment patterns make use of loss development patterns that may need to be estimated by a professional. These patterns are also likely to be incorporated in the pricing or reserving of the insurance program. Along with development, frequency and severity trends are used to forecast losses and can be easily estimated for an individual insurance program and benchmarked against industry trends.
Management may also opt to examine both traditional and more creative, nontraditional statistics. For example, an expected loss ratio or cost per exposure unit, such as payroll, will show management results of the program, while the ratio of the amount of indemnity dollars spent compared to allocated loss adjustment expenses spent may show the effectiveness of defending and settling claims. Innovative statistics can be designed to monitor any current or potential concerns.
Interpreting the Results
Sorting information and interpreting results may seem daunting and puzzling at first, especially if actual results do not support initial intuitive expectations. A common question is how much to rely on this information. The answer depends on how much management trusts both its own data and the data of the benchmarked source.
When a trusted source is used, with all the data adjustments made to put the program on comparable levels, management should feel more at ease. Consistent results may also confirm benchmark credibility.
Management may decide up-front how the results will be evaluated before estimating benchmark statistics. Is there a particular threshold the results need to meet? If so, what is an acceptable tolerance level knowing that losses are still likely even with a successful safety program? If a single cutoff standard is not ideal, a letter grading scale (such as A, B, C) might be implemented for the purpose of creating bands and smoothing the results. The scale should be agreed upon before the study begins, but might need to be recalibrated after results are received.
When developing a benchmark system, it is helpful to keep in mind that most benchmark results are typically not an “all-or-nothing” game. In fact, some losses take many years to develop. Depending on the snapshot of losses used, the same view a year later may portray a different story. If losses change significantly between evaluations, management may choose to complete the study more routinely or incorporate a longer experience period or utilize caps on losses.
Creating Value From a Benchmarking Study
A benchmark study is a valuable tool in a risk manager’s arsenal that can provide instant and proven benefits. By exposing the loss drivers, risk management will be able to prioritize short- and long-term goals as well as identify which types of claims or divisions to focus on first in order to have the most impact, promote greater safety and reduce costs. This detailed information will also serve as a basis for additional business intelligence about the program.
Even initial adverse results can create an opportunity by providing concrete information to support a request for additional funds to invest in risk management with the end goal of reducing total future losses. A cost-benefit analysis of a new safety initiative will determine how much the benchmark results improve.
While monitoring the cost per unit is important, risk management should also keep the total dollars in mind. Risk managers may want to share study results to revitalize emphasis in the insurance program among divisions, internal management, brokers and commercial insurers, which would each utilize the results differently. In addition, risk management may seek to use the benchmark results to achieve lower excess or reinsurance premiums, display significance to and connect with executive level management, and, most importantly, highlight and boost the value of their work.
Effectively administering and customizing the structure of a benchmark based on the program’s exposure adds strength to the process. From compiling consistent data to estimating program-specific statistics, interpreting results and investigating loss drivers, a thorough benchmark study may require a substantial amount of time.
Use of an outside source might be ideal for those with limited time or those looking for expertise or a layer of independence, particularly when the end user of the study is an investor or high-level executive. Devoting an individual to monitoring the benchmark results might also ensure that the study remains a priority and retains consistency across multiple studies, which is key to tracking the results over time.
In the end, a properly conducted benchmark study will finally answer the question of “How does your insurance program compare?”