In 2007, the average temperature in the U.S. in was 54.2 degrees. The average household spent $1,300 for energy. The average increase for health insurance premiums between 2007 and 2008 was 5%, according to the 2009 Employer Health Benefits Survey from Kaiser and the Health Research & Educational Trust.

What do these numbers have in common? Well, for one thing, they're averages, which means they represent lots of numbers that can be significantly different from the average.

Further, the separate numbers that are averaged together each can have very different factors that contribute to their individual values - such as the weather in Alaska vs. Texas, electricity costs in San Francisco vs. Des Moines, or health care cost increases for a unionized manufacturing company in Baton Rouge vs. a software company in Cleveland.

Averages create a blur, a number that in many cases is overly general and not well-defined. Even when some facts are known about a statistical average, it often has little real-world value. For example, it's highly doubtful that residents of southern Florida would make clothing choices based on the average U.S. annual temperature.

With that in mind, looking more closely at the reported average increase in insurance premiums, one finds that four of 10 surveyed employers had increases more than 120% above or below the average.

This is a wide variance, yet many employers will be quick to compare their health insurance cost increases or decreases to an average statistic without questioning its underlying definition or relevance. They will judge their performance (or their renewal rate) good or bad, depending on whether they are below or above this mark, with no knowledge about its comparative value.

Absurdity of benchmarks

In the benefits world, employers come across average benchmark statistics over and over again. Health plan reporting often contains book-of-business benchmarks for data such as in-patient visits, emergency room use, number of prescriptions, and other utilization and cost variables. So, your insurer is comparing your health plan performance to the average of every company to whom it sells insurance.

If this strikes you as absurd, it should.

A statistical sound-off such as, "The average health plan cost increase for 2009 was ..." should send your mind racing for relevance: Does this reflect data before or after employers made plan design cutbacks and/or employee contribution increases? Is this average for all industries or your industry? Your company size and geographic location? For self-insured or fully insured firms, or both? For PPOs? HMOs? CDHPs? Plan designs like yours? With employee contributions similar to yours? What about comparable workforce demographics and health status?

And to cut to what really matters for most employers: Does the benchmark offer a comparison to companies with whom I compete locally for employees? Do the benchmarked companies have the same or different total compensation profile as my organization? The same expense structure? Similar or dissimilar wage-to-benefits ratios?

If you're thinking that this pretty much disqualifies most benefits benchmarks you've seen, you're right, it does.

Here's a rule of thumb for gauging the relevance of a statistical average for your personal circumstances: The more the individual values that make up the average differ from your situation, the less likely the average will have meaning for you.

Scoring benefits

To score your benefits, first determine your financial and labor market constraints and priorities in offering benefits, regardless of what the benchmarks say. If you're not in a sustainable position now, where do you want to be? And what is your plan for getting there? Over what time period?

Second, benchmark against yourself and your benefits goals for cost-containment, employee satisfaction, recruitment/retention or other factors.

Third, if you must use external benchmarks, use them with caution and clearly identify where comparison validity starts and ends. The best benchmarks to use are those that match your most important comparison attributes.

Benchmarks that are relevant to your situation and provide a range of performance comparison data are most useful. For example, data by quartile from high- to low- performing companies is much more informative than a simple average. This lets you compare yourself to more than middling. After all, most of us would not be content, on average, with our head in an oven and our feet in ice.

Neil Quinn is vice president and director, health management services, with Oswald Companies in Cleveland, Ohio. He can be reached at nquinn@oswaldcompanies.com.