Take a look at your process metrics. Are you using any mean scores? If so, you might be missing performance problems. More importantly, you are reducing your ability to make decisions and take action.

For example, suppose that your help desk has a goal of getting issues resolved within 90 seconds. It resolves resolving half of the issues in 60 seconds and the other half in 120 seconds. This would look if your target (metric) was set at an average of 90 seconds, yet you’d be be missing your goal fifty percent of the time!

That’s the problem with a simple average. By its very nature, it can mask what is really happening. For example, here are three scenarios in which you could have an average time of 90 seconds per call each of which would require a different action:

• Fifty percent of the calls take 30 seconds, and fifty percent take 150 seconds
• Twenty-five percent of the calls take 60 seconds, fifty percent take 90 seconds, and Twenty-five percent take 120 seconds
• All calls take 90 seconds

In the first case, you probably have some significant and broad reengineering work ahead of you. In the second case, you probably have some focused process improvement. In the third case, you are right where you want to be.

The point is that understanding variance is as important, if not more important, than understanding the mean. This is the basis of quality processes such as Six Sigma. There are a lot of good books and articles written on this subject.

But, there is a simple way to get started – switch to using frequencies instead of means. In the example above, your metric might be that 80% of the issues must be resolved in 90 seconds (you would set the percent based on your business model). This way, you can’t be fooled. You can also set a lower bound frequency – less than 5% of all calls should take more than 90 seconds. While the frequencies won’t tell you the cause of the problem, they will give you information that is more actionable.

For example, when reviewing customer satisfaction data, I used to use three frequencies (the data based on satisfaction surveys where 1 was very dissatisfied and 5 was very satisfied):

• Percent of ones and twos (people who weren’t satisfied
• Percent of fours and fives (people who were satisfied)
• Percent of fives (people who were very satisfied)

I’d look at questions or departments that fell outside the bounds on any of them in the following way:

• First focus on areas that were having more ones and twos – usually major changes were necessary
• Then focus on areas that were having fewer fours and fives combined – usually there were a couple of key issues
• Finally, focus on areas that had fewer fives – usually “polishing” some rough edges was all that was needed

Those three simple frequencies allowed me to triage and prioritize my effort on improving satisfaction. (Of course, your prioritization also depends on impact – it could be that moving the fours to fives has a greater impact than moving the ones and twos.)

A classic question from Statistics 101 is, “Would a six foot tall person always be able to walk across a pond whose average depth is 5 feet without drowning?” Of course not and your business can’t either. Hitting a target that’s based on averages might just mean that you are having average performance.