An Average Can Be Correct And Still Useless

Arithmetic that is beyond dispute can still produce a figure that misleads everybody who reads it.

The average is the most trusted number in business reporting and the least examined. It is trusted because everybody understands how it is calculated and nobody suspects a figure they could work out themselves. But an average compresses an entire distribution into one value, and in doing so it throws away the shape. If your orders cluster tightly around one hundred dollars, the average tells you almost everything. If nine orders are ten dollars and one is nine thousand, the average is about nine hundred, and it describes nothing that ever happened and no customer who exists.

So the average is not wrong in that second case. It is exactly right and completely useless, which is a more dangerous combination than being wrong, because a wrong number can eventually be caught while a useless one just quietly misinforms. Before you report any average, look at the extremes and the middle. What is the largest value, what is the smallest, and what value sits in the middle when everything is arranged in order? If the middle value and the average are far apart, the average is being pulled around by a few unusual records and should not travel alone.

There is a second question that matters just as much: what is the average over? An average order value and an average customer value are different figures that people conflate constantly. And an average calculated across records where some values are absent is an average of only the records that had a value, over a smaller population than you think. Always report the count alongside the average. Two numbers together are harder to misread than one, and the count is often the number that reveals the problem.