How Measuring Performance Makes Things Worse

Here's a quick thought on measurement to contemplate as you head into the weekend.

Scientists constantly wrestle with the problem of how their very presence in a research setting can change the behavior of what is being studied. This same principle was brought to a business realm by Robert D. Austin in his groundbreaking 1996 book, Measuring and Managing Performance in Organizations.

What Austin learned was that the very act of measuring something often creates the wrong behavior as a result, behavior that is against the goal of the measurement. So if a company chooses to study the failure rates of parts coming off an assembly line, workers respond by slowing down production to raise quality. And if the company measures unit production, quality may suffer in the rush to get product out the door. In both cases the measures are achieved (yippee!), but the overall goals of the company are perverted.

Microsoft's Steven Sinofsky, who ran the Window's 7 product team, writes about this phenomenon in a new book written with Harvard Business School professor Marco Iansiti, One Strategy. Measuring the "bug fix rate", for example, encourages the withholding of bugs until the fix is assured. He says:

"Thus understanding that if we choose to measure something, we are also choosing both to exclude some data and realize the direct action of the measurement. Trying to correct this by adding yet another counter-measure only yields a Rube-Goldberg status report that few can understand and even fewer acknowledge or act on."

The way to beat this problem is admittedly difficult: measure everything, not selectively.