The Metric That Learned to Smile
By Ayman Fouad Abdelgawad
A measure becomes dangerous when improving its appearance is easier than improving the work.
A service announces that its response time has fallen. The figures are clear and the chart looks encouraging. Then someone asks what counts as a response. If an automatic acknowledgement now stops the clock, the organisation may have become quicker at recording contact while leaving the original problem exactly where it was.
This imagined example captures a recurring tension in measurement. A useful indicator begins as a rough window onto something important. Once it becomes a target, people have reasons to concentrate on the window itself. The measure acquires an audience, a deadline and consequences. Its relationship with the underlying purpose may gradually become the least convenient part of it.
That does not make measurement foolish. Without records, an organisation can substitute anecdotes and confidence for evidence. Waiting times, completion rates and error counts can expose failures that senior people would otherwise never encounter. The difficulty is deciding what an improvement in the number actually entitles us to conclude.
Suppose staff are judged by how many cases they close. They might solve problems more effectively. They might also classify difficult cases differently, discourage uncertain applicants or close requests that later return through another channel. These possibilities are not morally equivalent, but they can produce similar figures. The number needs a surrounding account of how it was achieved.
A good measure therefore requires permission to disappoint. If a worsening result automatically produces blame, people learn that the organisation wants reassurance more urgently than knowledge. Honest reporting becomes an individual risk. The dashboard grows calmer while the work becomes more difficult, and its calmness is cited as evidence that the difficulties cannot be serious.
The answer is not to collect every conceivable statistic. Too many indicators can reproduce the same problem at greater expense. It is to keep asking a few stubborn questions: what experience is this figure meant to represent, what behaviour does rewarding it encourage, and whose experience disappears when we celebrate it? A small sample of actual cases can sometimes disturb a very polished aggregate.
I am most persuaded by institutions willing to retire a flattering measure when it stops being informative. That decision gives up an easy story about success in order to recover a harder understanding of the work. A metric should remain capable of bringing unwelcome news. When it has learned to smile on command, the organisation may still be counting carefully, but it has begun to lose the ability to find out.