Goodhart's law
expressionops culturegovernance & risk
'When a measure becomes a target, it ceases to be a good measure.'
Economist Charles Goodhart observed the effect in 1975 monetary policy; anthropologist Marilyn Strathern gave it the crisp modern phrasing. Engineering lives it daily: optimize ticket-close counts and quality dies, optimize line coverage and tests stop asserting, optimize velocity and estimates inflate. Any metric worth managing by is worth gaming, so pair every target with the behavior it is supposed to proxy.
When a measure becomes a target, it ceases to be a good measure. Charles Goodhart wrote it about monetary policy, and the mechanism generalizes to anything measured: the moment people are rewarded for a number, they optimize the number, and its correlation with the thing you actually cared about weakens.
Software supplies endless examples. Measure test coverage and you get tests that execute code without asserting anything. Measure tickets closed and issues get split, or closed as works-as-designed. Measure lines of code and you get verbosity; measure velocity and estimates inflate quietly until the chart looks healthy. In each case nobody is cheating exactly, they are responding to the incentive that was set.
The uncomfortable conclusion is that there is no metric immune to this, only metrics not yet load-bearing enough to distort. Partial defences exist: use several measures that are hard to game simultaneously, treat metrics as prompts for conversation rather than as verdicts, and keep the number away from compensation. The most reliable safeguard is asking what behaviour this measure would produce if someone optimized it ruthlessly, before publishing it.