Goodhart's law has a bibliography

The version everyone quotes, "when a measure becomes a target, it ceases to be a good measure," is not Goodhart's sentence. It is anthropologist Marilyn Strathern's summary, written in her 1997 paper on audit culture in British universities, per the law's documented history. Goodhart's original was about monetary policy: a statistical regularity collapses once authorities lean on it for control.

The university context is the instructive one for engineering leaders, because Strathern was not describing fraud. The academics gaming research-assessment metrics were mostly doing legitimate work, reshaped: publishing earlier and smaller, choosing safer questions, optimizing the unit of output the audit counted. The measure did not get gamed by villains. It got satisfied by reasonable people responding reasonably to what was rewarded.

That is the failure mode worth planning for. Not the engineer who fakes a number, who is rare, but the team that legitimately closes more tickets by slicing tickets thinner, legitimately raises deploy counts with deploys that carry less, legitimately improves review turnaround by reviewing less deeply. Every one of those is a defensible local decision. The dashboard improves. The system does not.

The working defenses are modest. Metrics in tension, so improving one at another's expense shows up. Measures read in trend and in context by someone who can ask why. And expiry dates, stated at adoption. The reversal condition is a metric so close to the actual outcome that satisfying it is the job (revenue, uptime for an uptime product), where Goodhart's teeth are dullest. Everything else on your dashboard is a proxy, and proxies age the day you start paying them attention.