When Measurement Makes You Worse
A measurement has value only if it reduces uncertainty.
There’s a general principle from Doug Hubbard’s “How to Measure Anything” that most organisations ignore: a measurement has value only if it reduces uncertainty. Instead, teams surround themselves with dashboards, KPIs, and reports that create the appearance of control without improving decision-making. The problem isn’t a lack of measurement—it’s that much of what is measured doesn’t actually help anyone make a better decision.
In product delivery, this shows up clearly through vanity metrics like velocity, downloads, logins and clicks. Teams track story points completed, celebrate increases, and use it as a signal of progress. But velocity doesn’t tell you whether you’re delivering value, reducing risk, or moving closer to an outcome that matters. It creates confidence without clarity—a precise number that answers the wrong question.
If a measurement only has value when it reduces uncertainty…Then:
A measurement is harmful if it increases uncertainty or creates false certainty.
This is where measurement becomes harmful. If it doesn’t reduce uncertainty, it often increases it by creating false certainty. Leaders feel informed, teams optimise toward the metric, and the system becomes efficient at producing outputs that may not matter. The metric stops reflecting reality and starts distorting it.
To get out of this trap, shift the question. For any metric, ask:
What decision changes because of this?
What uncertainty does it reduce?
What behaviour will it drive?
If you can’t answer these clearly, the metric isn’t helping—it’s misleading. And once you see that, it becomes much harder to unsee.
If you find yourself surrounded by bad measures, don’t try to fix the metric—fix the question. Start by identifying the decision that actually matters (e.g. “Should we continue investing in this feature?” or “Are we reducing delivery risk?”), then work backwards to what you’re uncertain about. From there, define a measurement that directly reduces that uncertainty, even if it’s rough or probabilistic. It’s better to have an imperfect measure of the right thing than a precise measure of the wrong one. Over time, remove or de-emphasise metrics that don’t inform decisions—because every metric you keep shapes behaviour, whether you intend it to or not.
References: How to measure anything - Douglas Hubbard