Your lab results come back perfect. Cholesterol fine, glucose fine, everything within range. And you know something's wrong, because for three weeks you've been sleeping four hours a night and your back hurts in a way it never did before.
Teams go through the same thing. The dashboard is green, lead time went down, the OKRs are being hit. In the retro someone says "we're doing fine" and nobody pushes back. But there's one person who hasn't turned their camera on in two months, another who no longer proposes anything, and a weird vibe since the last change in priorities. Nobody says it out loud. There's no data.
Robert McNamara was brilliant with numbers. He had modernized Ford through number crunching and in the sixties he became the US Secretary of Defense. In Vietnam he applied the same logic: the progress of the war was measured in enemy body counts, because that was what could be counted. What the population thought, morale, whether the strategy made any sense at all, all of that weighed much less than a number you could show every week. And the number said they were winning.
That's why this mistake is known as the McNamara fallacy: deciding only with what can be measured and ending up concluding that everything else doesn't matter, or simply doesn't exist.
The sociologist Daniel Yankelovich explained it in four steps that I think are some of the best things ever written about metrics. First you measure what's easy to measure. Then you ignore what isn't, or you make up a number for it. Then you assume that what can't be measured isn't that important. And finally you say it doesn't exist. He called that last step suicide, just like that, no sugarcoating.
Almost nobody gets to the fourth step on purpose. You get there little by little, meeting after meeting, every time someone brings up a feeling and the answer is "do you have any data to back that up?".
The team with the best lead time in the company, that nobody wants to depend on. The person who closes the most tickets and that everyone avoids in code reviews. The project that's green on the dashboard while the client is about to cancel. And the resignation "we didn't see coming", that the whole team had been seeing for months.
All of that was there. It just didn't fit on the dashboard.
Lead time tells you how long a task took, and nothing about how the person who finished it is doing. The problem is asking metrics to answer questions they can't answer. A team's trust, whether someone is about to burn out, whether a decision was really understood: that shows up somewhere else. In a one-on-one, in what gets said over coffee, on people's faces at the daily, in who stopped talking.
That's information too, even if it doesn't come in a chart. Not having a number only makes it easier to ignore.
The job of whoever leads, I think, is to hold both things at once. Look at the dashboard and also ask. And when someone says something is off, ask them what they're seeing before asking for a chart.
A while ago I wrote about quantophrenia, that fever for measuring everything. McNamara is where it takes you if you don't stop it in time: after measuring everything, you start believing that what you didn't measure didn't happen.
The lab results can come back perfect. Someone still has to ask you how you're sleeping.