The dashboard delusion

Here is the blunt version. A dashboard is a decision you have stopped making. You built it to watch a number, then you watched the number, and somewhere along the way watching replaced acting. The screen glows green. Everyone nods. Nobody owns the call the green was supposed to inform. That is the delusion, and it bites even the dashboards you can justify on paper.

I am not arguing against measurement. Measure away. I am arguing that a metric on a wall does nothing on its own, and that the act of staring at it tricks a whole team into feeling busy and in control while the actual decision rots. The dashboard becomes a comfort object. A worry blanket with a refresh rate.

Most advice about dashboards stops at “is this worth building.” Fair question, and a real one. But it skips the harder problem: the dashboards you already decided were worth it, the ones people really do watch, can still fool you rotten.

Why does watching feel like managing?

Because your brain rewards the watching. You open the tab, the chart loaded, the number sits inside the band you wanted, and a small hit of relief lands. Job done. Except no job was done. You looked at a thing.

This bugs me more than almost anything else in operations. In building Tallyfy over 10+ years, the pattern that keeps showing up is teams who can recite their metrics to two decimal places and cannot tell you what they would actually change if the metric moved. The number is a pet. They feed it attention. It feeds them calm. Neither of them does any work.

And the pets breed. Once watching a number counts as managing it, a team has every reason to make more numbers to watch. Each new worry gets its own tile. The wall fills up. Six months later you have a 30-tile dashboard nobody reads top to bottom, where every tile is a past worry hardened into a chart and then forgotten. When I teach this, the point that tends to land is a plain one. Half the tiles on a typical dashboard survive because removing them feels riskier than keeping them. Which is a daft reason to keep anything, when you say it out loud.

W. Edwards Deming saw this decades before the dashboard era. In Out of the Crisis he quotes Lloyd Nelson: the most important figures that one needs for management “are unknown or unknowable”, yet management must still take account of them. The stuff that matters most is often the stuff your dashboard cannot show. So a wall of visible numbers can lull you into managing only what is easy to plot, which is rarely the thing that will sink you.

There is a name for the failure mode where the easy-to-plot number turns into the goal itself. Worth knowing.

Goodhart law and the metric that turns on you

The moment a number becomes a target, it stops being a good number. That is Goodhart law, in the phrasing the anthropologist Marilyn Strathern gave it in 1997: “When a measure becomes a target, it ceases to be a good measure.” Charles Goodhart, a British economist, coined the original in 1975 while needling the Thatcher government over monetary policy. The idea travels well beyond central banking. Put a metric on a dashboard, tie someone bonus or status to it, and people game the metric rather than the thing the metric was a proxy for. Support teams close tickets faster and resolve fewer problems. Sales books more demos and qualifies worse. The chart goes the right way. The business goes the wrong way. You are now worse off than if you had never measured it, because the dashboard is lying to you with a straight face and a confident upward slope.

Eric Ries gave a sibling idea its label. He calls them vanity metrics in The Lean Startup, and his test is plain: “They might make you feel good, but they don’t offer clear guidance for what to do.” Registered users. Total signups that never come back down. These only ever climb, so they only ever flatter. A metric that cannot drop in a way that forces a decision is decoration. I have shipped vanity metrics myself, more than once, and felt clever about the line going up before realizing it could not tell me anything I could act on. Easy mistake. I still catch myself doing it.

Now stay with me, because there is a second, sneakier failure here.

The green dashboard that is quietly red.

When green is actually red

Project people have a word for this and it is a good one. The watermelon. Green on the rind, red all the way through. The status report shows green, everyone relaxes, and underneath the project is on fire. Watermelon reporting, as the PMO Professionals describe it, is when the status looks green on the outside but is “actually red right through” once you look inside. It happens because people learn what the dashboard rewards. Report amber, get questioned. Report red, get blamed. So the rational move, for your own skin, is to keep the rind green and hope.

I have watched this dynamic up close. The dashboard does not cause the lie, but it creates the incentive to lie, because it turns a living situation into a single coloured pixel that someone gets judged on. The richer the truth, the worse a one-pixel summary serves it. And once a team stops trusting the green, every status meeting turns into an interrogation about whether green really means green, which is slower and nastier than just talking about the work.

So you have three traps stacked on top of each other. Watching feels like managing. Targets corrupt the measure. Green hides red. The thread running through all three is the same. A dashboard with no owned decision behind it controls nothing. It just performs control. Expensive theatre, paid for in the scarcest currency a team has, which is attention.

What rescues a dashboard from all of this is almost embarrassingly simple to state, and surprisingly hard to do.

Name the decision, name the owner

For every standing number you keep, write two things next to it. The decision it should trigger when it moves. And the one person who owns that decision. Not a committee. Not “the team”. One name.

Walk a real example. “On-time delivery rate.” Fine. What decision does it trigger? Say: if it drops below the agreed line for two weeks running, we pull a person off new work and onto the backlog, and the head of ops makes that call. Now the number has a job. It is wired to an action and an owner. If it moves and nobody does the thing, the owner is accountable for the gap, not the chart.

Run every dashboard through that filter and watch what happens. A few light up. They have a clear decision and a clear owner behind them, and those you keep and act on. The rest go quiet. No decision attached. No owner attached. Just a number someone built once because it felt responsible to track it. Those are the comfort objects. Delete them. I mean actually delete them, not “archive for later”, because a parked dashboard still costs glance-time every time it sits in a list.

This is the encyclopedic version, the bit you can hand a colleague. A dashboard earns its place when three things are true. One, it maps to a decision a named owner will make when the number crosses a line you agreed in advance. Two, the metric is a real proxy for the outcome and not a vanity number that only climbs, so it can move in a direction that hurts and forces the decision. Three, the owner is accountable for acting, which kills the watermelon incentive because the question stops being “what colour is it” and becomes “what did you do when it turned”. Strip any one of the three and the dashboard slides back into theatre: watched, soothing, and inert. Most fail at least one. A surprising number fail all three and survive for years on the momentum of having once been built.

Most failed analytics work traces back to the same gap: a number with no decision wired to it.

Does any of this kill dashboards entirely? No. A dashboard wired to a decision and an owner is one of the better tools you have, because it shortens the loop between something changing and someone doing something about it. The problem was never the screen. It was the quiet assumption that the screen was doing the managing for you.

So go count yours. My guess is you will find more comfort objects than controls, and that the act of deleting the dead ones will feel weirdly good, like clearing a desk you had stopped seeing. I might be wrong about the proportions for your shop. I am not wrong that a number nobody owns is a decision nobody is making, dressed up to look like one.