A dashboard nobody opens is a failed dashboard
Why they get abandoned, the one-screen-one-decision rule, why cadence is part of the product, and the three-question test that tells you whether yours is alive.
Every dashboard looks good on the day it ships. The question that matters is different: six weeks later, does anyone open it?
In most cases the answer is no, and it is almost never the tool’s fault. It is a decision-design problem, not a visualisation one.
Why they get abandoned
| Symptom | Actual cause |
|---|---|
| Nobody has opened it in weeks | It answers no question anybody has |
| Everyone looks and nobody acts | It shows state, not deviation: the “against what” is missing |
| Each person reads a different number | The same metric is defined two ways |
| Only opened when something went wrong | No cadence; it is an autopsy, not a dashboard |
| It has forty charts | Built from what was available, not from what was needed |
The second symptom is the most deceptive, because from outside it looks like success: people open it, look, and go back to work. It happens when the board shows state without showing deviation. “Sales were 40” is not actionable. “Sales were 40 against a plan of 55” is, because now there is a gap and somebody has to explain it.
The third is the most corrosive. When two teams read the same label with two definitions — “revenue” including or excluding cancellations, “leads” counting or not counting duplicates — every meeting starts by arguing about the number instead of the decision. By the third time, people stop trusting the whole screen.
One screen, one decision
The rule that saves the most dashboards: each view is designed for one concrete decision somebody takes at a known frequency. Not for a topic, not for a department, not for “seeing how we are doing”.
Written as a question, before opening the tool:
- Do I move budget between channels this week?
- Do we have a lead response problem?
- Are we going to hit the month?
Each of those is a separate view, with the three or four metrics that answer it and nothing else. A board trying to answer all three answers none.
A dashboard without a cadence is a poster
A board nobody is obliged to look at, at some defined moment, becomes decoration. Cadence is not an organisational detail: it is part of the product.
And it works better the other way round from how it usually gets built: instead of waiting for someone to come in, have the board go out and find them. A Monday summary carrying the week’s deviation and a link to the full view gets read far more than the best screen somebody has to remember to open.
The other half is alerting, and the same warning applies as in what to automate first: the most important alert is not a bad number, it is a number that stopped arriving.
Every number needs an owner
Two different owners, in fact, and confusing them is what destroys trust:
- Who acts on the number. The person who does something different when it moves. If there is nobody, the metric is informational and does not belong on a decision board.
- Who answers for its definition. What it includes, what it excludes, which source it comes from and in which time zone. Written down somewhere, and reachable from the board itself.
The second almost never exists, and it is the one that prevents the meeting argument. When two numbers disagree, the answer should not be an investigation: it should be opening the definition. We wrote about that separately in when your sources disagree.
What to remove first
- Vanity metrics. Impressions, reach, followers. They almost never change a decision and they occupy the place of something that would.
- Charts nobody knows how to read. If it needs explaining, it does not work on a board.
- Comparisons against nothing. A number with no prior period, no plan and no target is a loose figure.
- Anything that could be looked at once a quarter. That is a report, not a dashboard.
A board with six metrics somebody checks weekly is worth more than one with forty nobody understands. Removing is harder than adding, and it pays better.
The test that tells you if it works
Three questions, answered honestly:
- When was the last time somebody changed something because of what they saw there?
- If the board stopped refreshing tomorrow, how long before anyone noticed?
- Do two different people read the same number the same way?
The second is the most honest of the three, because good intentions cannot answer it. If the answer is “I don’t know”, the board is already abandoned — nobody has noticed yet.
FAQ
- Which tool should I use? It matters far less than people think. Most abandoned dashboards were built in perfectly good tools. What decides adoption is whether the screen answers a question someone actually has, on a rhythm that matches how they work.
- Should executives get their own view? Yes, and it should be smaller than the operational one. A director needs to know whether things are going well and where the problem is. Everything beyond that belongs to whoever can act on it.
- Is real time worth it? Almost never. Real time only earns its cost when a decision would change within the hour. Otherwise it adds fragility, load and noise in exchange for a figure nobody acts on until Monday.
- What if the data is not reliable enough yet? Then that is the project, not the dashboard. A well-built screen over unreliable data makes bad decisions faster and gives them an authority they have not earned.