Why nobody opens the dashboard you paid for

A dashboard nobody opens was not designed badly. It was delivered as a product rather than built into somebody's week. Its panels answered the questions that were interesting on the day of the design workshop, and no person was given a decision to make with them on a Tuesday morning. It survives when each panel belongs to a named owner who acts on it at a fixed time, and when the number is fresh enough to be worth acting on.

What keeps a dashboard alive past its first month?

A routine. The dashboard is not the deliverable; the ten minutes in which somebody reads it is the deliverable, and the screen is what makes those ten minutes cheaper.

That reframes the project. Instead of asking what data you have, you ask who will sit in front of this, when, and what they will do differently because of it. A dashboard that cannot answer those three questions is a wall decoration with a login.

How many decisions does one panel carry?

One. A panel showing stock levels, outstanding invoices and monthly sales teaches the reader nothing, because attention divides and no single figure becomes urgent.

The test is a sentence: "If this number moves, we do ___." If you cannot finish the sentence with an action, an owner and a rough threshold, the panel is not ready. Some panels exist to answer a question once — then it is a report, not a dashboard, and keeping it on a screen every morning afterwards is how a page fills with things nobody reads.

Who owns the number?

One person per number, and the owner is the person who would be asked why it moved.

That produces an uncomfortable conversation, which is the point. It also produces a definition: what is counted, from which system, at what moment, excluding what. Two panels showing revenue from two definitions is worse than one panel showing nothing, because the argument moves from the business to the numbers and never comes back.

Write the definition next to the panel, in one line, with the source system named. When the definition changes, the change is dated — a dashboard whose history was silently restated cannot be used to compare this quarter with the last.

What does stale data teach people?

That the dashboard lies. One morning of wrong numbers is enough: somebody makes a decision on a figure that turns out to be a week old, is embarrassed in a meeting, and never trusts the page again.

So freshness has to be visible, on the panel, without a click. And a broken feed must look broken: an empty chart or a row of zeros reads as "we sold nothing", which is a different and more dangerous message than "the export failed at 03:10". A dashboard that cannot distinguish zero from unknown is not a reporting tool; it is a way to be confidently wrong.

Alerts or charts?

If somebody must act within hours, push the number to them. A chart requires three things that are never all present: remembering to look, having the tab open, and knowing what normal looks like.

Alerts have their own discipline. One alert per real threshold, with the threshold written down, and an owner who receives it. A channel that fires ten times a day is muted within a week, and once it is muted the whole system has stopped working, quietly.

Which panels should you delete?

Panel	Keep it if	Delete it if
Daily sales	someone acts on it daily	a monthly figure sits on a daily page
Stock below minimum	it starts a purchase	it lists items nobody reorders
Overdue invoices	someone chases them weekly	the ledger is read instead
Visitor counters	a decision depends on them	they exist to look busy

Deleting is the cheapest improvement available. A screen with four panels gets read; a screen with thirty gets scrolled. The panels to remove first are the ones that were never clicked, the cumulative totals that only ever go up, and anything included because a competitor's dashboard had it.

What does the weekly ten minutes look like?

1. Same time, same person, every week, in the calendar as a meeting rather than a task.
2. Open three panels, not thirty.
3. For each, check the freshness marker before reading the number.
4. If a number is stale, log the feed problem and stop; a decision on old data is worse than a late decision.
5. Name one action, one owner and one date.
6. Write that line in a shared log, so next week starts from it.
7. Anything that produced no action for three weeks is deleted or replaced with a better question.
8. Once a month, check one number by hand against the source system.

That last step is not bureaucracy. It is the only thing that catches a pipeline that has been quietly wrong for six weeks while everyone nodded at the chart.

Where this goes wrong

Built for the buyer, not the user. The owner commissioned it, the owner looks at it, and the operations manager who could act on it does not have a login.

Thirty panels and therefore no priority. Everything visible means nothing is urgent. The one number that mattered on Monday is somewhere in the fourth row.

Access friction. A separate login, a VPN, a layout that only works on a desktop. Each of those removes most of the audience, and the dashboard dies of inconvenience rather than of bad data.

The dashboard as a substitute for a decision. A red panel that stays red for a quarter is evidence that the process is broken, not that the chart needs a redesign. Rebuilding the screen is the comfortable response and it changes nothing.

Silent restatement. Someone corrects last month's figure without a note. The trend line now says something that did not happen, and the meeting spends its ten minutes reconstructing history.

No owner for the pipeline. The person who built the exports leaves, a source system is upgraded, and the dashboard shows zeros. Nobody notices for a month, because nobody was assigned to notice.

What this does not cover

Nothing here is about chart types, colour or which tool to buy; those choices matter less than the routine and are easier to change later. It does not cover the architecture behind the numbers — a warehouse, a semantic layer, a single definition of revenue across systems — which is a separate job and, where it is missing, the first one to do. It also assumes the business has somebody who can act on each number; where a panel exists because a manager wants visibility but nobody has authority to change anything, the problem is not reporting.

We build dashboards around one weekly decision, with the owner of each number named and the feed monitored.