A dashboard should end an argument, not start one

If a meeting opens by disputing whose number is right, the reporting has failed before anyone has made a decision.

There is a particular kind of meeting where the first twenty minutes are spent establishing whose number is correct. Two people have two figures for the same thing, both defensible, both produced differently. By the time it is settled, the meeting is half over and no decision has been made.

That meeting is a reporting failure, not a discipline failure. The reporting was built to display data rather than to settle a question.

Define the measure once

Most of these disputes come from the same root: a KPI that means slightly different things depending on who built the extract. Different date basis, different exclusions, different treatment of adjustments. Everyone is being honest and the numbers still disagree.

Fixing this is not glamorous. It means agreeing the definition with the people who own the process, writing it down, implementing it once in the model, and refusing to maintain a second version because a stakeholder prefers the old figure. That refusal is the hard part and it is the part that matters.

Design from the decision backwards

I try to start with a question rather than a dataset. Who looks at this, what are they deciding, and what would make them act differently today than yesterday?

That question kills a lot of charts. A visual that cannot change anybody's action is decoration, however attractive it is, and every one of them costs attention that the important number needed.

A dashboard is not successful because it looks good. It is successful when someone makes a better decision because of it.

Different audiences, different reports

An operational audience needs to know what needs attention today — exceptions, queues, what is stuck and who is holding it. A management audience needs trend, direction and where structural attention should go.

Trying to serve both in one view usually serves neither. The operational team drowns in trend they cannot act on, and management gets detail they cannot use. Two focused reports beat one comprehensive one.

The check nobody runs

The most useful question about a dashboard is asked three months after launch: is anyone opening it? Not whether it refreshes correctly. Whether it is used.

If nobody opens it, the honest response is to retire it rather than let it sit there consuming refresh capacity and implying the organisation is data-driven. Reporting that nobody reads is a form of overproduction, which is a waste in exactly the Lean sense.

A practical test before building: ask the requester what they will do differently depending on what the report shows. If there is no answer, the report is not ready to be built.