Dashboards people actually open: why most are abandoned in two months

Almost every company we talk to has at least one abandoned dashboard. It was built with enthusiasm, shown in a meeting, and today nobody opens it. People still message the person who knows to get the number.

The diagnosis is rarely technical. The dashboard works perfectly. The problem is that it answers questions nobody was asking.

Start from the decision, not the data

The right question isn't "what data do we have?" but "what will you do differently tomorrow if you see this number?"

If there's no answer, that metric doesn't go on the dashboard. It can go in a monthly report, it can go in an archive, but it doesn't occupy space on a screen someone looks at every day.

Compare these two:

• "Month-to-date revenue." → Interesting. Changes no decision tomorrow.
• "Orders unshipped for more than 48 hours." → Someone opens the list and calls someone. That's a dashboard.

The second kind of metric has a property the first lacks: it's actionable and it has an owner. A specific person is responsible for making that number go down.

Four numbers, not forty

A dashboard with forty metrics isn't informative — it's an elegant way of refusing to prioritise. Nobody reads it, because reading it costs more than asking the person who knows.

The discipline that works: one screen, no scrolling, four to six large numbers, each with a comparison against target or against the previous period. Everything else lives one click deeper, for when one of those numbers turns ugly and needs investigating.

A number without a reference is noise. "Sales: 1,240" says nothing. "Sales: 1,240 — 12% below the same weekday last week" is information.

Real time is usually unnecessary

"Real time" is what everyone asks for and almost nobody needs. The question is the shortest interval in which somebody can actually act on the data:

• An executive dashboard is looked at once a day. Refreshing it hourly is money burnt.
• An operational dashboard — dispatches, support queue, production line — genuinely needs minutes, because someone reacts in minutes.
• A financial dashboard is read weekly or monthly, and there freshness matters far less than the number being right.

Real time is expensive: infrastructure, complexity, and a permanent source of "the number doesn't match the system." It pays for itself when someone truly acts inside that window. Otherwise it's a feature bought for the demo.

Where the data comes from

Here's the part of the project nobody budgets.

The data lives in the ERP, the storefront, spreadsheets, one legacy system, and one person's head. Consolidating it forces agreement on definitions the company never wrote down:

• Does a sale count when invoiced, when paid, or when shipped?
• Do cancelled orders subtract from the day they were placed or the day they were cancelled?
• Is an "active customer" one who bought in the last 90 days or the last 12 months?

These arguments are the real value of the project, even though they don't look like it. A dashboard where the commercial and finance leads read the same number two different ways is worse than no dashboard: it generates meetings about where the number came from instead of what to do about it.

Write the definitions somewhere visible and have the dashboard surface them one click away. It costs an afternoon and saves a year of argument.

The trust rule

A dashboard is abandoned the day somebody cross-checks it against the source system and it doesn't match. After that it never recovers: every time it shows something odd, the first hypothesis will be that it's wrong.

So before announcing it, cross-check against the source for two or three weeks and fix the differences. And when a difference is legitimate — because the dashboard excludes voided orders, say — explain it on the screen itself, not in an email nobody will find.

How to know it worked

There's only one honest metric: does anyone open it unprompted?

At the two-month mark, look at the access logs. If it only gets used when you mention it in a meeting, it didn't work — and what needs fixing is almost never the design, but having picked metrics that change no decision.

If you'd like us to sort out what can be measured with what you already have and what needs connecting first, tell us which systems you run.