Why most dashboards are ignored — and how to design one that is not

Start from decisions, not from available data. List the recurring decisions each role makes — which sites to visit this week, which accounts are at risk, which orders need expediting — then show only the numbers that actually change those decisions.

Dashboards built the other way around, starting from 'what data do we have,' tend to accumulate charts nobody asked for. Each one seemed useful in isolation, but together they bury the two or three numbers someone actually needs under a screen of noise.

Dashboards built the other way around, starting from 'what data do we have,' tend to accumulate charts nobody asked for.

Define every metric once, in writing, and reuse that definition everywhere. Two definitions of revenue — one that includes tax, one that doesn't — will end the dashboard's credibility faster than any bug, because the moment two numbers disagree, people stop trusting either one.

A short metrics glossary, even a single shared page listing how each number is calculated and which system it comes from, prevents this from happening as more dashboards get built by more teams over time.

Show comparison by default: versus target, versus last period, versus other branches. A number without a reference point cannot trigger action — '142 tickets closed this week' means nothing on its own, but '142 tickets closed this week, versus a target of 120' immediately tells someone whether to act.

This is a small design change with an outsized effect: it turns a dashboard from a reporting artifact into something that actually prompts a decision the moment it's opened.

Finally, review usage after a month and delete what nobody opens. A shorter dashboard is a used dashboard — every chart that stays on screen but goes unread makes the ones that matter harder to find, which quietly pushes people back toward asking someone for a number instead of looking it up themselves.

Most BI tools already log which panels get viewed. Checking that log after 30 days, rather than assuming every chart earns its place forever, is usually enough to cut a dashboard down to the handful of views people actually rely on.

Most BI tools already log which panels get viewed.

Before publishing a dashboard, it's worth confirming: every panel maps to a named decision someone makes on a recurring basis; every metric has one written definition used consistently across the organization; every key number is shown next to a comparison point; and there's a review scheduled at 30 days to remove what isn't being used.