The dashboard nobody opens — how to build BI that actually gets used

Most enterprise dashboards are abandoned within ninety days of launch. Not because the data is wrong. Because the dashboard answers questions nobody is asking.

Here’s how we approach BI engagements differently.

Decisions first, data second

Before we touch Power BI, Tableau, or any visualisation tool, we list the decisions the dashboard has to enable. Not the metrics. The decisions.

“Should we add a shift to plant 3 next month?” is a decision. “Plant 3 utilisation %” is a metric — and only one of several inputs to that decision.

Dashboards that get used are decision-shaped. Dashboards that get abandoned are metric-shaped.

Three numbers, not thirty

The most-used operational dashboards we’ve shipped have three to five numbers visible at a glance, with progressive disclosure for everything else. The CFO who opens the dashboard at 7 AM doesn’t have time for thirty tiles.

If your stakeholder can’t tell you which three numbers matter most, the dashboard isn’t ready to be built.

Refresh latency matches the decision cadence

A daily decision needs a dashboard that’s at most a day old. A 5-minute trading decision needs sub-minute freshness. Most dashboards get this wrong in both directions — daily-decision dashboards refreshed every minute (expensive, no benefit), real-time dashboards refreshed nightly (useless).

Operations team owns it, not the data team

The single biggest predictor of a dashboard surviving past 90 days: someone on the operations team — not the data team — feels personal ownership of it. Otherwise it goes stale and nobody fixes it.

We hand over BI deliverables with the operations owner identified, trained, and committed to maintaining the model. That’s the actual delivery, not the dashboard file.