Building a BI dashboard people actually use

Many dashboards are opened twice and then forgotten. The ones that survive answer a small number of questions that someone asks every week.

Start with the decisions

Ask each audience what they decide and how often: a finance manager approving spend weekly, a field supervisor checking daily output. Each decision suggests two or three measures. If a chart does not support a decision, leave it out.

One page, one audience

• Put the most important number at the top left.
• Show trends over time rather than single values.
• Compare against a target or the previous period so a number has meaning.
• Use the same colours for the same things everywhere.

Make the data trustworthy

State where each figure comes from and when it was last refreshed. A dashboard that is wrong once loses its audience. Reconcile key totals with the source system before launch.

Automate the refresh

If a dashboard depends on someone exporting spreadsheets, it will go stale. Connect it directly to the source systems or a regularly updated data store.

Review after a month

Check which views are used and which are not, and ask users what they still look up elsewhere. Remove what nobody uses.

See our data, AI and analytics services.