A client once told us their new BI suite had 4% weekly active usage. The dashboards were technically excellent — fast, accurate, beautifully laid out. They just were not part of anyone's job.
Problem one: it answers a question nobody asked
Dashboards get built from available data rather than from decisions. Start the other way round: name a recurring decision, name who makes it, name when they make it. "The plant head decides Monday morning which orders to prioritise." Now build for that, and delete anything that does not serve it.
Problem two: the timing is wrong
A report that lands on the 8th about the month that ended on the 31st cannot change anything. Either move the data closer to real time, or accept it is an audit artefact, not a decision tool — and stop measuring adoption on it.
Problem three: nobody trusts the number
This is the quiet killer. If a user once found revenue on the dashboard disagreeing with the finance report, they will never trust it again — and no amount of visual polish recovers that.
The fix is a governed semantic layer: one definition of revenue, one of active customer, one of on-time delivery, defined once and reused everywhere. Then publish those definitions where the user can read them, and show the last-refreshed timestamp on every screen.
What changes when you fix it
At that same client we cut 40 dashboards to 6, tied each to a named recurring meeting, added a refresh timestamp and a definitions link. Weekly active usage went past 60% within two months. We did not add a single new feature.
Adoption is not a training problem. If people are not using the dashboard, it is usually because it does not help them do the thing they are already doing.