Why Don't People Use the Dashboards They Asked For?
People do not use dashboards because nothing in their working week requires them to. A dashboard is a pull source: it waits to be visited. Unless a named owner, a defined threshold and a fixed review cadence sit around it, there is no moment at which failing to look costs anyone anything - so nobody looks, regardless of how good the build is.
Why this matters
The pattern is consistent enough to be predictable. A manager requests a dashboard urgently, chases it for months, and then opens it twice in four months. Analysts who check usage logs find this constantly, and it is rarely about the quality of the work.
What was actually satisfied was the request, not the need. Sometimes getting a dashboard built was an objective on someone’s annual goals, and the delivery closed it. Sometimes a single question needed answering once, and a report would have done. And sometimes the request was a way of deferring a decision by asking for more visibility first.
Adoption follows obligation. Where a number has an owner, a threshold that defines a problem, and a standing point in the operating cadence, it gets looked at. Where it does not, the screen is optional, and optional things lose to whatever is urgent.
The thing most people miss
The honest test is not how to increase usage. It is whether a decision changes based on what the dashboard says.
Ask what action follows if the number comes back high, and what action follows if it comes back low. If both answers are identical, no dashboard will fix that, because there was never a decision waiting on the data. The build was the deliverable, and the request was closed the moment it shipped.
Where a real decision does exist, the fix is usually delivery rather than design. Push the exception out to its owner with the evidence attached, instead of waiting for someone to come and find it. Usage stops being a behavioural problem once not looking has a consequence.
This is the same open loop that explains why dashboards fail to improve performance even when the numbers on them are entirely correct.