Licences for sixty, a launch, then adoption down to a handful within months. The reason is never the tool. Adoption is a trust problem, not a licensing one.

Why your self-service BI rollout collapsed

Self-service analytics is sold as a destination. Buy the licences, build the dashboards, and the business will help itself. It rarely happens. The pattern is grimly familiar: licences bought for sixty people, a launch with a flurry of interest, and then adoption collapsing to a handful within months while everyone drifts back to spreadsheets and emailed attachments.

This connects with our data strategy and leadership work and, separately, with Consolidating multiple Business Central companies in one model.

The reason is almost never the tool. It is that people do not trust the numbers, cannot find the right report, or were never shown how any of it fits their actual job. Adoption is a trust and habit problem, not a licensing one, and you cannot buy your way out of it with more dashboards.

Adoption is a trust problem

People do not abandon a report because it is hard to open. They abandon it because the first time they checked it against something they knew, the numbers were off, or appeared to be, and trust once dented does not heal on its own. So most of the work of adoption is the work of earning and protecting trust: numbers that reconcile to the systems people already believe, definitions that stay stable, and a visible, fast answer to the question "how was this calculated". Win that and usage follows. Lose it and no amount of training will save you.

A thin layer of structure does most of the work

Self-service does not mean a free-for-all, and the organisations that treat it as one get chaos and then retreat from it. It means a curated space people can trust, and a small amount of structure carries most of the load. One certified, shared model, so everyone builds on the same truth. A tidy, well-named set of workspaces, so people can actually find things. A clear marking of what is trusted versus what is a draft, so nobody builds a board paper on someone's experiment. And a simple, visible route to request something new, with a known turnaround. This is light governance in service of freedom, not a policy binder.

Fewer, better reports

More dashboards is not more value. A long tail of similar, half-trusted reports is the enemy of adoption, because every near-duplicate is another chance for two numbers to disagree in public. Pull the usage data, find the handful of reports that carry real decisions, make those excellent and trusted, and retire the rest. People adopt a small set they believe far faster than a large set they have to second-guess. Pruning is not destruction. It is the precondition for trust.

People move adoption, not features

Tools do not train people; people do. The most reliable lever is a champion embedded in each team: someone who knows the work, uses the reports daily, and whom colleagues can ask without feeling foolish. And train for the job, not the tool. Show a person the two or three things relevant to their week, not a tour of every feature. Short, role-based sessions that answer "how do I do my job with this" beat a comprehensive course nobody finishes.

Measure it honestly

You cannot improve what you will not look at. The platform records who logs in, which reports are opened, and which are ignored. Think of it as a funnel: licensed, logged in, using regularly, deciding from it. Most rollouts leak badly between "logged in" and "using regularly", and that is where to focus. A report nobody opens is not a sunk cost to defend. It is a candidate for retirement.

So this week, pull your adoption numbers and find where the funnel leaks. If it is between logged in and using regularly, the fix is trust and relevance, not more licences or more dashboards. Rescuing and rightsizing self-service this way is part of our analytics strategy work and how we shape an Analytics Acceleration Programme, and the full playbook is in our Self-Service That People Use guide.

If any of this sounds familiar, talk to us about your data.

Related reading

• In the loop or on the loop: governing analytics that AI helped build
• The Spreadsheet That Became A System Nobody Maintains
• Analytics Is A Garden, Not A Project