40 Reports, Zero Decisions: What Self-Service BI Actually Produces
Forty reports. I have walked into Power BI workspaces with forty, sixty, sometimes more. Each one built by someone who needed an answer at the time, got it, and moved on. The workspace grew. The trust collapsed.
Self-service BI was sold as empowerment. The pitch made sense: every team gets its own answers, no more waiting on a central BI team, the tool is there, the data is there, go build what you need.
You know what happens next. A report named "Sales Report v3 FINAL." Another called "Sales Report v3 FINAL revised." A third that finance built when they stopped trusting the first two. Nobody knows which version is current. Nobody can name who created the original. When the numbers matter, three people send three different files from three different reports, all reading the same source system, all showing different totals.
This is not a data problem. The source is accurate. The tool did what it was supposed to do. The breakdown happened somewhere else entirely.
Most organisations treat this as a governance problem. That framing is accurate, but it misses where governance actually broke down. It did not break down at the data layer. It broke down at a decision that nobody made before people started building.
The question is simple: which reports should exist? Not which ones do exist. Which ones should.
In an environment where BI actually works, someone answered that question first. They mapped which decisions the organisation needs to make, then designed just enough reports to serve those decisions. One authoritative source per decision type. Clear ownership. Exploratory analysis kept separate and clearly labeled as such. Every report has a purpose. Every team knows which one to trust.
In a report swamp, that conversation never happened. Each team built for its own need, which is individually rational. The swamp is the systemic result of forty individually rational decisions made without a shared frame.
We reach for governance tooling at that point. Power BI has endorsement, sensitivity labels, workspace access tiers. These are useful. They cannot fix the absence of an architecture decision. You cannot govern what was never designed. Endorsing a report means nothing when your teams have not agreed on which report contains the right revenue number and which ones should not exist at all.
What works is a conversation most organisations skip. Not a technology conversation. A business one: which decisions does this organisation actually need to make, and which reports serve those decisions? Run that exercise honestly and the number of reports you actually need is almost always smaller than what you have. Often far smaller.
What comes out of that conversation is structure: enough reports for every team, every role, every perimeter, no more and no fewer. Each report tied to a decision. Each piece of exploratory analysis separated clearly from the sources of truth. Not because governance policies say so, but because someone sat down and decided it.
Without that decision, self-service BI keeps producing reports. Each technically valid. None of them trusted.
If you are a BI developer managing a workspace that has outgrown its own logic, the answer is not stricter policies. It is the session that maps purpose to report and retires everything that has neither.
If you are the executive who approved self-service BI and is now hearing that teams are running from different numbers, the failure is not the tool. An architecture was never designed. That is fixable. The longer it runs without fixing, the more decisions get made from the wrong source.
The forty reports you have did not appear by accident. They appeared because good governance was assumed to follow naturally from broad access. It does not. Forty reports and zero decisions is the proof.
Forty reports and zero decisions is not a self-service problem. It is the absence of a decision about which reports should be the authoritative source and which ones can be retired. That decision does not emerge from the bottom up. It has to be designed. If you are managing a report library that has outgrown any possibility of trust, the investment in designing the right structure now is always less than the cost of continuing to maintain one that nobody believes. → Design the architecture
Design first. Then build.
Every team that builds without a shared architecture adds another report nobody fully trusts. One structured session maps which reports should exist, who owns them, and which ones can be retired.
Book a 30-Minute Discovery CallFrequently asked questions
Why do self-service BI environments end up with so many reports nobody trusts?
Self-service BI removes the bottleneck of a central team, but it does not replace the architecture decision that team was implicitly making. When every group builds for its own needs, reports accumulate without a shared structure. Nobody decides which one is authoritative. Over time, you get volume and no trust.
What is the difference between a governed BI environment and a report swamp?
In a governed environment, someone decided before build which reports would exist, what each one answers, and who owns it. In a report swamp, those decisions were never made. The technology is often identical. The missing element is a structured conversation about purpose and ownership, not better tooling.
How do you know when your BI environment has become a report swamp?
Three signals: people check figures from different reports before important meetings, nobody can confidently name who owns a given report, and new reports get created because existing ones are not trusted rather than because something genuinely new is needed. Any one of these is a warning. All three together is confirmation.
What is the first step to fixing a report library nobody trusts?
Map existing reports to the decisions they support. A report that cannot be tied to a specific decision is a candidate for retirement. Most organisations find that a small number of well-designed reports replace dozens of overlapping, semi-trusted ones. Start with the reports used in the most consequential decisions and agree on one authoritative source per decision type before building anything new.