Why I Stopped Building Dashboards and Started Building Trust

The biggest mistake I see companies make with data is treating it as a reporting problem.

They invest in BI tools. They hire analysts. They build dashboards. The dashboards look good in the demo. Then six months later nobody is looking at them.

This pattern is so common it almost feels normal. But it is not inevitable. It is a symptom of solving the wrong problem.

It Is a Trust Problem

Here is what actually happens.

Leadership gets a new dashboard. They look at the revenue number and something feels off. They ask the analyst to check it. The analyst spends two days tracing back through the pipeline. They find the issue, fix it, and report back. Leadership nods. But the seed has been planted.

Next month the same thing happens with a different number. Then again. After a few cycles, the dashboard is technically still there. People technically still have access. But nobody opens it anymore because nobody trusts it. Decisions go back to being made from spreadsheets and gut feel, which is exactly where they were before the dashboard project started.

The problem was never the dashboard. The problem was the data underneath it.

The Boring Work Matters Most

I have spent six years working across ten-plus enterprise clients and the pattern is consistent. The teams with the most trusted data are not the ones with the best dashboards. They are the ones that did the boring work first.

Clean models. Consistent metric definitions. Automated testing. Proper documentation.

These things are not exciting. Nobody puts them in a project proposal. But they are the difference between a data team that gets listened to and one that gets questioned.

When I build a dbt model, every metric is defined in one place. Not across five slightly different versions buried in five different dashboards. One definition, version controlled, with tests that run before anything reaches production. If a column should never be null, there is a test that says so. If a customer ID should be unique, there is a test that confirms it. If something breaks, it breaks loudly and early, not quietly in a report someone is presenting to the CFO.

That is what builds trust. Not the visualisation layer. The layer underneath it.

Start Small

The other mistake I see constantly is trying to boil the ocean.

A company decides to get serious about data and immediately wants a single dashboard that answers every question for every team. That project takes six months, involves forty stakeholders, and produces something so generic it is useful to nobody.

The best data work I have delivered started small. One clear use case. One stakeholder. One problem solved well.

When you solve one problem well, people trust the output. When they trust the output, they come back with more questions. You solve those too. Over time the platform grows, but it grows on a foundation that people believe in. That is how you build something that actually gets used.

Trying to build everything at once means you build nothing people trust.

What AI Changed

I was an early adopter of AI tools in analytics work and I have thought a lot about what it actually changes versus what it just makes faster.

The thing I keep coming back to is this: AI is only as good as the data layer underneath it.

When I built a report generation tool using OpenAI and Claude, the output was impressive because the dbt models underneath it were clean, well-tested, and properly documented. The AI knew what each metric meant, what the business rules were, and what data it could and could not access. It produced reliable output because it was working with a reliable foundation.

I have seen what happens when you try to layer AI on top of a messy data layer. The AI confidently produces the wrong answer. It hallucinates metrics that do not exist. It joins tables at the wrong grain. The output looks polished and is completely untrustworthy.

Without a solid foundation, you are not improving your analytics with AI. You are just automating bad data faster.

The Shift

The work I am most proud of is not the dashboards I built. It is the moments when a CFO asked why two reports showed different revenue numbers and I could trace it back through the pipeline, show exactly where the logic lived, and explain it clearly in thirty seconds.

That is what trust looks like in practice. Not a beautiful visualisation. The ability to stand behind the numbers completely.

That is what I mean when I say I stopped building dashboards and started building trust. The dashboard is still there. But it is the last step, not the first one.