A category manager at a retailer once told me, quite cheerfully, that she didn't use the dashboard we'd built her because she "didn't trust the numbers." I asked which number was wrong. She couldn't name one. She just knew, in the way people know things, that the spreadsheet she'd kept for years was right and this glossy new thing was suspect. And here's the part that took me too long to accept: she was being completely reasonable. The dashboard was correct. It didn't matter, because correctness and trust are two different things, and I'd built the first while assuming it would deliver the second for free. It never does.
I've spent a stretch now running Power BI trainings for exactly these people — the non-technical teams who are supposed to use the dashboards we lovingly engineer — and it has quietly changed how I build. Trust turns out to be buildable, but not with better DAX. With better communication.
Why a correct dashboard gets distrusted
Start by taking the distrust seriously instead of being wounded by it. When someone doesn't believe a dashboard, it's almost never because they've found an error. It's one of these, and none of them are about the maths:
- It disagrees with their spreadsheet, and nobody explained why. They have a number they've trusted for years. The dashboard shows a different one. In the absence of an explanation, the new thing is the suspect thing — obviously. The disagreement is usually a definition difference (the dashboard counts "active customers" differently than their sheet did), but until someone surfaces that out loud, it just reads as "the dashboard is wrong."
- It's a black box. The spreadsheet, however janky, was theirs — they built it, they can see every formula. The dashboard arrives as a finished object from the data team with its workings hidden. People don't trust what they can't inspect, and "trust me, the pipeline is correct" is not inspection.
- It was wrong once. If the dashboard showed a genuinely bad number even once, early on, that memory outlives ten thousand correct refreshes. Trust is asymmetric: slow to earn, instant to lose, and the first impression is doing far more work than you'd like.
- Nobody they trust has vouched for it. People believe numbers that people they respect already believe. A dashboard blessed by no one carries no social proof, however sound its engineering.
What actually builds the trust
So in the trainings — and then in how I build — the fixes are communicative, not technical:
- Reconcile out loud, once, together. The single most powerful move: sit with the sceptic, put their spreadsheet next to the dashboard, and walk through why the two numbers differ until you land on the definition gap. Nine times in ten it ends with "oh — you're counting returns and I wasn't," and the moment they understand the difference, the distrust evaporates. You cannot skip this by asserting correctness. You have to show the seam.
- Make the definitions visible. Every key number on a dashboard should be one hover away from a plain-language definition of what it means and where it comes from. "Revenue = invoiced sales excluding VAT and returns, refreshed nightly from the finance system." When the definition is inspectable, the black box becomes a glass box, and glass boxes get trusted.
- Teach people to read it, don't just hand it over. This is what the trainings are for. A dashboard dropped on someone's desk is a black box; a dashboard someone's been walked through — this is where this comes from, this is what this filter does, this is how to check a number you doubt — becomes a tool they own. The training isn't about Power BI features. It's about transferring enough understanding that trust becomes possible.
- Never ship a wrong number, especially early. Given how asymmetric trust is, the launch period matters disproportionately. I'd rather delay a dashboard a week to be certain of its numbers than launch on time with one embarrassing figure that poisons the well for a year.
The training room is where you learn what you built wrong
The unexpected gift of running these sessions is that a training room is the most honest usability test you'll ever get. Watching a non-technical team try to use your dashboard live, in front of you, is where you discover that the label you thought was obvious means nothing to them, that the filter you're proud of confuses everyone, that the metric you led with isn't the one they care about. You built the thing inside your own head; the training shows you the inside of theirs, and the gap between the two is where all the distrust was living.
I've redesigned more dashboards off the back of a training session than off any amount of solo polishing, because the session tells you the truth: not "is this correct?" but "can the person it's for actually read it and believe it?" Those are different questions, and only the second one determines whether the thing gets used.
A dashboard nobody trusts is just an expensive screensaver. The engineering earns you the right to be correct; only the conversation earns you the right to be believed — and being believed is the entire point.
The quiet part
Here's what I'd say to data teams who find this frustrating — who feel that building the thing correctly should be enough, and that trust ought to follow from truth as a matter of justice. I sympathise, and it doesn't, and it never will, because humans don't extend trust to correctness they can't see. This is the same conviction I came into data with from the communication side and have never had reason to abandon: a number is a message before it's a fact, and a message only lands if the person receiving it understands and believes the sender. You can be completely right and completely ignored.
Which reframes the job. Building the dashboard is half of it. The other half — the reconciling, the defining, the teaching, the being-careful-at-launch — isn't a soft add-on to the real technical work. It is the work, or at least the half of it that decides whether the technical half was worth doing. A brilliant dashboard nobody trusts delivered exactly nothing. The category manager with her spreadsheet taught me that, cheerfully, and I've built differently ever since.