Nobody wants data governance. Everybody wants to trust their numbers.

Let me start by conceding the obvious. “Data governance” is a terrible name for something important. It sounds like a committee. It sounds like a policy document nobody reads, a SharePoint folder full of PDFs last opened in 2019, a meeting that could have been an email about a meeting. If it makes your eyes glaze slightly, I don’t blame you. It has spent years being presented in the most boring way imaginable.

So let me describe what it actually is by describing its absence, because the absence is something everyone recognises.

It’s two reports that answer the same question with different numbers, and a meeting that turns into a debate about whose figure is right instead of what to do about it. It’s the field that means one thing in the student record system and something different in the finance system, so the two never quite reconcile and everyone has privately given up expecting them to. It’s the definition of “active student” that turns out to be three different definitions wearing the same name. It’s the report only Dave knows how to run, and the slow realisation, the week Dave leaves, that nobody wrote any of it down.

None of that feels like a governance problem while it’s happening. It feels like a bad afternoon. But that is exactly what governance is: the presence or absence of the boring agreements that decide whether your data means the same thing tomorrow as it did today, and the same thing in your office as it does in mine.

Here is the reframe I’d offer anyone who thinks this isn’t for them. Data governance is not paperwork about data. It is the thing that decides whether you can trust your own numbers. And trusting your own numbers turns out to be the foundation under almost everything else you might want to do.

Think about what sits on top of it. You want to make decisions based on evidence rather than the loudest person in the room – that needs numbers everyone agrees are real. You want to submit a clean regulatory return without three weeks of archaeology – that needs data that meant the right thing when it was entered. You want to build dashboards people actually use – which only happens if they trust what the dashboards say, and they only trust it if the numbers stopped contradicting each other. You want to use AI on your data, the ambition of the moment – and AI applied to badly governed data doesn’t fix the mess, it launders it and scales it, confidently. Every one of these rests on the unglamorous layer underneath. You cannot do clever things with data you don’t trust. You can only do fast things wrong.

This is why I get a little impatient when governance is treated as the optional, do-it-later part – the tidying up you’ll get to once the exciting projects are done. It’s backwards. The exciting projects are the ones standing on top of it. Skimping on governance to get to the interesting work faster is like skipping the foundations to get to the fun part of building the house sooner. It does genuinely feel faster, right up until the point it very much doesn’t.

And I want to be clear about what good governance actually looks like, because the caricature puts people off before they start. It is not a thick policy nobody follows. It is not a committee that meets to discuss the committee. At its most useful it is surprisingly modest: a shared understanding of what your key data means, agreed definitions that don’t quietly differ between teams, someone who actually owns each important dataset, and enough written down that the institution doesn’t lose a critical capability every time someone changes job. That’s most of it. It is much closer to good housekeeping than to bureaucracy, and the institutions that do it well tend to have less process than the ones that don’t, not more, because they’re not constantly firefighting the consequences of not having it.

The reason it gets skipped isn’t that people don’t value the outcome. Everyone values trustworthy numbers. It’s that governance is invisible when it’s working and only noticeable when it’s absent, which makes it perpetually easy to defer. Nobody is ever thanked for the reconciliation error that didn’t happen, the return that went smoothly, the report that survived a resignation. The wins are quiet by nature. The failures are loud, and they arrive at the worst possible time, usually wearing the disguise of a bad afternoon rather than the governance problem they actually are.

So I’d retire the phrase, if I could, and replace it with the question it’s really about. Not “do we have data governance”, which sounds like a compliance exercise nobody wants to own. Just: can we trust our numbers? Do they mean the same thing everywhere? Would we survive the right person leaving? Those are the questions the boring word is standing in for, and they are not boring at all. They decide whether everything you build on top of your data is standing on something solid, or something you’re quietly hoping holds.

Nobody wants data governance. But everybody, without exception, wants to trust their numbers. They are the same thing. One just has a much better name than the other.