The Definitions Problem: Why Your Data Can't Talk to Itself

Ask three departments how many active customers you have and there's a decent chance you get three different numbers. Not because anyone's lying, and not because the systems are broken. It's because "active" doesn't mean the same thing to all three of them.

Sales might define active as anyone with a live contract. Finance might mean anyone who's paid in the last 90 days. Support might mean anyone who's raised a ticket this quarter. Each definition is internally consistent. Each team can defend theirs perfectly. And none of them can be reconciled with the others, because they were never designed to be reconciled in the first place. Every department built its own version of the truth, independently, for its own purposes, long before anyone tried to join them up.

This is the quiet failure mode behind a lot of data projects that get blamed on the technology. The integration was fine. The pipeline worked. The dashboard loaded. What broke was meaning.

It's a governance problem wearing a technical costume

When an integration project stalls, the instinct is to look for a plumbing issue: the wrong tool, a schema mismatch, a system that doesn't talk to another system. Sometimes that's genuinely it. But just as often, the systems talk to each other fine. They pass the data across without a hitch. The problem only shows up once someone tries to use it, and discovers that "region," "customer," "revenue" or "active" means something different depending on which system it came from.

That's not a technology failure. It's the absence of a decision that was never made: who owns what a term means, across the whole business, not just within one department's four walls. Most organisations have plenty of people who own systems. Very few have anyone whose job is to own definitions.

The scale of this is bigger than most leadership teams realise. In a recent survey of UK business leaders, 36% named a lack of clear data governance and ownership as one of their biggest data-related risks, and 33% pointed specifically to the difficulty of managing data consistently across multiple platforms and systems. Those aren't small numbers for a problem that's rarely on anyone's roadmap as its own line item.

Siloed working doesn't just hide data, it fractures its meaning

There's a second layer to this. It's not only that definitions drift apart, it's that once they have, nobody finds out until it matters. Almost 3 in 10 leaders in the same research say data is genuinely difficult to access because of siloed working, and one in five suspect data isn't being shared across teams at all, whether that's deliberate hoarding or simple neglect. The same report's conclusion goes further, naming siloed data as one of the critical root causes behind data being hoarded and left dark in the first place.

Put those together and you get the real shape of the problem: teams that can't see each other's data, building their own definitions in isolation, for years, until someone tries to merge the picture and discovers the pieces don't fit. By then, unpicking whose definition is "right" is a political problem as much as a technical one, because every team has been making real decisions on their version the whole time.

The cost of leaving this unaddressed shows up right at the top of the business. 67% of the c-suite in that same survey say the data-based intelligence they rely on is too contradictory to even help them reach informed conclusions. Not incomplete. Not slightly out of date. Contradictory: multiple sources claiming to answer the same question, disagreeing with each other, with no way to tell which one to trust.

What actually fixes it

Buying a bigger data warehouse doesn't fix this. Neither does a more ambitious BI tool. Both will happily ingest three conflicting definitions of "active customer" and present all three with equal confidence, which is arguably worse than not having the number at all.

What fixes it is unglamorous: a shared glossary of the terms that actually matter to the business, written down, with an accountable owner for each one, and a process for what happens when a new team wants to redefine something that already has a definition. That's a people and process problem before it's ever a technical one. The technology to store and enforce those definitions has existed for years. What's usually missing is the decision to make anyone responsible for doing it.

It's a less exciting project than an AI rollout or a new analytics platform. It also tends to be the reason those bigger projects either succeed or quietly disappoint everyone twelve months in.

This is the third in a short series on why more data isn't always better data. Next: what happens to analysts when nobody's solved this problem, and they're left to manually reconcile the mess, buried under more data than any one person can reasonably hold in their head.