Here is a scene I've watched play out in organisation after organisation. An analyst needs a particular dataset — customers, say, or last year's transactions. They don't know if it exists, and if it does, they don't know where it lives, what it's called, whether it's the right version, or whether they're allowed to use it. So they do what everyone does: they ask around. They message a colleague, who points them to someone else, who half-remembers a table in a database somewhere, maintained by a person who left last year. Hours or days later they've either found something and aren't sure they can trust it, or they've given up and rebuilt it themselves — creating yet another slightly-different copy of the data nobody will be able to find next time. Multiply that by every analyst and every dataset, and you have one of the largest, most invisible drains on productivity in any data-rich organisation. The fix has a name — a data catalogue — but the fix is more subtle than buying one, and getting that subtlety wrong is why so many catalogues sit empty.

What a catalogue is actually for

A data catalogue is, at its simplest, an inventory of your data assets with enough information attached that people can find, understand, and trust what they find. Think of it as a library catalogue for data: it tells you what exists, where it lives, what it means, who owns it, how fresh it is, and whether you can rely on it. Done well, it turns "ask three people and hope" into "search and find," and that shift is worth an enormous amount of recovered time and avoided duplication.

The information that makes a catalogue useful falls into a few kinds:

  • Technical metadata — what the data physically is: tables, columns, types, where it's stored. The easy part; tools can scan most of this automatically.
  • Business metadata — what it means: that "cust_type_2" is the customer segment, defined thus, and here's the difference between revenue and revenue_net. This is the part that makes data understandable, and it can't be scanned. A human has to supply it.
  • Ownership and trust signals — who's responsible for this, is it the certified version, how current is it. This is what lets someone trust what they found rather than nervously double-checking it.

Why catalogues fail

Here's the trap, and I want to be blunt about it because it wastes so much money: organisations buy a cataloguing tool, point it at their systems, let it auto-scan all the technical metadata, and declare victory. Six months later the catalogue is a graveyard nobody visits, because the auto-scanned technical metadata — table and column names — is exactly the part that doesn't solve the problem. Knowing a column is called cust_type_2 and is an integer tells you nothing about what it means or whether you can trust it. The valuable metadata — the business meaning, the ownership, the trust signals — is precisely the part a tool can't generate for you. It has to be supplied and maintained by people, and if you don't build the habit and the ownership to do that, the fanciest catalogue on the market becomes a comprehensive index of things nobody understands.

So a catalogue is not a purchase. It's a practice. The tool is maybe a quarter of it; the other three-quarters is the ongoing human work of describing, owning, and maintaining the entries — the same lesson that haunts every governance effort.

How to make one that lives

  • Start where the pain is, not everywhere. Don't try to catalogue the entire estate on day one; you'll produce a vast, shallow, unmaintained index. Catalogue the most-used, most-argued-over data first — the datasets people actually fight to find — and earn trust with a small, genuinely useful, well-described core.
  • Assign owners, or it rots. Every important dataset needs a named person accountable for its catalogue entry being accurate. Ownerless entries go stale, and a stale catalogue is worse than none because it actively misleads.
  • Make the business meaning mandatory, not optional. The description, the definitions, the "what this actually is" — treat that as the point of the entry, not a nice-to-have. It's the part that does the work.
  • Put it where people already look. A catalogue people have to remember to visit will be forgotten. The more it's woven into the tools people already use to find and query data, the more it gets used.

The communication heart of it

Underneath, a data catalogue is a communication tool, which is why it's really a people problem wearing a technical costume. The reason nobody can find anything is rarely that the data doesn't exist — it's that the knowledge about the data lives in people's heads instead of somewhere shared. A catalogue is an attempt to get that knowledge out of individual heads and into a common, findable place, so the organisation stops depending on who happens to remember what. That's a cultural change as much as a technical one: it asks people to write down what they know, and to keep it current, for the benefit of colleagues they may never meet. Get the culture right and the tool amplifies it. Get only the tool and you've bought an empty library with beautiful shelving. The shelves were never the problem. The books nobody wrote down are.