The Excel glossary trap: When definitions become noise
“Customer.”
A small word. Just 4 letters in Swedish (kund). And yet it managed to create more chaos than a whole container full of flat packs with no instructions. In a corporation with over 200,000 employees, “customer” becomes more of a philosophical problem than a business concept.
Many years ago, I was new in the role of Enterprise Architect at a large international retail company. I was responsible for information architecture and integration when I was introduced to an ongoing “business glossary” project. The project’s ambition was healthy: to create a shared business language that could bridge both people and IT. So far, so good.
They had collected words from everywhere: IT systems, domain experts, spontaneous corridor whispers, and crammed them all into Excel sheets. These lists were then distributed to individual “experts”, a title some received with the same joy as an unexpected dentist appointment. One word = one sentence = universal truth, preferably before lunch. As if semantics were something you could force out of people with enough spreadsheet tabs.
From semantic friction to shared understanding
I started reading. And I started laughing.
Not at the people. But at the idea that “customer” means the same thing for production, warehouse, marketing, HR and finance. It’s like assuming everyone uses the same recipe for meatballs, just because they’re called “meatballs.”
It was a bit like gathering all dialects in Sweden and asking them to agree on one way to say “kex” (or tomato in the English-speaking community). Or trying to summarise the taste of cinnamon buns with a single adjective. Technically possible, but you lose something essential in the process.
So, I started to build concept models, the first one based on the existing definitions, to demonstrate the problem with the current lists. And that first model revealed everything we already knew but hadn’t quite dared to admit: We mean different things in the same words, and standalone definitions without context don’t make sense. And that’s okay, as long as we talk about it, and then do something about it!
And that changed something. It sparked conversations. Laughter in the hallways. Real insights. People could see where the friction actually was. We stopped debating the “correct” definition and started talking about real needs and how different concepts relate to one another. There was a beautiful pattern to discover behind the mess.
Lessons for leaders: Ontologies, ownership, and the voice of the business
Things I learned and still carry with me today:
- Don’t start with the glossary. Start by understanding a concept model or ontology.
- Don’t always try to unify a definition. Show why people interpret things differently.
- And please: never let IT own the business language. Let the business keep its voice.
Language isn’t about forcing the world into cells. It’s about creating shared meaning. And sometimes, that starts by admitting we actually mean different things, truly.
The experiences from that large company also taught me something else: that real understanding must be built, not forced. That a common language doesn’t emerge solely through standardisation, but through dialogue. And that business language must be owned by those who actually speak it, not by IT.
…and yes, you’re right. Those who recognise the story have guessed it: That large international retail company was IKEA.
And it’s important to say they’ve come a long way since. In fact, IKEA now works in a far more model-based and structured way and has even employed ontologists (which in itself is a fantastic word to drop on someone in the checkout line).
So no, this isn’t a takedown. It’s a story about how difficult it is to organise language, even in the most ambitious organisations. And about how it’s possible to get better. Much better.