What a Lingua Franca for Business Data Actually Means

By Aaron Gammon · Founder, Inferrex · June 2026

I describe Inferrex as the lingua franca of business data more than any other way, and I've come to realise the phrase does a lot of quiet work that's worth unpacking. People hear "lingua franca" and reach for "translator" — and that's wrong in a way that matters, because translation is exactly the trap the whole industry is stuck in.

This is a piece about the concept, with the analogies I actually think in. Where it touches what Inferrex does, I keep to what it produces, not how. Live platform figures are at inferrex.com/claims — I don't bake them into essays because they move.

Imagine a port city before any shared language existed. A merchant arrives from somewhere far off, needing to agree price, quantity, quality, and terms with a local trader who shares not one word with him. So they gesture. They point. They hold up fingers, draw in the sand, and spend an afternoon establishing what could have been said in a sentence — and at the end of it they're still not sure they agreed the same thing.

That is manual field mapping. An integration consultant pointing at a Salesforce field, then at a HubSpot field, drawing the line between them on a whiteboard, spending a week confirming the two fields mean the same thing.

The first fix humanity found for this was the translator. And the translator is genuinely useful — but look at the shape of the cost. A translator knows one pair of languages. Each new trading partner means learning a new language from scratch. The translator is a bottleneck, a cost centre, and a single point of failure, all at once.

That's a systems integrator. Someone learns Salesforce-to-SAP. Then learns SAP-to-Workday, separately. Then Workday-to-Oracle, separately again. Each connection is its own expedition, each costs a fortune, and each one breaks the moment either side changes. Translation between pairs never stops being expensive, because the number of pairs grows far faster than the number of systems. That's the trap, and almost the whole integration industry is still standing in it, just with better whiteboards.

Then, in human history, something different happened — and it's the thing the word "lingua franca" actually names.

Certain languages became bridges. Aramaic across the Near East. Greek across the Eastern Mediterranean. Latin across Western Europe. Arabic across science and mathematics. English across global commerce. None of them started as universal languages, and none of them won by being "better" than the local tongues. They spread through network effect: each new speaker made the language more useful to every existing speaker, until the cost of not speaking it was isolation from the network. Adoption became inevitable — not by force, by sheer utility.

And here's the part people miss when they hear "lingua franca" and think "translator." The bridge language didn't replace the local ones. People still spoke their own dialect at home, in their community, in their own context. The lingua franca sat above the local languages as a shared layer, and let everyone communicate through it without anyone abandoning who they were.

That distinction is the entire point. A translator stands between two parties and converts, pair by pair, forever. A lingua franca stands above all parties as one shared understanding they each speak through. The first scales like the integration industry's costs — quadratically, painfully. The second scales like a language — learn it once, and you can talk to everyone who also speaks it.

A lingua franca for business data is that second thing. Not a converter wired between Salesforce and HubSpot. A shared layer that understands what "customer," "invoice," "payment," and "lifecycle stage" mean — above every system's local dialect — so each system communicates through the shared understanding rather than through a fragile bilateral wire.

Here's what that buys you, structurally.

When integration is translation, every system needs a wire to every other system it wants to talk to — and the number of those wires explodes as you add systems. When integration is a lingua franca, every system maps to the shared language once, and through that shared understanding it can reach every other system that has also mapped to it. One act of understanding per system, not one hand-built wire per pair.

I make the economic version of this argument in the pillar piece on comprehension over connection and the maths of it in the N×N problem. The linguistic version is simpler to feel: nobody learns a separate private language for every person they'll ever meet. You learn the shared one, and the whole network opens at once. That's the difference between translating and speaking — and it's the difference between a connector catalogue and a comprehension layer.

It's also bidirectional for free. Once a system's dialect is mapped to the shared meaning, understanding flows both ways — reading from it and writing back to it are the same mapping run in opposite directions. A pile of one-way translators never gives you that; a shared language does, because understanding isn't directional.

The thing the shared language is built out of is meaning, and that's what makes it durable.

A fishing village develops thirty words for types of wave; a mountain people develop a dozen for kinds of snow. Nobody was wrong — each dialect is perfectly adapted to its own reality. The same is true of every software vendor that ever shaped "customer" to fit its own data model. None of them was being difficult. They adapted to their context, and the dialects drifted apart until they couldn't understand each other.

A translation tool keys off the surface — this field is called customer_email, wire it there. So when a vendor renames the field, the wire snaps, because the name was the only thing holding it together. A lingua franca keys off the meaning — this field, whatever it's currently called, is a customer's email address. Rename it and the shared language still recognises the same concept wearing a new label. The rename is a new bit of vocabulary to absorb, not a break to repair. That's why a comprehension layer can heal itself where a mapping tool can only break: meaning survives a rename; a wire doesn't. I make that case in the self-healing piece.

The last property is the one I find most beautiful, and it's why I insist the corpus is a language and not a dictionary.

Lingua francas don't just translate. They absorb. Latin took in Greek philosophy, Arabic mathematics, Germanic law, and grew richer for it. English borrowed "entrepreneur" from French, "tsunami" from Japanese, "algorithm" from Arabic — each word arriving because no existing word captured the same thing. The language absorbed the concept because the concept was useful, and within a generation it felt native, as if it had always been there.

If that feels abstract, consider dinner. There was no tomato in Italy until the sixteenth century, and now Italian cooking is unimaginable without it. The chilli is American, and it now defines the food of India, Thailand, Sichuan, and Korea. Nobody legislated these ingredients in. They spread by usefulness, got woven into existing traditions, and came to feel native — and the cuisines didn't lose their identity by absorbing them. They became more themselves.

A shared language for business data absorbs the same way. One provider contributes a concept the others don't model — a particular progression of payment states, a specific lifecycle of a sales lead, a clinical observation hierarchy. It proves useful, enters the shared vocabulary, and the next provider that connects finds its own version of that concept already understood, as though it had always belonged. The shared model doesn't get diluted by each new dialect it takes in. Like a cuisine taking on a new ingredient, it gets richer and more complete. Every new system that learns to speak the language makes the language more fluent for everyone already speaking it.

That's the difference between a dictionary and a living language. A dictionary is a frozen list. A living language grows, absorbs, and stays current by watching how the world actually speaks. A connector catalogue is a dictionary. A comprehension layer is a language.

Strip away the analogies and here's what a lingua franca for business data actually delivers that a stack of translators can't.

You map each system once, to shared meaning, instead of wiring every pair by hand. You get every connection in the network as a consequence of that understanding, bidirectionally, rather than building each one as its own project. You survive renames and provider changes, because the language is built on meaning and meaning outlasts labels. And the whole thing gets better as it grows, because every new dialect it absorbs makes it more fluent for everything already connected — the opposite of a connector catalogue, which only gets heavier to maintain.

That's why I keep using the phrase. "Integration platform" describes a box of wires. "Lingua franca of business data" describes what's actually needed: not a better translator, but a shared language — so everyone can stay exactly as different as they already are, and still understand each other.