A data scientist walks into a meeting with 40 slides and six months of hard work.

The first 35 slides explain the methodology: there are correlation matrices, feature importance charts, a model architecture diagram that looks like the Paris metro map. The words "statistically significant" appear four times.

The business stakeholders nod. Nodding is polite. Nodding tells everyone else: I am smart enough to understand.

Then comes Q&A:

  • "Can we get more accuracy? 90% is not enough"
  • "If the market crashes, will the model self-learn?"
  • "Three-month forecasting horizon is a good start, can we do 24 months?"

The data scientist's eye twitches.

The meeting ends. No pilot gets approved. Instead, everyone agrees to try out the forecasting on all remaining regions, each with its own edge cases and local quirks.

Two years later, nothing has been implemented.

Sounds familiar? So what went wrong?

Not the data. The data was fine.

If you're in the data team, the diagnostic is very simple: "business people don't understand the data." If you are on the business side, it's more like: "these overpaid nerds don't know anything about the real world."

Both are kind of true. But they are missing the point.

This is not a knowledge problem. It's a translation problem.

Data people and business people are each fluent in completely different languages. When they meet in a conference room, they are both doing their best.

And they are both, from the other side of the table, completely incomprehensible.

Who owns this mess?

Everyone. Here's the breakdown:

The data team Built something technically impressive. Presented it in a language their audience couldn't speak. Seriously, no one in the business world wants to go through 40 slides of methodology.

The business stakeholders Nodded politely instead of asking basic questions. Then expanded scope instead of running a pilot. They probably weren't able to generalise their problems to the extent that a machine could grasp and solve them.

The organisation Let both of the above happen. Never built a bridge between the two groups. Never asked the most obvious question: will the people who need to use this actually be able to use it? Everyone assumed someone else had checked. Nobody had.

The gap nobody talks about

It's not a data gap. It's a translation gap: the missing conversation between what data can do and what the business actually needs.

That conversation requires a specific kind of fluency. It's rarely taught. Rarely hired for. Rarely taken seriously.

This blog is about that conversation.