Numbers say what. People say why.

What I noticed at a sentiment analysis conference in Tokyo.

Last month I went to Tokyo for a few days of conferences. One of them was about sentiment analysis for market research. I wasn't sure what to expect - I work in performance marketing, not in academic research. But the conversation stayed with me, more than I thought it would.

What I noticed is that a lot of what we do in marketing today is still based on assumptions.

We have so much data. CTR, ROAS, CPA, conversion rates, session duration. We can see every step of the funnel. But after the conference, I started thinking about what those numbers actually tell us.

They tell us what people do. They almost never tell us why.

The gap between data and understanding

In performance marketing, when a campaign underperforms, the first thing we check is the data. Did the CTR drop? Is the CPA too high? Did the audience change?

That works for a while. But after some point, you hit a wall. The numbers go up and down for reasons you can't really explain.

I think this is the point where most performance marketers start guessing. We A/B test ten versions of an ad and hope one works. We change the targeting. We rewrite the landing page. But honestly, we're often guessing what the customer is thinking.

The conference reminded me that there's another type of data right in front of us, and we usually ignore it: what customers actually say.

Customer language is data too

Reviews. Comments under ads. Support questions. The exact words people type in the search bar. The objections people send to sales. The things people complain about on Reddit or X.

This is also data. It's just less structured.

In my work, I've noticed that the best landing pages I've written for clients almost always came from listening to actual customers, not from brainstorming sessions with the team. When a customer says "I was scared this would be too complicated" - that's a real objection. When thirty of them say it in slightly different ways, that's a pattern.

But going through hundreds of comments manually is hard. This is where AI starts to help.

What AI helps with - and what it doesn't

Sentiment analysis tools, and lately language models, are quite good at grouping comments by topic and emotion. They can sort through thousands of reviews and show you that, for example, 40% of negative ones are about delivery times, even if no one used the exact same wording.

To be fair, that's a real shortcut. A year ago I would have read everything manually.

But there's a limit.

The tools can show you patterns. They can't decide what's important. They can tell you 40% of negative reviews mention delivery, but they can't tell you whether that's the most strategic thing to fix, or whether operations can actually solve it.

That's still our job.

What I think is changing is the balance: less time reading everything, more time deciding what to do with what we read.

How this applies to performance marketing

For me, this changes how I think about a few things.

When I write a new ad, I want to start from the customer's actual words. Not from a tagline I made up.

When I redesign a landing page, I want to know what objections people had before they bought - or before they didn't.

When I look at a campaign that's not converting, I want to compare what the ad promises with what customers say after they buy.

These three things - search terms, objections, post-purchase feedback - give a clearer picture than ROAS alone.

Takeaway

What stayed with me from the Tokyo conference is that good marketing isn't only about being efficient. It's about understanding.

Numbers tell you what people are doing. Customer language tells you why.

If you only look at one, you'll keep optimising in the dark.

If you can use both, you start to see your audience more clearly. And once you see them more clearly, the ads stop being guesses.