Data should inform judgement, not replace it

Bruna Marquezine scores 90 for audience demand in Brazil and 3 in the UK. Same week, same data.

That figure comes from a study our colleague Simon Jacobs ran in 2026. He took thirty women who fronted fashion and beauty campaigns between May 2025 and May 2026 and scored each for audience demand in ten markets. A brand asking who should front its campaign would naturally reach for the overall ranking. It turned out to be the least useful thing in the data. The name at the top had no social accounts at all. Most of the thirty were past the peak of their current wave of attention.

What the data didn't do was pick. People who know the sector had chosen the thirty, and nothing in the data showed any of them was a mistake. What it showed was which name was right for which market, which platform and which moment. The study's own line puts it best: demand "narrows a shortlist. It does not choose from one."

That's the case for data in one example. It has two jobs, and choosing isn't either of them.

The first is to test a hunch. A leader's gut feeling usually comes from years of watching. Research can say whether it holds, how strongly, and above all for whom. That last part turns one person's instinct into something the rest of the business can see and back. Sometimes the answer is yes, and the leader moves faster. Sometimes it nudges, and the hunch gets sharper.

The second is to narrow the field, as the shortlist study did.

We learnt this in the creative industries, where data that dictates produces safe, average work. But it holds anywhere. Leaders should feel free, now and then, to make a call the data doesn't support. If they're right, say so. If they're wrong, learn from it. If they ignore good evidence every time, that's a conversation worth having plainly.

The risk has grown. In March 2026 David wrote about a Wharton study of what its researchers call "cognitive surrender": when a machine supplies the answer, people stop questioning it and come to believe it was their own. "When the AI was wrong, participants followed it 79.8% of the time." A dashboard, a model or a tidy ranking can do the same thing to a leader. The more fluent the answer, the more deliberately someone has to own it.

What to do on Monday

• Before the next research project starts, write down the hunch it's meant to test and which audience it's about.
• Ask for answers as "for whom", not "yes or no".
• When the data produces a ranking, ask what it's ranking for: which market, which channel, which moment.
• Keep a short log of calls made against the data, and look back at it in six months.

Steer starts with the decisions, not the data: Steer

← All points of view