When the CEO of one of Belgium's biggest radio groups tells you what is keeping his attention, it is worth listening.
For Kim Beyns, who leads NGroup — the group behind NRJ, Nostalgie and Chérie — the answer over the past few months has been the persona panel. “What fascinates me the most these last months is the personas,” he says in the conversation above. NGroup is among the first European radio groups to use audience modelling as a live decision-making tool, and this is what that looks like in practice.
A panel of 100 listeners that never closes
NGroup built a panel of 100 modelled listener personas — MediaDatak's predictive audience modeling — from real-world public data gathered with AI. The panel does not keep office hours. It is available 24 hours a day, seven days a week, and the team can put a question to it the moment a decision comes up.
There is a discipline built into it. The personas answer from data, not imagination — when there is nothing real to ground a response, the panel holds back rather than inventing one. That is the line between a research instrument and a chatbot.
Planning a radio summer
The most recent use was seasonal. “Last week we asked them: what do you expect from a radio during the summer?” Kim says. The questions were specific — should the morning show stay the same through the summer months? Should NGroup put on concerts? The panel came back with answers, and the team could go a step further by asking the same questions of one segment only: listeners under 35.
A complement to traditional research — not a replacement
Kim is deliberate about how he frames this, and the distinction matters. NGroup still runs its traditional focus group every year — seven real people, its own listeners and competitors' listeners, answering its questions in a room. The persona panel does not retire that work. It surrounds it.
“It's important to say this kind of research doesn't replace traditional research. … This is an extra research, and this is so rich. This is really very, very interesting.”
That is the honest version of where AI belongs in a research budget: not as a replacement for the focus group, but as the layer that lets a team ask a hundred more questions between focus groups.
Testing the decision before spending the budget
The second theme Kim returns to is prediction. “With AI, it's predictive — so we can now test,” he says. His example is one every programmer and marketer will recognise: there is rarely enough budget to pre-test a television ad the traditional way.
A persona panel changes that calculation. The team can show a new idea — or a draft script — to the panel and ask the questions that usually only surface after launch: what do you think of this idea, what is missing, what do you like? “And that is amazing,” Kim says.
The same logic, applied to the music
The predictive angle reaches into the playlist itself — the domain of MusicDatak, MediaDatak's sister company for music intelligence. Kim describes the output without any jargon: here are the 50 tracks your listeners already like — “and look, they even like these 10 tracks you don't play yet.”
Those last ten are the point. Surfacing songs an audience would respond to before they ever reach the playlist is exactly the kind of call that used to rest on instinct. And, as with the persona panel, Kim is clear that it stands next to traditional music research rather than over it — “a very, very interesting complement.”
None of this is a demo reel. It is a working account of how one of Belgium's largest radio groups now makes decisions — faster, across more questions, with a model of its own audience it can query at any hour. The traditional research still happens. There is simply a great deal more intelligence around it.