Your personas are lying to you

Somewhere in every organisation there’s a slide deck with customer personas on it. Names like “Enterprise Eva” and “Startup Steve.” Demographics, motivations, pain points, preferred channels. Created in a workshop. Last updated: never.

We’ve seen these decks in a dozen engagements. They’re always wrong. Not completely — the broad strokes are usually reasonable. But the specifics? The messaging assumptions? The tone preferences? Those are guesses from a room of internal people projecting what they think their customers want.

We ran the Living Personas experiment to find out how wrong.

Within 60 days, the system corrected 4 out of 7 initial persona assumptions using actual behavioural data from live campaigns. The biggest one: the “enterprise evaluator” persona was assumed to prefer formal, professional tone. The data showed conversational tone converted 22% better. The team had been writing stiff, corporate copy for a segment that wanted to be talked to like humans.

But here’s what really caught our attention. The behavioural clustering didn’t just refine existing personas — it found two entirely new segments that nobody had identified:

Champions — internal advocates who were already evangelising the product to their teams. They didn’t need convincing. They needed ammunition — case studies, ROI data, comparison sheets to share internally.
Silent evaluators — people who trialled extensively, used every feature, read every doc, but never responded to a single marketing email. They weren’t disengaged. They were doing their own research and didn’t want to be interrupted.

Both segments had completely different messaging needs from anything in the persona deck. And both were invisible until the data surfaced them.

The takeaway we keep coming back to: static personas are better than nothing, but they’re worse than they think they are. The gap between “what we assume about our customers” and “what the data shows” is almost always wider than teams expect. And the corrections that matter most — tone, timing, messaging approach — are exactly the things humans are worst at intuiting.

Full experiments: Living Personas and Persona & Cohort Builder. The thread we’re still pulling: can you detect when a persona has drifted far enough from reality that it’s doing more harm than good — a staleness signal?