Six clusters nobody understood became six personas the board approved
The setup
Pinnacle Fitness runs 40 gyms and an app that tracks check-ins, class bookings, and renewals. Dwayne handles brand. Every year he has to walk into a board meeting and justify the marketing plan to executives who think in P&L, not percentiles. This year he wanted the plan built on real member segments, not vibes.
The problem
The segments existed. Dwayne had run the data and had six clean clusters: check-in frequency, class mix, tenure, renewal behavior, all neatly separated. On screen it was a beautiful heatmap summary of numbers.
In a boardroom, it was death. “I put up a table where Cluster 4 has a check-in index of 1.8 and a renewal rate two points above mean, and I watched every executive’s eyes glaze,” he said. “It was correct. It was also unusable. Nobody funds Cluster 4. They fund people they can picture.”
He needed to turn statistics into human beings, fast, without inventing things the data didn’t support.
The turning point
Dwayne ran AI persona cards on his clusters. Divisio fed each segment’s summary to Claude, which read the numbers and wrote a persona for every group, grounded in the actual stats rather than made up.
Cluster 4 became “The Streak Keeper”: a tagline, a plain-language description that cited the real numbers (checks in four or five times a week, books classes constantly, renews like clockwork), three to five sharp trait phrases, and, the part Dwayne loved, one ready-to-brief campaign idea aimed right at them. The tiny outlier cluster came back honestly flagged as probably outliers, treat with caution instead of dressed up as a fake archetype.
Six clusters, six people: The Streak Keeper, The New Year Hopeful, The Weekend Warrior, The Quiet Lapser, The Class Devotee, The Corporate Perk User. The board didn’t see a heatmap. They saw members they recognized from their own gyms.
How he did it
- Ran pattern detection on member behavior and confirmed six clean segments in the heatmap summary.
- Renamed segments with working labels so the AI had hints about how he thought of them.
- Generated AI persona cards, one per cluster, each grounded in the segment’s real numbers.
- Dropped the persona names, taglines, and traits straight onto board slides.
- Used each card’s suggested campaign as the seed for the year’s brief.
The payoff
The plan got approved in the room, with no “come back with more detail.” Pinnacle reorganized its entire annual calendar around the six personas, and the suggested campaign for The Quiet Lapser (a low-pressure “your gym misses you” reactivation) became a recurring play that lifted win-back sign-ups by a double-digit percentage.
The personas outlived the meeting, too. They became shared language. When someone said “that’s a Weekend Warrior message,” the whole team knew exactly who they meant.
“I’d had the right data for weeks. The personas are what made it land. The board doesn’t fund Cluster 4. They fund The Streak Keeper.”
Dwayne, brand marketer at Pinnacle Fitness
Feature spotlight: AI persona cards (Pro)
Divisio hands each segment’s summary to Claude, which writes a persona grounded in the actual numbers: a memorable name, a one-line tagline, a description citing the most telling stats, three to five trait phrases, and one ready-to-brief campaign idea. It never invents attributes the data can’t support, and it’ll tell you when a tiny segment is probably just outliers. Statistics in, people the whole org can picture out.