One focus group, start to finish.
Follow a crew's marketing persona through one focus group for Keepslot, a made-up booking tool that takes deposits so small appointment businesses lose fewer slots to no-shows. On test: a new homepage headline.
Every specimen is an illustration. The panelists are real library entries; their words and every number were written for this page.
Choose the room
On the crew's Cohorts page, you filter the library as you would an ad audience — type, generation, tier, channel reach, Tapestry LifeMode, NAICS sector — and tick seven owners and managers, the two front-line practices at two seats each.
Add a rationale, tags, and a product-context block saying what Keepslot does and costs. The builder refuses fewer than six members and warns when more than a third of them are barely reachable on your test's channel. Saving makes version 1, and the box applies it on its next sync.
appointment-owners · version 1 · 9 seats
Dr. Chloe Tsaiveterinary practice owner · 2 seats
Tanya Ruizclinic practice manager · 2 seats
Tonya Mitchellowner of two dental practices
Tameka Williamshealthcare operations manager
Brian KowalskiHVAC owner-operator
Diane Hartmanninsurance agency owner- Stephanie Liang-Parkboutique law firm partner
Shaped like owners and front-desk managers of small appointment businesses.
An experiment, then a focus group
Earlier, the persona's ideate task pre-registered the experiment and named this cohort because its tags matched; klingbar.com refuses any later change to the registration. On its next scheduled run, the focus-group task writes the headline and its page, brand-checks both, fixes what fails, and opens the focus group. You can also start one from the cohort's page.
- experiment
x-20260915-keep-the-slot- arms
control(today's headline) ·keep- primary
- Trial signups per homepage visit, higher is better
- guardrail
- Demo requests, must not fall
- cohort
appointment-ownersv1
Seven reactions, each alone
Each panelist is one engine call on your box, in its own empty directory holding the copy and a rendered page. The five questions never change — first reaction, the promise said back, who it is for, what to keep and cut, whether it would act — and the persona may add two.
Each answer closes with STANCE and WOULD_ACT lines, and the counts come from those. A round counts once two thirds of the cohort answer; all seven did.
Never lose a slot to a no-show again. Keepslot takes a card deposit when a client books and texts a reminder the day before.
A score, and what to change
A separate call with a fixed rubric reads the answers, so the persona never grades its own headline. The score is the resonant seats plus half the mixed ones, over the seats that answered: 56.
The persona reads the synthesis before any transcript, notes which objection each change answers and which landed line stays, revises, and brand-checks again.
- counts
- Resonant 3 · mixed 4 · dissent 2, in seats of 9, from the panelists' own
STANCElines - landed
- The deposit at booking; reminders by text
- objections
- “Never” overpromises; a deposit reads as distrust of regulars; no word on which calendars
- their language
- “lost slots” “my regulars” “a worried owner”
- revision brief
- Drop “never”. Let owners skip the deposit for clients they know. Name the calendars.
Did the revision answer you?
Round 2 shows “Keep the slot. Keep the regulars.” Each panelist also finds its own round-1 answer — only its own — and one more question: did this address what you raised? A third closing line, ADDRESSED, records the answer.
It scores 72. Three panelists say the revision answered them, three say partly, and one says no.
The stop rule
Round 3 lets the owner decide what a same-day cancellation costs, and scores 78. One seat still dissents: the insurance agency takes calls, not bookings.
After round 3 the persona stops if the score rose less than five points, or if dissent is a quarter of the seats or fewer. Otherwise a fourth round runs, never past five. The verdict is earned: 60 or more, dissent within a third of the seats, and the brand check passed.
- scores
- 56 → 72 → 78
- rose under 5?
- No, it rose 6
- dissent ≤ ¼?
- Yes, 1 of 9 seats, so it stops
- verdict
- Earned
One card, and it ships nothing
The persona closes the focus group, commits the creative to its own branch, opens a pull request in your app's repository, and proposes one card. Its title names the cohort and version, so you know which room shaped the work, and the detail links each asset on klingbar.com. Approving marks the creative ready to test; nothing is merged yet.
AP-41 — Final creative for x-20260915-keep-the-slot (2 assets, cohort appointment-owners v1, score 56 → 78 over 3 rounds)Landed: deposits for new clients only, reminders by text, the owner sets same-day terms. Still open: agencies that take calls. The cohort predicts keep wins.
A launch card, then a daily look
The launch card merges the pull request, deploys, checks the revision now serving, and starts the experiment in Statsig on half the homepage traffic. The same card approves bounded daily re-weighting and the stops back to control. Every morning the persona takes a look; it posts only when something changed, and klingbar.com keeps every look.
This step runs because Keepslot's app is wired the way the persona expects: Statsig for the arms, and a conversions source for the metric that decides. Yours needs the same before a launch card can start anything.
keep against control- day 3
- 212 and 205 exposures · too young to decide
- day 9
- P(beat) 0.81 · expected loss 1.6% · continue
- day 16
- P(beat) 0.93 · expected loss 0.4% · winner:
keep
No winner before seven days and 100 exposures per arm.
A winner waits for its rollout card
On day 16 the look clears the bar registered before launch — P(beat control) of 0.90 or more, expected loss under 1% — and the persona records the decision. The rollout card sends everyone to keep, merges the cleanup, deploys, and checks what is serving.
The learning it then proposes quotes the low end of the lift, and the cohort's page records that it called the winner. Had a guardrail shown harm on two looks running, the persona would have gone back to control without asking, and said so.
AP-47 — Roll out x-20260915-keep-the-slot → keep (P(beat) 0.93, expected loss 0.4%, lift +12% (+3%, +22%))Bring the headline you are unsure about
Early access is free, and we set up the marketing persona with you; it is not self-serve yet. Read the focus groups overview, browse the panelists, or see how AI focus groups fit beside real research.
Request early access