Test a product concept before you build it
Before you spend a quarter building something, write it up as the page a buyer would read and show it to a synthetic cohort shaped like that buyer. A crew's marketing persona runs three or more rounds on your box and sharpens the framing between them. You come away with the objections to answer and a clearer way to say what the product is. Then you ask real users.
Write the concept the way a buyer would meet it
Panelists react to what is in front of them, so the concept has to be something a buyer could actually be shown: a one-page description, or a rough landing page. Say what it is, who it is for, what it replaces, and what it costs, if you know. Write it for the buyer, not for an investor. A panelist who has never heard of your market will not fill in what the page leaves out.
You do not write the questions. Every round asks the same fixed ones: what the panelist thought this was in the first ten seconds, what it promises and whether they believe it, who it is for, which line they would keep and which they would cut, and what would stop them acting. The persona may add up to two questions of its own for the concept.
stimulus/ copy.md the concept, as a buyer would read it context.md what a panelist needs to know first page.html a rough landing page, if there is one
What one panelist's working directory holds in round 1. From round 2 it also holds that panelist's own previous answer, and never anyone else's.
A cohort shaped like the buyer, including the ones who say no
On klingbar.com, build a cohort of six to sixteen panelists from the public library. For a business concept, filter the business panelists by tier and NAICS sector. For a consumer one, filter by generation, Tapestry LifeMode, or channel reach. Give more seats to the people who sign the order, tag them by role, and add a product-context block.
Then write one sentence saying who the cohort is shaped like, because the builder will not save it without one. Leave in the skeptics: a room chosen to agree with you produces agreement and tells you nothing. The builder also refuses a cohort under six, because a group that small has no room to dissent.
Independent warehouses · v2 · shaped like operators of five to forty dock doors who book carriers by phone and whiteboard.
- owner · 3 seats
- family freight business
- ops manager · 2
- regional third-party logistics
- dispatcher · 1
- tagged
daily-user - receiving lead · 1
- tagged
daily-user - carrier · 1
- owner-operator, tagged
skeptic - + 3 more
- 8 members, 11 seats
For a concept, the objections matter more than the score
The first question tells you whether the concept is legible at all. If three of eight panelists describe a load board when you meant a dock scheduler, fix that before anything else. The second question surfaces what they would need to check before believing it. The synthesis gathers the objections, what landed, and the phrases panelists used unprompted, and writes a brief for the next draft.
Read the transcripts, not just the list. An objection one panelist made in passing is sometimes the one a real buyer leads with.
- objections
- Assumes carriers will book through a link, and half of them call. Silent on what happens when a truck is two hours late.
- landed
- "No more whiteboard."
- their language
- "the six a.m. scramble", "detention charges"
Three rounds, each one answering the last
Round 1: is it legible?
The first reading shows what people think the concept is and the objections it invites. The persona revises the page, notes which objection each change answers, and keeps the lines that landed.
Round 2: did the revision answer them?
Each panelist sees the new page beside its own round 1 answer and says whether the revision dealt with what it raised: yes, partly, or no. Nobody sees what another panelist said.
Round 3: what is left?
A third round shows which objections survive two revisions. The persona stops when the score rose by less than five points or dissent falls to a quarter of the seats, and never runs more than five rounds.
What remains after the last round is your interview guide: the objections that would not go away, the framing that finally read clearly, and the words panelists reached for. How each round runs covers the mechanics.
A rehearsal, not a verdict on the market
A concept that does well with a synthetic cohort has passed a rehearsal. It has not found a market. The panelists are invented, and they cannot tell you how many buyers share an objection or whether anyone will pay. Take what the rounds produced into five or ten real conversations with the people the cohort was shaped like, and lead with the objections that survived.
Once the concept is ready to meet traffic — a landing page or a waitlist email — and your app runs Statsig and reports conversions the way the persona expects, the persona can propose it as an experiment, and real behavior settles what conversations can only suggest. Message testing follows that path through launch and rollout. Synthetic respondents vs real people covers the research on where this method holds.
Find the objections before you write the code
Klingbar is free during early access. You bring the box and a Claude Code login, we set up the marketing persona with you, and you can browse the panelist library today without an account.
Request early access