What they are, how to build them, and how to use them to price, package and position with confidence, in a fraction of the usual time. Written from our own runs, including the mistakes, the costs, and the checks we now run before a single number reaches a client.
Half of it is method. The other half is what we learned the hard way, which is the part you will not find in a vendor deck.
How a persona panel is built, which research instruments to run against it, and how to aggregate the answers into a directional read rather than a false-precise number.
If a price is visible anywhere the panel can see it, you do not get an independent read, you get a fixed fraction of the number you showed it. We show the evidence, including a run where 62 of 100 personas named the identical figure.
The same effect applies when a buyer asks an assistant to evaluate you. Whichever price it meets first becomes what you cost. What that implies for how you structure a pricing page.
Measured token counts at published API rates, by model, per 50 and per 100 replies. Plus the reason cost is rarely the real decision.
Any synthetic panel hands you confident-looking numbers. These are the checks that tell you whether to believe them.
Free PDF. We will send occasional updates as the method develops, and nothing else.
Want to run one yourself? The Agent Persona Survey is a free download that runs the full study on your own machine, and signal.clearseam.ai gives you a directional read in minutes with no setup at all.