How to test messaging on personas

A persona from Hiasynth is not a workshop invention. It is a composite of a real cluster in the population, with a number behind it: this many people, this share of your audience. This guide shows how to sample them and put them to work. One call to sample_persona returns two, for 3 credits.

Ask for two personas

Name the audience and the place, as narrow or as wide as you like: "families with young children in Bavaria who own an electric car", or simply "people in Porto". Then say what the second persona is for.

Similar gives you the next most common type, a second voice from the same crowd.
Contrasting gives you the part of your audience the first persona doesn't speak for.

Give me two contrasting personas for people aged 30 to 45 in the top income quartile in Bavaria.

Read what came back

A size. How many people the persona stands for and what share of the audience. One that stands for 37% of your market deserves more of your copy than one that stands for 6%.
Who they are. Median age, sex, education, work, children and commute, with income and wealth in local currency.
What they value, in words. "Notably below average on life satisfaction, notably above on enjoying life."
How to find them. Media diet, digital engagement, the affinities that over-index, and the regions where this type is most common.
How they live. When you name a place: the spending categories where this persona's household is well above or below the national average, its diet, and its carbon footprint with flying, driving, heating and food split out. It is what lets a persona answer "would you pay for this?" from a budget and not from a mood.

Put them to work

Ask your assistant to stay in character and react:

Take the first persona. Read this landing page as them and tell me where you stop reading, and why.

Here are three subject lines. Which would each persona open? Answer separately, in their voice.

As the second persona, what would make you distrust this offer?

Your assistant does the role-play. Hiasynth supplied the facts that keep it honest: what this person earns, values, pays for and cares about.

Going further

Pair it with the reach profile. Plan reaching an audience tells you which values to lead with. The personas tell you how that lands with the two largest types in the audience.

Make the contrast yourself. When an audience is very uniform, the two personas can come back close to each other, which is itself worth knowing. For a sharper contrast, split the audience on what you care about, renters and owners, or under and over 40, and sample each half.

Compare places. The same audience in Milan and in Naples gives different personas. Useful when one creative has to work across markets.

A small panel. Sample three or four audiences, keep the personas in the conversation, and run every draft past all of them. It doesn't replace research with real people, but it removes the obvious misses before you pay for that research.

Worth knowing

A persona is a composite. The medians and most common values of a cluster of similar synthetic people. It is never one record, and there is no real person behind any of it.
It is not a count. For "how many", size the market.
It needs people behind it. At least 100 per persona, and more for an area smaller than a municipality. The answer tells you when to widen.
What it says in character is your assistant's. The facts are from the population. The voice, the objections and the reactions are a language model reasoning from those facts. Treat them as a well-briefed colleague's guess and not as a survey result.

That is the whole loop: size it, choose where, understand the place, plan the reach, test the message. When an answer surprises you, Reading an answer explains every number in it.