Somewhere in a shared drive near you sits a slide called "Marketing Mary". She is 34, lives in a suburb, juggles two kids and a career, drinks oat-milk lattes, and "values authenticity". Ask who interviewed her and the room goes quiet. Mary was assembled in a workshop, from a template, in about forty minutes.

Then she gets handed to whoever writes the ads. This is where it gets expensive, because you cannot write a hook for a person who does not exist. I have tried. The copy comes out generic because the input was generic, and the account learns nothing because there was never a hypothesis to test.

The takeaways

A persona earns its place in paid ads only when every claim in it traces back to a real quote. My floor is 3 verbatim quotes per persona.
4 to 6 personas per project is the working range: enough to cover Eugene Schwartz's five awareness stages, few enough that each one gets a properly funded creative test.
Persona demographics are almost always estimates. Label them that way, and treat customer vocabulary as the part that does the targeting.

What makes a buyer persona useful for paid ads?

A persona is useful for paid ads when it changes what you ship: the hook you write, the objection the ad answers, the words in the first line. If two media buyers could read the persona and still write the exact ad they had in mind before, the document is decoration. That is the whole test.

Template personas fail it because their contents do not write copy. "Values authenticity, aged 30 to 45" gives you nothing to type. A sentence a customer left under a one-star Amazon review ("I gave up after the third sync failed") gives you a hook, a pain, and the customer's own vocabulary in one line. Buyer persona research for paid ads is the work of collecting enough of those lines and arranging them into a handful of people you can write for.

Where the raw material comes from

Five places reliably hold the language buyers use when nobody is marketing at them: Reddit threads, Amazon reviews, App Store reviews, Google Play reviews, and the niche forums of your category. Seed the search with your own product and your competitors', because the angriest, most quotable material usually sits under a rival's one-star reviews.

Each quote you keep needs more than the words. Note the emotion behind it, the register (clinical, frustrated, jokey), and 2 to 4 phrase markers: the exact word combinations you would never have invented yourself. Mining those sources well is its own craft and deserves its own post. Here I am taking the pile of quotes as given and dealing with the harder question: how do you compress a few hundred of them into personas you can run a campaign on?

How many buyer personas do you need for paid ads?

Four to six per project. The range comes from how paid ads consume personas: each one has to map to a distinct angle you will put media spend behind, so the count is bounded on both sides.

The lower bound is awareness. Eugene Schwartz laid out five stages of buyer awareness in Breakthrough Advertising back in 1966, from Unaware to Most Aware, and an ad that works on one stage usually dies on another. With only three personas you end up averaging stages together, and the copy averages with them.

The upper bound is budget arithmetic. Every persona implies its own hooks, its own creative test, its own minimum spend before results mean anything. Eight personas on, say, a 10,000 euro monthly budget means no test ever reaches a readable sample. Six is where I stop.

Three rules that keep a persona honest

Every persona needs at least three verbatim quotes. Three is a floor. If you cannot find three separate people describing the same pain in their own words, the persona is a guess wearing a name badge. The quotes then double as the brief: I paste them at the top of every creative request for that persona.

Keep the vocabulary attached. The phrase markers travel with the persona into production. When a quote says "I felt like I was paying a subscription to feel guilty", that phrasing, or something within arm's reach of it, belongs in the hook. Paraphrasing it into marketing language is how the evidence leaks back out of the document.

Rank pains by intensity, not just frequency. Fifty mild mentions of a clunky login do not outrank one furious, detailed story about a refund that took six weeks. Frequency tells you what is common. Intensity tells you what people will stop scrolling for.

Should a buyer persona include demographics?

Yes, but label them as the estimates they are. Almost nobody states their age, income, or family situation in a product review, so any demographic line in a research-built persona was inferred from context. That is fine as orientation. Presented as a measured fact, it is a small lie at the base of the whole document.

It also matters less for delivery than it used to. In Meta's broad-targeting era the algorithm finds the people; your creative is the targeting layer. The demographic sketch helps a designer picture the scene in a static, and that is roughly the extent of its job. The vocabulary, the dominant pain, and the awareness stage carry the campaign.

I built this rule into Adscalr as a label: the audience module assembles 4 to 6 personas per project, each backed by at least 3 real quotes from those 5 sources, and every demographic field is marked AI-estimated. Putting the uncertainty on the label cost nothing, and it means nobody has to wonder which parts were measured.

From persona deck to live campaign

The deliverable from buyer persona research for paid ads is small: 4 to 6 named people, each with quotes, a pain ranked by intensity, an awareness stage, and demographic guesses labelled as guesses. Small is the point. A document that fits on one screen gets used; a 40-page persona report gets admired and ignored.

If you want to see the assembled version, the audience intelligence side of Adscalr runs this pipeline end to end: it pulls quotes from the five sources, groups them into themes, maps them to Schwartz's awareness stages, and hands you personas with the evidence still attached. And once the persona-led creatives are live, the discipline moves downstream to reading the results without fooling yourself, which I covered in how to tell a winning ad from a lucky one.