RK Growth & Development

Build Your Customer Persona From People Who Already Buy From You

The customer persona method I actually use with clients: skip the demographics report, look at who really buys, and work backward from there instead.

EssayBy Rielly KeelerAugust 11, 2026

Most persona exercises start backward. A business pulls a demographic report, invents a fictional buyer with a stock photo and a made-up income bracket, gives her a name like “Marketing Mary,” and calls it strategy. Then nobody on the team looks at Mary again.

The better starting point is sitting in your own customer list already. Who actually buys from you? Not who you think should buy from you. Who has, this year, put money down.

Smashing Magazine’s research on this points at a specific, well-documented failure mode: personas built without real research, without interviews or actual behavioral observation, tend to end up full of details, a fictional name, a stock photo, a guessed income bracket, that never actually inform a real decision. The research itself has to come from somewhere real before the persona means anything. A good tip to get started: look at that buyer list mentioned earlier, specifically the postal codes of the people who are your clients. Then run those through Environics Analytics and look up their PRIZM clusters. This will give you a whole bunch of geodemographic information and at least a starting point with real data.

Wikipedia’s summary of UX persona research draws a distinction between “proto-personas,” which are assumption-based and used before you’ve done any real research, and true personas, built from it afterward. The sequencing matters: assumptions come first as a working guess, real data comes after and either confirms or replaces them. Always go straight to the real numbers. The assumption-based approach is only for new products and services, and it’ll just waste time otherwise. If you have data, paying customers, use it. Don’t guess. You’ll burn cash and time.

Then look at what you already have. Back to a Kingston heating and cooling company I work with: if you run a business like that, you don’t need to guess whether your customers are homeowners or renters. Your invoices already tell you. You don’t need to guess their age range. Your service call history tells you, close enough. The patterns are already sitting in the data you generate every week, and most small businesses never actually look at it before spending money on ads aimed at a guess instead.

This matters even more once you start running paid ads, because the platforms themselves aren’t targeting on demographics the way they used to. Meta’s Advantage+ and Google’s Smart Bidding both lean on behavioral signals, what people actually click, buy, and engage with, more heavily than static demographic data. Mailchimp’s own guidance on modern segmentation confirms the shift: targeting increasingly runs on behavior and preference, not just age and location. If the platform is already reading behavior, your persona should start there too, not with a stock photo and a made-up job title.

Both run on a feedback loop: who clicked the ad, and can the platform confirm who actually bought. That’s why proper tracking matters so much. If Google or Meta can’t see who actually converted, or doesn’t know what a real conversion looks like, it starts optimizing for the wrong things. One workaround, at least with Google, is uploading your customer list. This gives Google a confirmed signal of who already buys from you, and it works from that. This shows that both proper tracking and knowing your customer are important. We can only rely on the automation as much as we know our actual data. Garbage in, garbage out, in coder speak. If we let the algorithms work off flawed data, they’ll keep generating flawed results. Which is why the planning and understanding side of things is so important. If a business owner can describe their customer well, then their ads, content, whatever, will work, because they know who they’re selling to and why they buy.

None of this means demographic data is useless. It’s a filter, not a foundation. Start with who’s already buying, spot what they actually have in common, and let the platform’s own behavioral targeting do the rest of the work.

Pull last month’s invoices. That’s your first real persona draft, no stock photo required.