Strategy

Marketing personas: how to build one from real data

A persona is a short description of a typical customer, built on what your real customers say and do. It brings together what the customer wants to achieve, what's holding her back and which words she uses herself. In marketing, you use the persona to write messages and choose channels for a person instead of an average. Many personas end up in a drawer. They're called "Mette, 42", have two children and do yoga, and nobody on the team can say where the information comes from. That's a shame, because a persona built on data makes almost everything else in marketing easier. Here's how to make one that holds up, and how AI makes the work faster.
A member of the Amplify team sitting at a table with a cup of coffee
→ All articles

What is a persona in marketing?

A persona in marketing is a working tool for decisions about messages, channels and content. The concept comes from software design, where Alan Cooper made it popular in the late 1990s. In sales and marketing, many people call it a buyer persona, because it's about a purchase.

A useful persona answers six things:

  • The situation. What's happening when the customer starts looking?
  • The goal. What does she want to achieve, and what does success look like for her?
  • The objections. What makes her hesitate or choose someone else?
  • The criteria. What does she compare on: price, reassurance, time or quality?
  • The words. How does she describe the problem in her own words?
  • The channels. Where does she look, and who does she ask for advice?

Name, age and a photo are fine to include if they help the team remember her. The six answers do the work. Many of them are about buying motives, in other words why people buy.

What's the difference between a persona and a target audience?

A target audience is the group you want to reach and can measure. A persona is a person in that group you can write to. The target audience might be women aged 35 to 55 in the Capital Region of Denmark who have bought before. The persona tells you that she buys when she's promised to organise the Christmas lunch and doesn't have time to search for long.

Numbers describe the group. According to Statistics Denmark, 85 per cent of 16 to 74-year-olds had shopped online within the previous three months in 2026. The figure says how many people buy. It says nothing about why each individual does.

You need both. Segmentation and a target audience analysis divide up the market and show you where the money should go. The persona shows you what to say once you're there. In B2B, one target audience often has several personas. The person looking for a solution is rarely the one who signs.

How do you build a persona from real data?

You build a persona by gathering what customers say and do before anyone writes a word of the description. The order matters most. If you start with the description, you'll look for data that confirms it.

Most businesses have enough data in-house. Here are the five places I'd start.

Sales and CRM

Look at the people who have actually bought. Who are they, what did they buy first, and what triggered the purchase? Look at the lost quotes too. The reason people say no is often the most valuable line in the whole persona.

Customer service

The questions before a purchase show what's unclear. The complaints after a purchase show what you promised that didn't hold. Read through a month of enquiries, and write down the questions that keep coming up.

Reviews

Reviews give you customers' own words. The good ones tell you what people value, and that's rarely what's on your homepage. The bad ones show the objections. Read your competitors' reviews too. They show what customers miss with them.

Searches

Keywords show how people phrase the problem before they know you. Look at the search terms in Google Ads, the queries in Search Console and what people search for on your own website. According to Google's Analytics Help, Search Console stores data for the last 16 months. So there's plenty to read. A customer searching for price is in a different place from one searching for "best for".

Conversations

Talk to a handful of new customers and a couple who chose someone else. Ask about their situation when they started looking, what else they considered and what nearly stopped them. Ask about what happened. Never ask what they would do in the future, because nobody knows.

Once you have the material, sort it by the six questions from earlier. Only write what you have a source for. Two to three personas are enough for most businesses. If a persona doesn't change anything in what you say or do, merge it with another.

Why don't made-up personas work?

Made-up personas don't work because they confirm what the team already believes. You learn nothing new about customers, and the messages stay the same as before.

There are three signs of a made-up persona. Nobody can point to where the information comes from. The persona has hobbies and a stock photo and no objections. And nobody has changed it since someone made it.

AI has given the pitfall a new version. If you ask ChatGPT for a persona for your industry, you get a neat average of the internet. It sounds convincing and is just as made-up as Mette. The AI knows nothing about your customers beyond what you give it.

How can AI help with a persona?

AI helps most in two places: gathering customer quotes and testing messages against the persona. It does both quickly, and both require it to work on your own data.

Gather data faster

Put reviews, support enquiries, meeting notes and search terms into Claude or ChatGPT. Ask it to group the statements into goals, objections and words, and ask for a quote with each group. Then you can check it hasn't made anything up. What used to take a workshop with sticky notes, you can now do in an afternoon.

Remove names, e-mails and phone numbers before you paste in customer quotes. Use a version of the tool where your business is in control of what happens to the data.

Test messages against the persona

Give the AI the persona and the source material, and ask it to read your ad or landing page as the customer. What's unclear? Which objection goes unanswered? Which words would she never use herself? It's a quick first filter that sorts out the weakest messages.

It's still a simulation. The AI can point out gaps. It can't tell you what customers buy. The market does that. So let two or three messages run against each other in your ads, and let the result correct the persona.

How Amplify works with personas

Amplify uses the persona as a living part of every client's plan and marketing strategy. These are the steps we follow:

  1. The buyer in the strategy sprint. Amplify starts with who buys and what they want to achieve. In B2B, we write down two people: the one who reads the ad and the one who signs.
  2. A source on every line. Every piece of information about the buyer has a source, such as a meeting, a review or a keyword. Whatever we can't document is marked "missing" with a date. Never as a guess.
  3. Customers' own words in the brief. Before we write an ad, Amplify's AI agent pulls praise and objections from the client's Google reviews into the brief.
  4. One file everyone uses. The persona lives in the client's file. The agent reads it every time it helps with copy or an ad, and the specialists work from the same file.
  5. Copy matched to how warm the reader is. Someone searching gets the answer and the proof straight away. Someone who doesn't know the client gets the problem first. An existing customer gets the proof and the answer to their objection.
  6. The ads correct the persona. Every Meta ad gets an angle, such as reassurance, price or social proof. When an angle wins or loses, the agent writes one sentence about the buyer based on it, and that goes into the next brief.
  7. The plan reviewed every quarter. Every quarter, Amplify reviews the plan with the client. If the ads or meetings have shown something new about the buyer, we correct the description there and then.

The AI gathers and suggests. The specialists and the client decide what the persona says.

Frequently asked questions about personas

How many personas do you need?

Most businesses are fine with two to three. Each persona should change something in your messages, channels or offers. If it doesn't, merge it with another.

What's the difference between a persona and a buyer persona?

There's no real difference. Buyer persona is the term sales and marketing use, and it emphasises that the persona is about a purchase. In UX and product development, people more often talk about personas for the users of a product.

Can you create a persona with ChatGPT?

Yes, if ChatGPT works on your own data. Give it reviews, support enquiries and meeting notes, and ask for a source for every claim. If you ask it for a persona out of nothing, you get a guess that just sounds convincing.

How often should you update a persona?

Review it at least once a year, and every time you launch something new or enter a new market. If messages that usually work start losing in your ads, that's also a sign your customers have moved.

Your next step

Take your latest reviews and a month of customer service enquiries. Remove names and contact details. Ask an AI to group them into goals, objections and words, with a quote for each group. Hold the result up against the persona you have today. Whatever doesn't fit is your first fix.

If you'd like someone to challenge the persona afterwards, we're happy to sit down with you.

Want help with this?

→ Book a chat