Customer research and personas that actually help

0:00 Customer research in a week

Here's a story that happens every week. A founder launches a product, writes ads about the thing she's proudest of, spends a month's budget, and gets almost nothing back. The product was fine. The ads were aimed at an imaginary customer. In this lecture you'll learn how to replace guesses with evidence in about a week, even with no research budget. You'll learn four sources of insight, the five interview questions that actually work, how to build a persona that fits on one card, and how to use AI to speed it all up without breaking privacy rules or inventing customers.

0:44 Why research pays

Why does this matter so much? Because the best ad copy you'll ever write is sitting in your customers' mouths. When a customer says, I just wanted lunch sorted without thinking, that sentence will beat anything a copywriter invents, because it's how the next customer thinks too. Research also saves money. Every ad you run to the wrong people, with the wrong message, is budget you can't get back. So think of research as the cheapest marketing spend you'll ever make. A week of conversations can change which channels you choose, what you say and even what you charge.

1:27 Four sources of insight

Let's look at the four sources. First, customer interviews: five to ten conversations of about twenty minutes. They tell you why people bought and what nearly stopped them. Second, review mining: reading reviews of your business and your competitors on Google, Amazon, Noon, Daraz, Trustpilot or the app stores. Third, search and social listening: Google Trends, Keyword Planner, TikTok search, Reddit and community groups show what people ask. And fourth, surveys and on-site polls, which tell you how common something is. Here's the key idea. Interviews and reviews explain why. Search data and surveys tell you how many. You need both.

2:11 Five interview questions

Now the interview itself. The golden rule is: ask about the past, not the future. People are terrible at predicting what they'll do, and they're polite, so they'll say yes, I'd definitely buy that. Instead, try these five. One: take me back to when you first started looking for something like this, what was happening? Two: what did you try before, and what was frustrating? Three: what almost stopped you from buying? Four: what happened after you started using it? And five: if we disappeared tomorrow, what would you do instead? That last one reveals your real competition, which is often not another brand at all.

2:57 Jobs to be done

Behind those questions is a simple idea called jobs to be done. People hire a product to make progress in a situation. A nurse in Riyadh on twelve-hour shifts doesn't want a meal-prep subscription. She wants to stop worrying about lunch at work. A parent in London doesn't want Arabic lessons. They want their child to read confidently before a family trip. When you understand the job, you find angles your competitors miss, because they're all describing features. So in every interview, listen for three things: the situation, the struggle and the progress they wanted.

3:38 Example 1: UK Arabic tutor

Let's make this concrete with a simple example. A London-based online Arabic tutor assumed parents cared most about native-speaker teachers. That's what every competitor says too. She ran eight short parent interviews. The trigger turned out to be something specific: a child struggling to read fluently before a family visit abroad. And the biggest anxiety was, will my shy child actually speak on camera? Then she read competitor reviews and saw one complaint again and again: the teacher changes every week. So her new message became: the same teacher every week, patient with shy children. Same service. Completely different pitch. And the enquiries that came in were far better matched.

4:26 Example 2: one-card persona (illustrative)

Now let's turn research into a persona. Forget the stock photo and the invented hobbies. A useful persona fits on one card and every line comes from evidence. Here's a realistic one for a Riyadh meal-prep service, and treat it as illustrative. Name: Shift-worker Sara. Situation: nurse on rotating twelve-hour shifts, sharing a flat. Job: eat properly at work without thinking about it. Triggers: a new rota, or a week of expensive delivery apps. Anxieties: will it still be fresh by two p.m., and is it clearly labelled halal? Alternatives: delivery apps, the canteen or skipping lunch. Her words: sorted, no thinking, not greasy. And where she spends attention: Snapchat, TikTok and WhatsApp groups with colleagues.

5:16 What the card decides

Look at how much that card decides for you. Channels: Snapchat, TikTok and WhatsApp. Hooks: lunch sorted, no thinking. Proof you must show: the packaging, the halal labelling, the delivery time. Timing: when rotas come out. That's the power of a good persona. And if you can't point to evidence for a row, mark it as an assumption to test. Now, how does AI fit in? Tools like ChatGPT, Claude, Gemini and Perplexity are brilliant at organising research. Paste anonymised reviews and ask them to cluster themes, count mentions and pull exact quotes. They can draft interview questions too, which you then check for leading wording.

6:02 AI safety rules

But there are rules. First, never paste personal data, names, phone numbers, order IDs, into a consumer AI tool unless your company has approved that tool and your privacy notice covers it. Anonymise first. Second, don't let AI invent your customers. A synthetic persona made from nothing is a guess in a nicer font. Use it to generate ideas, then test them with real people. Third, models can miscount and occasionally make up quotes, so check every quote against the source before it goes into an ad. The full prompt I use for review mining is in the lesson text, ready to copy.

6:47 Common mistakes

Let's cover the classic mistakes quickly. Interviewing friends and family instead of real buyers. Asking hypothetical questions like, would you pay twenty pounds? Building personas from demographics only, like women aged twenty-five to thirty-four, with no situation or motivation. Treating AI personas as evidence. And doing research once, then never again. Markets move. Revisit it every quarter, or whenever results change. And how do you know research is working? Your ads start using customer language, hooks drawn from reviews earn better click-through, and your DMs have fewer is this for me questions.

7:27 Recap and try this now

Quick recap. Research replaces an imaginary customer with a real one. Use interviews and reviews to learn why, and search and surveys to learn how many. Ask about the past, listen for the situation, the struggle and the progress, and capture exact words. Summarise it on a one-card persona backed by evidence. Use AI to organise, never to invent, and never with personal data you haven't cleared. Here's your try this now. This week, talk to three recent customers or read fifty reviews, yours and a competitor's. Fill in the one-card persona, and highlight the three phrases you'd put straight into an ad.