---
title: "Audience research and voice-of-customer mining"
description: "The best copy is found, not invented Experienced copywriters spend more time researching than writing. The reason is simple: customers describe their…"
url: https://optimizeall.com/learn/copywriting-that-converts/audience-research
updated: 2026-10-05
---

Copywriting That Converts · Copy foundations and audience research · lesson 2 of 19 · 11 min

# Audience research and voice-of-customer mining

## The best copy is found, not invented

Experienced copywriters spend more time researching than writing. The reason is simple: customers describe their problems, desires and doubts in words that resonate with other customers. When you reuse that language, your copy feels like it is reading the reader's mind — because, in a sense, it is.

## What you are looking for

Collect evidence in five buckets:

| Bucket | Question | Example finding |
|---|---|---|
| Pains | What frustrates them today? | "I waste every Sunday planning meals" |
| Desired outcomes | What does success look like? | "Healthy food without thinking about it" |
| Triggers | What made them look for a solution now? | "Doctor told me to cut sugar" |
| Objections | What makes them hesitate? | "Meal kits always have too much packaging" |
| Language | How do they phrase it? | "Decision fatigue", "no-brainer dinners" |

## Where to find customer language

- **Reviews** of your product and competitors — on marketplaces, app stores, Google Maps and review platforms. Three-star reviews are especially useful: they are balanced and specific.
- **Support tickets, live chat and WhatsApp conversations** — recurring questions reveal missing information.
- **Sales call recordings or notes** — objections in the prospect's own words.
- **Customer surveys** — especially "What nearly stopped you from buying?" and "How would you describe us to a friend?"
- **Interviews** — 20 to 30 minutes with recent customers.
- **Communities and social comments** — forums, Reddit threads, Facebook groups, YouTube and TikTok comment sections in your niche.
- **Search data** — questions people type into search engines ("how to…", "is X worth it").

## A step-by-step research sprint (one to two days)

```
1. Define the reader: product, segment, market (e.g. new mothers in the UAE buying baby formula online).
2. Collect 100-200 snippets from reviews, comments, tickets and surveys into a spreadsheet.
3. Tag each snippet: pain / outcome / trigger / objection / language.
4. Highlight phrases that are vivid, specific or repeated.
5. Count themes: which pains and objections appear most often?
6. Write a one-page "message bank": top 5 pains, top 5 outcomes, top 5 objections,
   20 verbatim phrases.
7. Share it with everyone who writes for the brand.
```

## The message bank template

```
READER: Freelance designers (UK, Pakistan, UAE) who invoice international clients

TOP PAINS
- "Chasing late payments is awkward"
- "Currency conversion eats my fee"

TOP OUTCOMES
- "Get paid on time without nagging"
- "Look professional to big clients"

TOP OBJECTIONS
- "Another subscription I'll forget to cancel"
- "Will clients trust a payment link?"

VERBATIM PHRASES
"awkward follow-ups", "invoice ghosting", "look like a real agency", "fees eat my margin"
```

## Interview questions that uncover gold

- "Take me back to when you first realised you needed something like this. What was happening?"
- "What did you try before? What didn't work about it?"
- "What almost stopped you from buying?"
- "What would you say to a friend who is considering it?"
- "What result have you noticed since using it?"

Ask follow-ups such as "Tell me more about that" and "What do you mean by…?". Avoid leading questions.

## Worked example: a meal-prep company in Dubai

A healthy meal-delivery brand assumed customers cared most about calories. Reviews and a short survey told a different story: the repeated phrases were "no time to cook after work", "stop ordering junk", and "Ramadan iftar sorted". The team rewrote their homepage headline from *"Calorie-controlled meals, delivered"* to *"Stop ordering junk after work — fresh, balanced meals delivered daily"*, and created a separate seasonal landing page for Ramadan. The new language came straight from customers.

## Respect privacy when researching

Use public reviews and comments for insight, not for copying identifiable content into ads. If you want to quote a customer by name, get their permission. Store interview recordings securely and use them only for the purpose you told participants about.

## Hands-on: AI-assisted review mining (with human checks)

Large language models are excellent at the tedious part of research — sorting hundreds of snippets — and poor at judging which insight matters. Use them to **tag and cluster**, then verify by reading.

1. Export reviews, survey answers or support tickets to a spreadsheet (one snippet per row). Remove names, emails, order numbers and phone numbers first.
2. Paste a batch (for example 100 rows) into your AI assistant with this prompt:

```text
You are helping with voice-of-customer research for [product] sold to [audience] in [market].
For each snippet below, return a table with columns:
id | bucket (pain / desired outcome / trigger / objection / language) | theme (2-4 words) | verbatim phrase worth reusing (copy exactly, do not paraphrase)
Rules:
- Quote the customer's exact words in the last column. Never invent or "improve" wording.
- If a snippet fits no bucket, write "none".
- Do not infer age, gender, health, religion or other personal characteristics.
After the table, list the 5 most frequent themes per bucket with counts.
Snippets:
[paste rows with ids]
```

3. **Spot-check at least 20% of rows** yourself. Models sometimes paraphrase while claiming to quote; search the source for each "verbatim" phrase.
4. Paste the verified output into your message bank.

If you prefer code, a few lines of Python count themes once they are tagged:

```python
import csv
from collections import Counter

counts = Counter()
with open("tagged_snippets.csv", newline="", encoding="utf-8") as f:
    for row in csv.DictReader(f):          # columns: id,bucket,theme,phrase
        if row["bucket"] != "none":
            counts[(row["bucket"], row["theme"].strip().lower())] += 1

for (bucket, theme), n in counts.most_common(15):
    print(f"{bucket:16s} {theme:30s} {n}")
```

## Research with data-protection rules in mind

Uploading customer data to an AI tool is processing personal data. Under the UK GDPR, EU GDPR and similar laws (including Saudi Arabia's PDPL and the UAE's federal data-protection law), you need a lawful basis, a trustworthy processor and minimal data. The simplest safeguard is to **strip identifiers before upload** and use business or enterprise AI plans whose terms say your inputs are not used to train models — check your provider's current data-use terms.

## Worked example: before and after message bank

A Karachi-based online pharmacy (details illustrative) had a homepage built on the team's assumptions: "Genuine medicines, best prices." Mining 300 app-store reviews and WhatsApp support chats surfaced a different priority: "delivered before my father's next dose", "they called to confirm the prescription", "no fake medicines like the market". The new message bank led to a headline about **reliable delivery windows and pharmacist-verified prescriptions**, with "genuine medicines" demoted to proof. The research did not produce clever wording; it produced the right priority.

## Common mistakes

- Relying on internal opinions about what customers care about.
- Only reading five-star reviews (too vague) or one-star reviews (often about delivery issues).
- Paraphrasing customer language into corporate jargon.
- Researching once and never updating the message bank.
- Treating a single loud comment as a theme.

## Research checklist

- [ ] Reader clearly defined (segment, market, awareness level)
- [ ] 100+ snippets collected from at least three sources
- [ ] Snippets tagged and themes counted
- [ ] Message bank written and shared
- [ ] Permission obtained for any named testimonial

## Video lecture: Audience research and voice-of-customer mining

Lecture coming soon · 11 chapters · about 8 minutes. Read the full transcript below.

1. Audience research and voice of customer
2. Why research beats guessing
3. Five things to listen for
4. Where to find customer language
5. Simple example: Dubai meal delivery
6. Realistic example: Karachi online pharmacy (illustrative)
7. Watch me do it: AI review tagging
8. Common mistakes
9. Interviews: finding the 'why'
10. Build the message bank
11. Recap

## Lecture transcript

### Audience research and voice of customer

Here's a secret professional copywriters rarely admit. The best lines they ever wrote, they didn't write. They found them. In a review, in a support chat, in a sales call where a customer said something so vivid it stopped the room. In this lecture you'll learn how to find that language on purpose. You'll learn the five things to listen for, where to look, how to run a research sprint in a day or two, and how to use AI to sort hundreds of snippets without letting it put words in your customers' mouths.

### Why research beats guessing

Why bother? Because your customers explain your product better than you do. You know every feature. They know what it felt like before they found you. When you reuse their language, readers think, that's exactly my situation. When you use internal language, they think, this isn't for me. Internal jargon is the silent conversion killer. The team says omnichannel fulfilment. The customer says, I want to pick it up on my way home.

### Five things to listen for

Think of research like tuning a radio. Before you start, you hear static: opinions, assumptions, what the founder likes. As you collect real snippets, the station comes in. You listen for five signals. Pains, what frustrates them now. Desired outcomes, what success looks like. Triggers, what made them look for a solution today. Objections, what makes them hesitate. And language, the exact phrases they use. Every snippet you collect goes in one of those five buckets.

### Where to find customer language

Where do you find snippets? Start with reviews, yours and competitors'. Three-star reviews are gold, because they're balanced and specific. Then support tickets, live chat and WhatsApp threads, where repeated questions reveal missing information. Sales call notes give you objections verbatim. Surveys work if you ask, what nearly stopped you from buying? And communities: Reddit threads, Facebook groups, YouTube and TikTok comments in your niche. Aim for one to two hundred snippets from at least three sources. Tag each one, count the themes, then write a one-page message bank: top pains, outcomes and objections, plus twenty verbatim phrases.

### Simple example: Dubai meal delivery

A simple worked example. A meal-delivery brand in Dubai assumed customers cared about calories. So the headline said, Calorie-controlled meals, delivered. They read two hundred reviews and asked one survey question. The repeated phrases were: no time to cook after work, stop ordering junk, and iftar sorted. Not a single person mentioned calories first. The new headline: Stop ordering junk after work. Fresh, balanced meals delivered daily. And a separate Ramadan page for iftar. Nothing clever. Just the customers' words, in the customers' order.

### Realistic example: Karachi online pharmacy (illustrative)

Now a realistic business scenario, with illustrative details. An online pharmacy in Karachi has a homepage that says, Genuine medicines, best prices. The marketing lead exports three hundred app-store reviews and support chats, strips names and phone numbers, and tags them. The top pain isn't price. It's timing. Delivered before my father's next dose. The top trust signal is a phone call confirming the prescription. And the top objection is fear of fake medicines. So the message bank reorders priorities. The headline promises reliable delivery windows and pharmacist-verified prescriptions. Genuine medicines moves down to become proof. Same company, same facts. A completely different lead.

### Watch me do it: AI review tagging

Watch me do the AI-assisted part. I've got a spreadsheet of a hundred reviews, and I've removed names, emails and order numbers first. I paste them into my assistant with a prompt that asks for a table: bucket, a short theme, and a verbatim phrase copied exactly, never paraphrased. I also tell it not to guess anyone's age, health or religion. It returns the table and a count of the top themes. Now the important step. I pick twenty rows at random and search the original reviews for each so-called verbatim phrase. If one was quietly reworded, I fix it. The model sorted. I verified. That's the division of labour.

### Common mistakes

Common mistakes. Relying on internal opinions, what the founder thinks customers want. Reading only five-star reviews, which are too vague, or only one-star reviews, which are often about delivery. Paraphrasing customer language into corporate language, which throws away the whole value. Treating one loud comment as a trend. And with AI, trusting the summary without reading the source. One more: privacy. Uploading customer data to a tool is processing personal data. Strip identifiers first, and use a business plan whose terms say your data isn't used for training.

### Interviews: finding the 'why'

Reviews tell you what people say after buying. Interviews tell you why they bought. So let's talk about interviews, because twenty to thirty minutes with five recent customers can change an entire campaign. Ask about the story, not opinions. Take me back to when you first realised you needed something like this. What was happening? What did you try before, and what didn't work about it? What almost stopped you from buying? What would you say to a friend who's thinking about it? Then use two magic follow-ups: tell me more about that, and what do you mean by that? Avoid leading questions like, you liked the fast delivery, right? That just hands them your answer. Record with permission, take notes on exact phrases, and store the recordings securely, used only for the purpose you told them about.

### Build the message bank

Once you have your tagged snippets, turn them into a message bank. That's one page, not a fifty-slide deck. At the top, the reader: who they are, where they are, and how aware they are. Then the top five pains, top five desired outcomes and top five objections, each ranked by how often they appeared. Then twenty verbatim phrases worth reusing, copied exactly. Finally, a short list of claims you can prove, with the evidence next to each. Why one page? Because the message bank is a tool for everyone who writes for the brand: the founder, the freelancer writing ads, the support team drafting replies, and the AI assistant you'll brief later in this course. Put it where people will actually open it, and refresh it every quarter or whenever a new product launches.

### Recap

Let's recap. Great copy is found in your customers' words. Listen for pains, outcomes, triggers, objections and exact language. Collect from at least three sources, tag and count, and write a one-page message bank that everyone who writes for the brand can use. Let AI do the sorting, and let humans do the verifying and the judging. Here's your try-this-now. Pick one product. Collect fifty snippets today from reviews and support chats. Tag them into the five buckets. Circle the three phrases you'd put in a headline. The lesson text includes the full prompt, a short Python counter and a message bank template.

## Key takeaways

- Customer language, reused faithfully, makes copy resonate.
- Collect pains, outcomes, triggers, objections and verbatim phrases.
- Three-star reviews, support tickets and sales notes are rich sources.
- Summarise findings in a shared message bank and update it regularly.

## Try it

Run a mini research sprint: collect at least 50 customer snippets for a product you market, tag them, and write a one-page message bank.

- [Previous: What converting copy actually does](https://optimizeall.com/learn/copywriting-that-converts/what-converting-copy-does)
- [Next: From features to benefits to outcomes](https://optimizeall.com/learn/copywriting-that-converts/features-benefits-outcomes)
- [All lessons of Copywriting That Converts](https://optimizeall.com/learn/copywriting-that-converts)