--- title: "Voice of customer: surveys, interviews and reviews" description: "The customer's words are your best copy Voice-of-customer (VoC) research captures motivations, objections and language directly from customers and…" url: https://optimizeall.com/learn/conversion-rate-optimization/voice-of-customer updated: 2026-10-05 --- Conversion Rate Optimization (CRO) · Conversion research · lesson 6 of 20 · 11 min # Voice of customer: surveys, interviews and reviews ## The customer's words are your best copy **Voice-of-customer (VoC) research** captures motivations, objections and language directly from customers and prospects. It powers both hypotheses and copy: when a page uses the exact phrases customers use to describe their problem, it tends to resonate more than internal jargon. ## Sources of VoC | Source | Best for | Tip | |---|---|---| | On-site polls (exit or on-page) | Why visitors hesitate right now | One short open question | | Post-purchase surveys | Why buyers chose you; what almost stopped them | Send soon after purchase | | Customer interviews | Deep motivation, context, alternatives considered | 20–40 minutes, recorded with permission | | Support tickets, chat and WhatsApp logs | Recurring questions and confusion | Tag themes monthly | | Sales call notes | Objections in B2B | Ask sales for the top five | | Product and competitor reviews | Language, desired outcomes, pain points | Mine 2–3 star reviews for nuance | | Social comments and community threads | Unfiltered opinions | Look for repeated phrases | ## Writing survey questions that work Good questions are open, neutral and specific: ``` Exit poll (on pricing page, after some time or exit intent): "What, if anything, is stopping you from starting today?" Post-purchase: "What nearly stopped you from buying?" "What made you choose us over other options?" "How would you describe [product] to a friend?" Avoid: "Would you like cheaper prices?" (everyone says yes) "How much do you love our site?" (leading) "Rate our checkout 1-10" (no actionable reason) ``` Keep surveys very short — one or two questions on-site, a few more post-purchase. Offer a clear purpose and respect consent and privacy obligations when collecting responses. ## Jobs to be done (JTBD) The **jobs-to-be-done** lens asks what "job" the customer hires your product to do. A useful interview structure explores the timeline of the decision: 1. **First thought** — when did you first realise you needed something? 2. **Trigger** — what happened that made you start looking seriously? 3. **Alternatives** — what else did you consider, including doing nothing? 4. **Anxieties** — what worried you about switching or buying? 5. **Decision** — what finally made you choose? These answers map directly to page content: the trigger informs headlines, alternatives inform comparison sections, anxieties inform FAQs and guarantees. ## Analysing qualitative responses 1. Export responses into a sheet. 2. Read 50–100 responses and create a first set of theme codes (for example: price, delivery time, trust, sizing, missing feature). 3. Code all responses; allow multiple codes per response. 4. Count theme frequency and pull representative quotes. 5. Link each major theme to a page element and a hypothesis. ``` | Theme | Count | Example quote | Page element | |----------------|-------|-------------------------------------------------|-------------------------| | Delivery time | 38 | "Didn't know if it'd arrive before Eid" | Delivery estimate on PDP| | Sizing | 27 | "Not sure about fit, sizes run small elsewhere" | Size guide + fit reviews| | Trust | 15 | "Never heard of the brand" | Reviews, press, returns | ``` ## Worked example: a B2B software company A UK-based HR software company runs interviews with eight recent customers. A repeated theme: buyers were triggered by a failed audit or a compliance scare, and their biggest anxiety was data migration from spreadsheets. The pricing page mentions neither. Hypotheses: a headline addressing compliance confidence, and a "we migrate your data for you" section with a timeline, will increase demo requests from pricing-page visitors. ## Hands-on: coding open-ended survey responses with AI and a codebook Coding hundreds of answers by hand is slow; letting an AI invent categories is unreliable. The middle path is a **codebook** you define, applied by AI, and checked by a human. 1. Read 50 responses yourself and draft 6–10 codes with definitions (e.g. `DELIVERY_UNCERTAINTY: unsure when the order arrives or whether it will arrive before a date`). 2. Remove personal data, then prompt: ```text Apply this codebook to each survey response. A response may have 0-3 codes. Codebook: DELIVERY_UNCERTAINTY - ... FIT_UNCERTAINTY - ... PRICE_VALUE - ... TRUST_PAYMENT - ... OTHER - use only if nothing else fits; give a 3-word label Return CSV: response_id,codes,short_quote (exact words from the response, max 12 words) Do not invent quotes. Do not infer demographics. Responses: [paste id + text] ``` 3. **Check agreement**: code 40 responses yourself without looking at the AI's output, then compare. If you disagree on more than roughly one in five, refine the definitions and rerun. 4. Count codes and pull quotes for your findings log. ## Jobs to be done interview script (30 minutes) ``` 1. Set the scene: "Take me back to the day you first started looking for something like this." 2. Push: "What was going wrong? What happened that made it urgent then?" 3. Pull: "What were you hoping would be different?" 4. Anxieties: "What worried you about switching or buying?" 5. Habits: "What were you doing instead? What almost kept you there?" 6. Decision: "What finally made you decide? Who else was involved?" 7. After: "What's different now? Anything disappointing?" ``` Record (with consent), transcribe, and tag answers to the four forces (push, pull, anxiety, habit). The anxieties and habits are usually where conversion copy and UX changes come from. ## Before and after: survey questions | Before | Problem | After | |---|---|---| | "How satisfied are you with our website?" | Vague; no action | "What almost stopped you from ordering today?" | | "Would you like faster delivery?" | Leading; everyone says yes | "How did you decide which delivery option to choose?" | | "Rate our checkout 1-10" asked to everyone | No context | Post-purchase: "Was anything confusing at checkout? (optional)" | ## Common mistakes - Asking what customers want instead of what they experienced. - Long surveys that few complete. - Counting only the loudest complaints rather than coding systematically. - Ignoring non-buyers — their objections are often the most valuable. - Paraphrasing customer language into corporate jargon. ## Recruiting interviewees Recent customers (who bought in the last month or two) remember their decision most clearly. Invite them by email with a short, honest request and, where appropriate, a modest thank-you such as a voucher. Include some people who **did not** buy — trial users who did not convert, or leads who went silent — because their objections are often the most instructive. Always ask permission to record, explain how notes will be used, and store recordings securely with limited retention. ## VoC checklist - [ ] At least one on-site poll live on a key page - [ ] Post-purchase survey with "what nearly stopped you?" - [ ] Five or more customer interviews per quarter - [ ] Support and sales themes reviewed monthly - [ ] Themes coded, counted and linked to hypotheses ## Video lecture: Voice of customer: surveys, interviews and reviews Lecture coming soon · 14 chapters · about 9 minutes. Read the full transcript below. 1. Voice of customer 2. Why VoC matters 3. Sources of VoC 4. Survey questions that work 5. Jobs to be done 6. Simple example: Abu Dhabi furniture 7. Mining competitor reviews 8. Realistic example: London clinic software (illustrative) 9. Watch me do it: AI coding with a codebook 10. Recruiting interviewees 11. The 30-minute JTBD script 12. Common mistakes 13. Close the loop 14. Recap ## Lecture transcript ### Voice of customer The most persuasive words for your website are sitting in your customers' heads. Voice of customer research is how you get them out: surveys, interviews, reviews and support conversations. In this lecture you'll learn the best sources, how to write survey questions that produce actionable answers, the jobs-to-be-done lens, and how to analyse hundreds of open-ended answers with a codebook, AI and a human check. You'll also see a thirty-minute interview script you can use this week. ### Why VoC matters Why does voice of customer matter for CRO? Because it explains motivation and hesitation in the customer's own words. Analytics says people leave the checkout. Recordings show them pausing at delivery options. Voice of customer tells you they're worried the order won't arrive before a wedding on Saturday. That last piece is what turns a vague problem into a precise hypothesis, and it often gives you the exact language to fix it. ### Sources of VoC Where do you find it? On-site polls at key moments, like on a product page: what's stopping you from adding this to your bag today? Post-purchase surveys: what almost stopped you from ordering? Exit surveys for people who cancel. Customer interviews. Reviews, yours and competitors'. Support tickets, live chat and WhatsApp conversations. And sales call notes for B2B. Each source captures a different moment, so combine them. ### Survey questions that work Writing survey questions is a skill. Ask about behaviour and moments, not general satisfaction. How satisfied are you with our website? produces a number you can't act on. What almost stopped you from ordering today? produces reasons. Avoid leading questions: would you like faster delivery? Everyone says yes. Instead: how did you decide which delivery option to choose? Keep surveys short. One to three questions on-site, open-ended where you want language, and ask at the moment the experience is fresh. ### Jobs to be done Now a powerful lens: jobs to be done. People don't buy products. They hire them to make progress in a situation. Four forces shape the decision. Push: what's going wrong now. Pull: the better future they imagine. Anxiety: worries about the new option. And habit: the comfort of what they already do. For conversion, anxieties and habits matter most, because they're what stop people who already want to buy. Your job is to reduce anxiety and make switching feel easy. ### Simple example: Abu Dhabi furniture A simple example. An online furniture store in Abu Dhabi adds a post-purchase question: what almost stopped you from ordering today? Out of the first hundred and fifty answers, the most common theme is assembly. Will someone help me put it together? The second is delivery to upper floors in towers. Neither was mentioned anywhere on product pages. The team adds a clear line on each product page about assembly service and delivery to any floor, with a link to details. That's a hypothesis built directly from customers' words. ### Mining competitor reviews Don't forget reviews, especially your competitors'. They're free, public and brutally honest. Read the three-star reviews first, because they're balanced and specific: the product was good, but delivery took ten days and nobody answered WhatsApp. Collect a hundred or so across your top three competitors on marketplaces, Google Maps and app stores. Tag them with the same codebook you use for surveys. You'll often find an anxiety the whole category ignores, like returns being a hassle, or sizes running small. If you can solve that anxiety and say so clearly, you have both a conversion improvement and a differentiator. Just remember: use reviews for insight, not for copying people's words into your ads. ### Realistic example: London clinic software (illustrative) Now a realistic scenario with illustrative details. A B2B software company in London sells scheduling tools to clinics. Trial sign-ups are healthy, but few convert to paid. The team runs six jobs-to-be-done interviews with recent buyers and four with people who trialled but didn't buy. The pattern is clear. Buyers had a push, a receptionist leaving, and the trial made it easy to import existing appointments. Non-buyers had a habit: their current paper diary worked well enough, and importing looked like hard work. The CRO opportunity isn't the pricing page. It's a guided import in the first day of the trial, and copy that speaks to switching without losing a single booking. ### Watch me do it: AI coding with a codebook Watch me code three hundred survey answers with AI, safely. First, I read fifty answers myself and draft a codebook: eight codes, each with a definition, like delivery uncertainty, fit uncertainty, price and value, and trust in payment. I remove personal data. Then I give the AI the codebook and ask it to apply zero to three codes per response, return a CSV with a short exact quote, and never invent quotes. Now the key step. I code forty responses myself without looking, and compare. If we disagree on more than about one in five, I sharpen the definitions and rerun. Only then do I count codes and pull quotes. ### Recruiting interviewees Recruiting interviewees is easier than people think. Recent customers are the best source, because their memory of the decision is fresh. Offer a small thank-you, like a voucher or a donation, and be clear about the time: thirty minutes. Include people who didn't buy, like cancelled trials or abandoned quotes, because their reasons are often the most useful. Aim for five to eight per segment. Record with consent, explain how you'll use the recording, and store it securely. ### The 30-minute JTBD script Let's look at the interview script itself. Seven prompts, about thirty minutes. Set the scene: take me back to the day you started looking. Push: what was going wrong, and why was it urgent then? Pull: what were you hoping would be different? Anxieties: what worried you about switching? Habits: what were you doing instead, and what almost kept you there? Decision: what finally made you decide, and who else was involved? After: what's different now, and anything disappointing? Tag every answer to one of the four forces. The anxieties and habits usually hold your next hypotheses. ### Common mistakes Common mistakes. Asking vague satisfaction questions. Leading questions. Surveys that are too long. Only surveying happy customers. Letting AI invent categories and quotes. Treating one vivid comment as a trend. And collecting VoC once, then never again. Customer worries change with seasons, competitors and prices, so make VoC a monthly habit. ### Close the loop Finally, close the loop from voice of customer to the page. Every coded theme should end up in one of three places. As a hypothesis in your backlog, like adding assembly information to product pages will reduce hesitation for furniture buyers. As copy, using the customers' own words in headlines, reassurance lines and FAQs. Or as an operational fix, because some anxieties can't be solved with words, like genuinely slow delivery. Write down which theme went where, and when you'll check whether it helped. That's how research stops being a report nobody reads and becomes a steady source of wins. ### Recap Recap. Voice of customer explains why people buy and why they hesitate, in their own words. Use polls, surveys, interviews, reviews and support conversations. Ask about moments and behaviour, not satisfaction. Use the jobs-to-be-done forces, and focus on anxieties and habits. Code open answers with a codebook, apply it with AI, and check agreement yourself. Try this now: add the question, what almost stopped you from ordering today, to your order confirmation page, and book two thirty-minute interviews with recent customers using the script in the lesson text. ## Key takeaways - Voice-of-customer research reveals motivations, objections and the words customers use. - Ask open, neutral, specific questions; keep surveys very short. - Jobs-to-be-done interviews map triggers, alternatives and anxieties to page content. - Code responses systematically and link themes to hypotheses. ## Try it Launch a one-question poll on a key page (or review 50 existing reviews) and code the responses into themes with counts and quotes. - [Previous: Qualitative research: heatmaps, recordings and user testing](https://optimizeall.com/learn/conversion-rate-optimization/qualitative-research) - [Next: UX research methods: card sorts, tree tests and prototype testing](https://optimizeall.com/learn/conversion-rate-optimization/ux-research-methods) - [All lessons of Conversion Rate Optimization (CRO)](https://optimizeall.com/learn/conversion-rate-optimization)