What is voice of customer?

Voice of customer research is the systematic collection of what customers and prospects actually say about their problems, in their own words. In CRO its value is twofold: it supplies the hypotheses behavioral data cannot generate, and it supplies the vocabulary that makes copy resonate.

Which sources are worth mining?

Ranked by value per hour spent, and the top two are usually free and already sitting in the business.

Sales call recordings. The highest-value source available to B2B companies and the most neglected. Objections stated in the prospect's language, the questions that recur, and the moment a deal turns. Most companies record every call and mine none of them.

Support tickets and chat logs. Where confusion lives, in the customer's phrasing.

Won and lost deal notes. Why people bought and why they did not, though lost-deal reasons are filtered through the salesperson's interpretation.

On-site surveys. The only source that reaches people who are on the site right now rather than after the fact. One open question placed at the right moment returns more usable material than a long survey nobody completes, and the placement decides which population you hear from. See survey design.

Customer interviews. Deepest signal, highest cost, and the source that saturates fastest. New interviews stop producing new themes sooner than most teams plan for.

Review sites and community threads, including competitors' reviews, which describe what your market dislikes about the alternatives.

Internal site search queries, which show the vocabulary visitors expect you to use.

How does customer language become copy?

The mechanism is transcription, and the discipline is refusing to paraphrase.

Collect verbatim phrases. When several customers independently describe a problem as "we could not tell whether SEO was working", that sentence belongs on the page in that form. Rewriting it as "attribution challenges in organic search" translates it out of the register that made it recognizable.

Three specific applications:

Headlines from objections. The most common objection, stated and answered, outperforms a feature claim.

FAQ sections from real questions. Not questions invented by the marketing team. Real ones, from sales calls, in the words used.

Proof aimed at stated fears. A testimonial that names the specific worry buyers hold does more than one praising the team.

The deliverable of voice of customer work is a file of quoted sentences. A summary of what customers said has already destroyed the thing that made it useful.

What are the common mistakes?

Asking customers what they want. People are unreliable predictors of their own behavior. Ask what they did, what they tried, and what nearly stopped them. Behavior and history are recoverable, preference is not.

Surveying only successful customers. The people who did not buy hold the information about why the page failed, and they are harder to reach, which is why they get skipped.

Leading the interview. "Was the pricing page helpful" gets agreement. "Walk me through what you did after you landed on pricing" gets an account you can act on.

Treating it as a substitute for testing. Voice of customer generates hypotheses. Whether a change built on one works is a separate question, answered by a test.

Sanitizing the language. Marketing tends to smooth customer phrasing into brand voice, and the specificity that made it work is exactly what gets removed.

Deeper reading: Building data-backed CRO hypotheses. Service: Conversion Rate Optimization.

FAQ

How many customer interviews do you need?

Six to eight recent buyers usually reaches saturation for a specific question, meaning new interviews stop producing new themes. Broader questions need more.

How many people have to say something before it belongs on the page?

There is no threshold. A phrase used unprompted by several people in the same segment is a pattern worth putting in front of visitors. A vivid phrase from one enthusiastic customer is an anecdote. The cheap check is reading it back to whoever runs the sales calls.

Is voice of customer research the same as market research?

They overlap. Market research typically addresses sizing, segmentation, and positioning. Voice of customer work in CRO focuses narrowly on the language and objections of people already in your funnel.

Ask AI about this term

I just read https://www.karpi.studio/cro-glossary/voice-of-customer about mining what buyers actually say. Help me apply this to my own customer research: - The source ranking by value per hour: sales call recordings first, the highest-value and most neglected source, then support tickets, deal notes, on-site surveys, and interviews, which saturate after six to eight recent buyers - The discipline that makes it work: collect verbatim sentences and refuse to paraphrase, since rewriting a customer's phrasing into marketing register removes the recognizability that made it useful - Where the language lands: headlines from the most common objection stated and answered, FAQ sections from real sales call questions, and proof aimed at stated fears - The questions that return evidence: what they did, what they tried, and what nearly stopped them, never what they would want, since people predict their own behavior unreliably - The population I am skipping: people who did not buy hold the information about why the page failed, and they are harder to reach - The most common mistake: delivering a summary of what customers said, which has already destroyed the thing that made it useful

Want more revenue from your existing traffic?

We run CRO for Webflow sites — from audit to A/B testing.

Work with us