Voice-of-customer research for copywriting

What voice-of-customer research is and why it matters

Most businesses describe what they sell in the language of the team: service names, features, process terms. Buyers describe what they want in the language of their situation: "my books are a mess before tax time", "I need someone who will actually call me back". When the two vocabularies differ, the visitor does not recognise themselves on the page.

Voice-of-customer research closes that gap. You collect real phrases from real buyers about three things: the situation that made them look, the outcome they wanted, and the doubts that nearly stopped them. Those phrases become your headlines, subheads, bullet points and objection answers.

It also protects you from the most common copy failure, which is writing what the team finds persuasive. Owners are the worst judges of their own page, because they already know what everything means.

Jobs to be done: the frame that makes the research useful

Asking "what do you want?" produces a wish list of features. A more useful question is "what were you trying to get done?" This is the core of jobs to be done, a theory associated with Harvard Business School professor Clayton Christensen. In a 2016 Harvard Business Review article, Christensen and three co-authors argue that companies focus too much on the product and not enough on why customers make the choices they do.

For a copywriter the idea is practical. People do not buy a service. They "hire" it to make progress in a particular situation. Your copy should name the situation, the progress and what stood in the way.

Bob Moesta, a practitioner credited with the method, teaches a way to find the job: run a switch interview with someone who recently bought, and reconstruct the timeline. His framework has four forces. Push (the pain of the current situation) and pull (the appeal of the new option) drive a switch. Anxiety about the new option and habit hold it back. As of his 2023 Business of Software talk, he says the anxiety and habit side is often the key to why people do not switch sooner.

Those four forces map neatly onto page copy. Push becomes your problem statement. Pull becomes your promised outcome. Anxiety becomes your objection answers and guarantees. Habit becomes your reasons to change now.

Where to find customer language

Use several sources. Each one has a different bias, and patterns that appear in more than one source are the ones to trust.

• Customer interviews: the richest source and the one you control. Aim for recent buyers, 30 to 45 minutes each. There is no verified magic number, so keep going until the answers start to repeat.
• Reviews and testimonials: yours and your competitors'. Read three- and four-star reviews as closely as five-star ones, because they name what was missing or nearly put the buyer off.
• Support tickets and live chat: show what confuses people after they buy, and the exact words they use for it. Your top ten ticket topics are often your top ten missing page sections.
• Sales call recordings and notes: show the questions prospects ask first and the objections that stall a deal. Get consent to record, and check recording laws where you operate.
• Survey free-text answers: scale well, but shallow. Useful for confirming patterns you first saw in interviews.
• Site search terms, emails and forum threads: show the vocabulary people use before they know your name.

How to run a customer interview

Nielsen Norman Group (NN/g) has clear guidance on what interviews are good and bad at. In a 2010 article, Jakob Nielsen says they suit exploring how people think about a problem and gathering vivid detail from specific cases. He warns they are poor at predicting future behaviour, because people struggle to imagine things they have not tried, and that memory of what people "usually" do is idealised.

So ask about specific past events. "Tell me about the last time you looked for an accountant" beats "What do you look for in an accountant?" NN/g also warns about leading questions and the "query effect": people will invent an opinion about almost anything you ask them. It suggests giving more weight to what people raise unprompted.

A workable interview, of the kind a switch interview uses, has five steps.

• Recruit recent buyers: people who chose you, or chose a competitor, in roughly the last few months, while the memory is fresh.
• Start with the first thought: ask when they first realised they had a problem, and what was happening then.
• Walk the timeline: what they tried, what they searched for, who they asked, what nearly made them stop, and what finally triggered the decision.
• Probe vague words: when they say "better" or "easier", ask "better than what, and how would you have known?"
• Capture exact phrases in quotation marks. Do not paraphrase while taking notes.

Surveys: ask carefully, because wording changes answers

Surveys help you check at scale what interviews suggested. They are also easy to get wrong. Pew Research Center, which publishes its survey methods, documents how question format and wording shift results.

In a 2008 Pew poll, 58% of respondents chose "the economy" when it was offered as a closed-ended option, while only 35% volunteered it when the question was open-ended. Adding a clause to a 2003 question about military action changed support from 68% to 43%. Pew also recommends avoiding double-barreled questions, which ask about two things at once, and notes that agree/disagree formats invite people to agree.

For copy research, the practical rules are simple. Start with open-ended questions to discover language, then use closed options built from the most common answers. Keep questions neutral. Ask about what people did, not what they would do. Put one idea in each question.

• Good: "What was happening in your business when you started looking for help?" (open, specific, past)
• Good: "What almost stopped you from getting in touch?" (surfaces anxiety)
• Bad: "How much did you love our friendly service?" (leading)
• Bad: "Is price and quality important to you?" (double-barreled)

Turning raw language into headlines

Raw quotes are not copy. Copy comes from patterns across many quotes. Put every phrase in a spreadsheet, one row per quote, with columns for source, topic and which of the four forces it shows. Group similar quotes. Count how many different people raised each theme.

Then write from the strongest themes. Use the customer's nouns and verbs. If five people say "the paperwork", do not call it "documentation management". Keep their emotion but trim the rambling, and never invent a quote or attribute one to a real person without permission.

A simple template: pick the most common situation for the headline, the most wanted outcome for the subhead, the most common doubt for an objection block, and the most vivid phrase for a pull quote. Headline formulas shows how to shape these into lines that work on a page.

Check your reading of the language

Patterns from a dozen interviews are a hypothesis. Test them before you rewrite a whole site. Show two or three recent customers a draft headline and ask them to say in their own words what you offer. If they cannot, the words are not working.

When you have enough traffic, a split test can compare versions on a real behaviour. NN/g notes in a 2005 article that A/B testing shows which version performs better, but not why, so it works best as a final check on ideas already shaped by qualitative research. Our guide to headline formulas covers testing honestly. For the wider craft, see what is copywriting.