.svg)
Most website copy is written from the inside out. It describes what the product does in the words the product team uses, and visitors can tell. Voice of customer research, the discipline of capturing the voice of the customer in their exact words, fixes that, but only if you finish the job: my position is that VoC research is worthless until every quote you collect has a specific home on a specific section of your site.
Plenty of teams run interviews, fill a spreadsheet, and then write the homepage the same way they always did. The research sits in a folder while the hero still says "streamline your workflow". That is a last-mile problem, not a research problem.
This playbook covers the whole path in one place. You will choose collection channels, run interviews without leading the witness, and code the language into themes. Then you will score quotes so you can pick a winner, map each winner onto your hero, subhead, features, FAQ, and CTA, and test the result before launch.
I will use real language from buyers we have spoken with at Storylane to show what good raw material looks like, and I will say plainly where our own product fits and where it does not.
What voice of customer research actually is (and isn't)
Definition: Voice of customer (VoC) research is the structured collection of the exact words customers use to describe their problem, their decision, and your product, so that your copy can echo their language back to them.
The key word is exact. A product marketer's summary is not VoC, because it has already been translated into internal vocabulary. The value is in the specific, sometimes awkward phrasing a buyer uses when nobody has coached them.
Here is how I draw the line:
VoC is:
- Verbatim phrases from interviews, sales calls, support tickets, and reviews
- Language about the buyer's situation before they found you
- The trigger that made them start looking, in their words
- The objections that nearly stopped them from buying
VoC is not:
- An NPS score on its own, which tells you how people feel but not what they say
- A generic testimonial like "great product, great team", which carries no specific language
- A demographic persona such as "Meet Shelley, 47, mom of three cats"
The persona point deserves a next step, not just a critique. Age and hobbies do not predict why someone buys B2B software, so replace the persona with buying-motivation data: what was broken, what triggered the search, and what the buyer needed to believe before signing.
Motivation is what your copy has to answer, and motivation is what good customer research should capture.
Why voice of customer research for website messaging matters more now, not less
AI answer engines now summarise category pages before buyers ever click. If your copy sounds like every competitor, the summary has no reason to single you out. Distinctive, human-sourced language is one of the few things a model cannot flatten into the category average.
Buyers are already feeling this shift. One GTM engineer we spoke with put it simply:
"I believe that what's kind of new channel that's getting up is referrals directly from the LLMs."
- [GTM engineer, IT asset management software]
The other reason is sameness. When every vendor in a category makes the same claim, positioning collapses into a feature checklist. A B2B content strategist described exactly why that happens:
"It's really hard for me to build a differentiated story if you don't already have a perspective and a point of view."
- [Founder, B2B content strategy consulting]
VoC is where that point of view comes from when you do not have one yet. Your customers have already described your difference, just not in the words your team uses.
I will not give you a vendor-reported ROI figure here, because the ones circulating in this category are usually secondhand. The better justification is one you can measure yourself: conversion on the pages you rewrite, compared against the pages you did not.
Step 1: Choose your collection channels
No single channel tells the whole story. Interviews are deep but slow, reviews are broad but shallow, and sales calls skew toward deals in motion. The organising principle I use is triangulation: when a theme shows up in several independent channels, treat it as ground truth.
| Channel | What it surfaces | Best for |
|---|---|---|
| Customer interviews (30 to 45 min, recorded and transcribed) | Deep context, trigger events, emotional language | Hero headline and subheadline |
| Post-purchase or onboarding surveys (open-ended) | Short, repeatable phrasing at scale | Validating frequency of a theme |
| Sales call and support transcripts | Objections, confusion, questions asked out loud | FAQ and objection handling |
| Review sites and public forums (G2, Capterra, Reddit, Slack and Discord communities) | Unprompted praise and complaints, including about competitors | Differentiation and feature copy |
| Win-loss and renewal conversations | Why deals were won or lost, what kept customers | Proof points and CTA microcopy |
| Website chat transcripts (AI agent or live chat) | Questions and objections asked at the moment of hesitation, in the visitor's own words | FAQ, objection handling, and spotting new themes |
If you only have time for two channels, pick interviews and sales calls. Sales calls are the most underused source in B2B marketing because they already exist, and every quote in this article came from them.
A cybersecurity marketer told us their site pushes one heavy "book a demo" call to action and gives visitors no way to see the product first. That website messaging insight came from a sales call, not a research project.
Chat transcripts: voice of the customer at scale
Customer interviews give you depth from a handful of people. Website chat transcripts give you breadth: every visitor who types a question is telling you, unprompted, what they came for and what is stopping them. Nobody wrote the question for them, which removes the leading-question problem entirely.
We treat our own RepX conversation transcripts this way, reading them alongside sales calls for recurring questions, objections, and phrasing. If you already run chat on your site, you are sitting on a VoC channel that refreshes itself every day. The work is pulling it into the same coding spreadsheet as everything else, so chat language is scored against interview language instead of living in a separate tool.
Step 2: Run customer interviews and surveys without leading the witness
The fastest way to ruin an interview is a question that contains its own answer, like "Did you choose us because of our integrations?" Ask open questions, ask why, and follow the thread when the answer surprises you.
I use a five-part interview arc because it maps well onto how a website is read. Here is a question bank for each stage:
- Goals. "What were you trying to get done when you started looking?" and "What would success have looked like six months later?"
- Pains. "What were you doing before, and what was annoying about it?" and "What did that workaround cost you?"
- Trigger event. "What happened that made this a priority that month and not the month before?"
- Objections. "What almost stopped you from buying?" and "Who internally needed convincing, and what did they ask?"
- Wins. "What changed after you switched?" and "How would you describe the result to a peer?"
On sample size: you do not need 100 interviews to find your homepage headline. For a single B2B SaaS segment, a small set of well-run interviews usually starts repeating itself, and the practical rule is to stop when new interviews stop producing new themes. If themes are still shifting, your segment is probably too broad, so split it.
For surveys, one open prompt such as "What nearly stopped you from signing up?" beats a ten-question form nobody finishes. Even a single form field counts: a well-designed "how did you hear about us?" question collects buyer language about discovery on every submission.
Step 3: Code the language into themes
Coding sounds academic, but it is just sorting: every row is one verbatim phrase, and every column says what it is and where it came from. Here is the process I follow:
- Transcribe everything. Use any recording and transcription tool. The tool does not matter; having the full text does.
- Pull verbatims, not paraphrases. Copy the exact phrase into a row, including the grammar. Clean it up later, never at this stage.
- Tag the source and role. Record the channel and the speaker's role and segment so you can check frequency across sources.
- Assign one category per row. Use the five categories below.
- Merge near-duplicates into themes. Five people describing the same pain five ways become one theme with five candidate quotes.
Most frameworks stop at four categories: goals, pains, objections, and exact phrasing. I add a fifth that most SaaS teams need: the buying trigger, the moment that turned a background annoyance into an active search. Triggers are what your subheadline should name.
| Category | What it captures | Example from our calls |
|---|---|---|
| Goal | What the buyer wants to achieve | Tie website engagement to booked and qualified leads |
| Pain | What was broken before | Good paid traffic that still does not fill out the form |
| Objection | What nearly stopped them | A form feels like too much commitment on a first visit |
| Buying trigger | Why now | Referrals from LLMs appearing as a new inbound channel |
| Exact phrasing | Distinctive words worth reusing | "filter intent more accurately" |
Step 4: Score and prioritize which quotes to use
This is where most VoC projects stall. You have hundreds of coded quotes, many saying the same thing, and no rule for choosing, so someone picks the one they like.
I use a lightweight rubric. Score each candidate quote from 1 to 3 on three criteria, then add them up. Anything scoring 7 or more goes on the shortlist.
| Criterion | 1 point | 2 points | 3 points |
|---|---|---|---|
| Frequency across sources | Appears once, in one channel | Appears several times in one channel | Appears across two or more independent channels |
| Emotional specificity | Generic ("it's hard") | Names a situation | Names a situation and what it cost or felt like |
| Recognition test | A buyer would shrug | A buyer would nod | A buyer would say "that is exactly us" |
Worked example: three buyers, one pain
Here are three real quotes from our calls that describe the same pain: visitors who are interested but not ready for a sales conversation.
- Quote A: "We have decent traffic coming in but if you look at our model, it is not really somebody coming and filling out the form in the first go." [Product Marketing Manager, cloud database software]
- Quote B: "What I think we're missing is the opportunity for people to potentially get a demo without speaking to a salesperson." [Senior Digital Marketing Manager, cybersecurity and compliance software]
- Quote C: "So we've got lots of volume coming in." [President, HR tech]
The theme appears across three segments, so each scores 3 on frequency. On specificity, A and B score 3 for naming the moment and the missing option, while C scores 1. On recognition, B scores 3 because any marketer with a single demo CTA sees themselves in it, A scores 2, and C scores 1.
That gives B a 9, A an 8, and C a 5. B wins the hero or subhead slot, A becomes a pull-quote or FAQ framing, and C is cut.
The point is not precision: two people using this rubric will usually pick the same quote. Keep the shortlist where your sales team can find it, because the same coded quotes double as sales collateral.
Step 5: Map quotes onto every section of your website
A scored shortlist still is not copy. The last mile is deciding which kind of quote belongs where, because a hero and an FAQ answer do different jobs.
| Website section | What kind of quote belongs there | Example direction from our calls |
|---|---|---|
| Hero headline | The most frequent, most specific phrase describing the core outcome, not the feature | "See the product before you talk to sales" |
| Subheadline | A phrase naming the audience or the buying trigger | Marketers with good traffic and a form nobody fills out on the first visit |
| Feature and benefit sections | Outcome language paired with the feature that delivers it | "Filter intent more accurately" next to the qualification feature |
| Objection handling and FAQ | The most common "what almost stopped me" language, answered directly | "A form feels like too much commitment" answered with a lower-commitment path |
| Social proof pull-quotes | High-scoring quotes that did not make the hero | Quote A from the worked example |
| CTA microcopy | How customers describe taking the next step | "Explore the product" instead of "Book a demo" |
Spoken language does not paste straight into written copy, so keep the words and the idea, then edit for each section's length and rhythm.
Next, split your messaging into two layers. A cybersecurity marketing leader described it this way:
"We have this little nasty document here. Which is kind of our top level, top level messaging, but exactly to your point, we have this FAQ section which goes through all of these various questions."
- [Global Integrated Marketing Leader, cybersecurity]
The top layer carries the hero and features. The FAQ layer carries objections in exact wording you control, which matters wherever legal or compliance language is involved.
"Like we prefer to have a very specific wording. That's why I put it in an FAQ format."
- [CBO, fintech lending]
Also check that your messaging document was written for customers, not your internal team.
Before and after. Generic: "Streamline your demo workflow." VoC-sourced: "Let buyers see the product before they're ready to talk to sales." The second line is longer, but a buyer recognises their own problem in it. For the writing craft itself, see our guide to B2B website copywriting.
For feature sections, the goal is turning that language into a product story rather than a list. For CTAs, remember that the same customer language can shape your demo script, so the promise on the button matches the experience after the click. Keep the messaging consistent across every page in the journey, because a visitor who reads one promise on the homepage and a different one on pricing trusts neither.
Where RepX fits in this workflow
Full disclosure: this is us. Storylane RepX is an AI agent that sits on your website, answers visitor questions, and qualifies intent before a sales conversation. I am including it here because the problem this playbook keeps surfacing, visitors who are interested but not ready for a form, is exactly the gap it was built for.
The mechanism is simple. RepX is trained on the material you give it, such as your website, docs, decks, and call scripts, so the FAQ layer from Step 5 can become part of what it answers from, and customer-sourced objection language gets used in real conversations. Visitors who would never fill out a form can ask their question in their own words and get an answer, then choose a lower-commitment next step such as exploring the product.
The side effect matters just as much for VoC. Every question a visitor asks is new raw language, captured at the moment of hesitation, which turns your website into a continuous collection channel for the Step 3 spreadsheet. Ask RepX lets you run AI prompts across those conversations to see what visitors are talking about, a quick way to surface candidate themes before you code them properly.
Qualified leads and conversation summaries also flow to your CRM and Slack, so what a visitor asked about can shape the follow-up. One cybersecurity marketer described wanting exactly that: a more qualified contact that arrives with context sellers can run with.
Where it does not fit: RepX is not a substitute for the research itself. If you have not done the interviews and scoring, an agent will answer in the same inside-out language your site already uses, so do the positioning work first.
And if your buyers never visit your website before sales engages them, your effort is better spent on demo scripts and sales conversations than on on-site tools.
Step 6: Test the messaging before it goes live
VoC informs the draft, and testing validates it. Skip this step and you get copy sourced from customers that still misses, because the quote was right and the placement was wrong.
You do not need a large research budget. Dedicated message-testing platforms such as Wynter exist for this, and the intent is worth borrowing: get outside reactions to draft copy before you rebuild the site. Here is the lightweight version I recommend:
- Internal review first. Have a colleague who was not involved walk through the draft and send back a list of changes. Buyers we spoke with did exactly this before launch.
- Five-second test. Show the hero to people in your audience for five seconds and ask what the company does. If they cannot say, the headline failed.
- Five customer reactions. Send the draft to five customers and ask which line sounds most like them and which sounds least like them.
- Start on high-intent pages. Test on the homepage and pricing page together, then roll the winning language out across the site.
- Compare against a baseline. A/B test the rewritten section against the original, as one accounting software PMM planned to do with an embedded experience on their site.
The same discipline applies beyond the website. Product launch messaging built the same way tends to land better, and one customer service software team we spoke with wants to spin up a product tour for each new feature launch to support its launch messaging.
How often to refresh your VoC research
A sensible floor is a full refresh at least once a year, and immediately before any redesign or relaunch. But the calendar is not the real trigger. The real trigger is when your sales team starts hearing objections or phrasing that is not in your spreadsheet.
Markets move faster than annual research cycles. New competitors change what buyers compare you against, and new channels change what they know when they arrive.
A practical cadence looks like this. Run a full interview round once a year or before a relaunch, review sales call transcripts monthly for new themes, and rescore your shortlist every quarter. This also works for teams without much inbound traffic: VoC-led messaging improves sales decks, outbound emails, and demo scripts as much as it improves web pages.
Run it as a VoC program, an ongoing input rather than a one-off project, and it becomes part of how voice of customer research fits into a broader GTM strategy. The spreadsheet from Step 3 should be a living document owned by product marketing, not a file that gets archived after launch.
Conclusion
Voice of customer research for website messaging only pays off when it reaches the page. Collecting quotes is the easy part. The discipline is scoring them so you choose consistently, and mapping them so every section of the site has a job and a quote to do it, then testing before anything ships.
If you take one thing from this playbook, make it the mapping table in Step 5. Fill it with your top-scoring quotes and hold your current homepage against it, and the gaps will show you exactly what to rewrite first.
Then keep the loop running, because voice of customer research for website messaging is never finished. Sales calls, support tickets, and visitor questions will keep handing you new language, and the teams that win on messaging are the ones that keep scoring and placing it instead of filing it away. Start with one page this week, and let the results make the case for the rest.
FAQ
How is voice of customer research different from a buyer persona?
A buyer persona describes who the buyer is, often with demographic details. Voice of customer research captures what the buyer says about their problem, trigger, and decision, in their own words. For website messaging, motivation beats demographics, so use VoC to decide what your copy says.
How many customer interviews do I need before I can rewrite my website?
Fewer than most teams expect for a single segment. Stop when new interviews stop producing new themes. If themes keep shifting, narrow the segment before running more interviews.
Can I do voice of customer research without talking to customers directly?
Yes, partly. Sales call recordings, support tickets, review sites, and community forums all contain unprompted customer language. Direct interviews still give you the deepest context on triggers and objections, so use indirect sources to supplement them rather than replace them.
What's the difference between voice of customer and win-loss research?
Win-loss research focuses on why specific deals were won or lost. Voice of customer research is broader, capturing language about goals, pains, triggers, and objections from customers and prospects at every stage. Win-loss conversations are one useful channel inside a VoC program.
Where should voice of customer quotes go on a website?
Use the most frequent outcome phrase in the hero, trigger language in the subheadline, and outcome quotes in feature sections. Put objection language in the FAQ and runner-up quotes in social proof. CTA microcopy should mirror how customers describe their next step.
What is a VoC program?
A VoC program is voice of customer research run as an ongoing process rather than a one-off project: a standing set of channels (customer interviews, surveys, sales calls, chat transcripts, reviews), a shared coding spreadsheet, and a regular cadence for rescoring and placing quotes. Product marketing usually owns it.
Sources
- No external statistics are cited in this article. Customer quotes come from anonymized Storylane sales and customer calls.
Ready to let buyers see your product before they talk to sales? Book a RepX demo.