Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Conversion Rate Optimisation › Conversion Research

Turning Customer Language into Page Copy

How to turn customer language into page copy: gather words from reviews, support, sales calls and polls, code objections and phrases, then rewrite and test.
Tutorialintermediate10 min readUpdated

Why customers' words beat the brief's words

A brief is written by people who know the product. It names features, uses internal terms and answers the questions the team thinks are important. Visitors arrive with a situation, a worry and a vocabulary of their own, and a page written in the brief's words makes them translate. The anatomy of a landing page guide makes the point for objections: they are discovered, not invented, from sales calls, support tickets, chat transcripts, review sites and on-page surveys, and the page answers them where they arise, in the visitor's own language.

Customers' words also carry the job. The Harvard Business Review article on jobs to be done, listed in this page's sources, argues, as the jobs-to-be-done framework page sets it out, that customers hire products to make progress in a particular circumstance, and that the job has emotional and social dimensions as well as a functional one. A review that says "I just wanted the hallway to stop looking like a building site before my parents visited" tells a page more about what to lead with than a feature list.

The conversion research methods guide lists support tickets, sales notes and chat transcripts as research that already exists and reads them for objections; this tutorial reads the same material for wording as well. The rest depends on one discipline: keep the exact words. A paraphrase written by the team turns the customer's language back into the brief's. The table shows what to code and where each kind of code usually ends up on a page.

CodeWhat it capturesWhere it usually goesFramework or guide
JobThe progress the customer is trying to make, in their situationHeadline and subheadJobs-to-be-done
OutcomeWhat success looks like to themBenefitsFeatures, advantages, benefits
ObjectionPrice, time, effort, risk, trust or fit, as the anatomy guide lists themBeside the point where the doubt arisesAnatomy of a landing page
AlternativeWhat else they considered or used beforeSubhead and how the offer differsValue proposition canvas
Proof cueWhat convinced them, or would haveProof placed beside the claimTrust and proof
Exact phraseThe words themselves, with their sourceHeadline, call to action and objection answersMessage match

Before you start

  • The anatomy guide, read. The prerequisite is the anatomy of a landing page guide; the codes are mapped to its sections.
  • One page and its decision. A product page, landing page or service page with a known conversion and a sense of where visitors hesitate, ideally from a funnel analysis or a research sprint.
  • Access to the sources. Review exports, the support and chat systems, sales-call notes, poll answers and the site search report, for a recent period.
  • A redaction routine. A way to remove names, contact details and order numbers before material is pasted anywhere, agreed with whoever handles data protection.
  • A coding sheet. One row per excerpt, with columns for the source, the date, the code and the exact words.

The steps

  1. Choose the page and the decision it supports

    Pick one page and write the decision a visitor makes on it: "Order this mattress for delivery to my flat", "Book a call about payroll". Note which section the research or the funnel suggests is failing, if anything does. A page and a decision keep the gathering focused; collecting language about everything the business sells produces a pile of quotes no page can use.

  2. Gather words from every source you already hold

    Collect a recent period of the business's own reviews and competitors' public reviews for the same kind of product, support tickets and chat transcripts, sales-call notes, returns and cancellation reasons, answers to on-page polls and post-purchase questions from the polls tutorial, and site search terms from the site search tutorial, which record intent in the visitor's own words. Keep each source separate and labelled, because the ranking step counts sources. The voice of the customer guide treats open text from surveys, support contacts and reviews as part of the same listening programme; this step borrows its material. Where the support team already tags contacts with a reason list, as the tutorial on tagging support contacts by reason sets out, pull the tickets under the reasons that concern this page rather than reading the whole queue.

  3. Remove personal data before analysis

    Support tickets, chat transcripts and returns emails are full of names, addresses, phone numbers and order references, and site search terms sometimes contain email addresses. Remove them before the material is shared, pasted into a spreadsheet or sent to any tool. The AI research synthesis guide sets out the redaction step: replace names with codes, remove contact details, account and order numbers, keep the key separately and read a sample of the redacted files by eye. The personal data guide explains why copies of personal data spread and how to avoid making them.

  4. Code jobs, outcomes, objections and alternatives

    Read a sample of each source and draft a short codebook using the table above: the jobs customers describe, the outcomes they want, their objections grouped as price, time, effort, risk, trust and fit, the alternatives they mention and what convinced them. Give each code a one-line definition and an example. Then apply it to all the material, allowing several codes per excerpt. Where a language model applies the codebook, use the synthesis guide's rules: codes only from the frame, excerpts copied exactly and a person checking a sample blind. Where the material includes a survey's open answers, the tutorial on coding open-text answers by hand shows how to write each code's exclusions and count respondents against the base.

  5. Keep exact phrases with their source

    For every code, keep the strongest phrases exactly as written, with the source, the date and the excerpt's reference, in a separate column. Do not correct grammar, merge two customers into one sentence or shorten a long answer into something more quotable; the synthesis guide warns that a tidied phrase reads naturally and is no longer evidence. Mark phrases from competitors' reviews as such: they describe the category's language, not your customers. The academy checkpoint on research sprints, usability sessions and customer words tests whether you can code customer language and keep each exact phrase with its source.

  6. Rank codes by how many sources share them

    Count, for each code, how many independent sources it appears in: reviews, support, chat, sales notes, polls, search. A worry that turns up in reviews, chat and sales notes is a stronger lead than one vivid review, however quotable. Within a source, count distinct customers rather than mentions, and take counts from the coded sheet, never from a model's summary. These counts describe the material, not all customers; a code that tops the list is the one to write for first, not a measure of how common the worry is.

  7. Map codes to page sections

    Place each high-ranking code in the section that answers it. The job and the strongest phrase go to the headline and subhead, where message match with the advert or search that brought the visitor still has to hold. Outcomes become benefits, built with the features, advantages and benefits chain so each benefit rests on a real capability. Objections go beside the point where the doubt arises, not in a FAQ at the bottom. Alternatives shape the subhead and the comparison, and the value proposition canvas checks that every pain the customers named has a reliever on the page. If the business keeps a messaging house, check the mapping against its pillars; the messaging hierarchy guide builds the message from strategy, and customer language supplies its wording and proof.

  8. Rewrite the page in the customers' words

    Rewrite each section using the phrases, adapted only as far as the sentence needs. Write for scanning: the writing for the web guide states, from Nielsen Norman Group's research on how users read on the web listed in this page's sources, that most users scan a page rather than read it word by word, so put the customer's words in headings, the first words of paragraphs and the call to action (CTA). Keep the Landing Page sections in the order the anatomy guide gives them. Where a phrase is a worry, answer it in the same words: if customers say "will it fit up my stairs", the page says "Will it fit up your stairs?", not "Delivery logistics".

  9. Check every claim can be substantiated

    Customers' words make promises sound natural, which is why every claim in the rewrite needs checking. The trust and proof guide sets the honesty rules: use only what is true, attribute what can be attributed, keep records of permissions, show reviews as they came and make sure every claim could be substantiated if someone asked. Quote a named customer only with permission. Reviews used as Social Proof appear as written, never edited for meaning. A phrase from a competitor's review is never presented as your customer's.

  10. Test the rewrite or ship and monitor it

    A rewrite is a change whose effect is uncertain. Where the page has the traffic for a test, run the rewrite against the original with the conversion the page exists for as the primary metric, using the landing page headline test design when the change is led by the headline and the A/B testing guide for the rest. Test the rewrite as a whole when it is bold, rather than one phrase at a time. Where traffic is too thin, ship it, record the date and watch the page's conversion and the questions customers still ask in support and polls, read cautiously because other things change too. Either way, record which codes the rewrite used in the research backlog of the conversion research hub, so the next rewrite starts from evidence.

A mattress retailer rewrites its delivery section (illustrative)

Illustrative example — not a real result

Dozefinch Mattresses rewrites in its buyers' words

Dozefinch Mattresses is an invented online mattress retailer. Its best-selling mattress page leads with "Seven-zone pocket springs for perfect support" and gives delivery one line near the footer. The marketer gathers the last six months of the shop's own reviews, competitors' public reviews of similar mattresses, chat transcripts and returns emails, and redacts names, addresses and order numbers before coding. The codebook covers jobs, outcomes, objections, alternatives and proof cues. Firmness appears, but the code that appears in every one of the shop's own sources is a fear about getting the mattress into the home: "will it go up a narrow staircase", "our stairs turn halfway", "I measured the landing three times". Chat transcripts show visitors asking whether the mattress arrives rolled, and some returns emails give the reason as "could not get it upstairs". The marketer rewrites the subhead as "Arrives rolled in a box that fits up narrow stairs, then opens to full size in a day" after checking with the warehouse that both claims are true, and moves the delivery section up beside the price, headed "Will it fit up your stairs?", with the box's dimensions and a line on what happens if it does not. The page receives around 20,000 visits a month, so the rewrite runs as a test against the original with orders as the primary metric and returns as a guardrail, and the codes it used are recorded beside the test.

Common mistakes

  • Collecting everything. Language for a page comes from one page and one decision.
  • Paraphrasing the quotes. A tidied phrase is the brief's language again, and no longer evidence.
  • Ranking by vividness. One memorable review is a lead; a worry shared across sources is a priority.
  • Skipping redaction. Tickets and chats are full of names and addresses that analysis never needs.
  • Treating counts as prevalence. Mentions in the material are breadth within the sources, not shares of customers.
  • Leaving objections in the FAQ. Answer a worry where it arises, in the words customers used.
  • Claims nobody can stand behind. Customers' phrasing makes promises easy to write and hard to keep.
  • Borrowing competitors' customers. Their reviews show the category's language, never quotes for your page.
  • Expecting a guaranteed uplift. A rewrite is a hypothesis; test it where traffic allows or ship and watch.

Frequently asked questions

Where can we find customer language without a research budget?

In material the business already holds. Its own reviews and competitors' public reviews, support tickets, chat transcripts, sales-call notes, returns and cancellation reasons, answers to on-page polls and post-purchase questions, and the terms people type into the site's search box all record what customers want and worry about in their own words. The work is reading and coding them, not commissioning anything new. Interviews add depth where the existing material is thin, and the customer interviews tutorial covers running them.

Can we quote customers' words on the page?

Use the words freely in your own copy, because a phrase that describes a common worry is not a quotation. Quoting a named customer is different: as general practice and as the trust and proof guide sets out, get permission to use the name and the words, keep a record of it, edit for length but never for meaning, and never write a testimonial for a customer to approve. Competitors' reviews are for understanding the category and its language, never for quotes on your own page. Nothing here is legal advice.

Does a rewrite in customers' words always need an A/B test?

Test it where the page has the traffic to support a test, because a rewrite is a change whose effect is uncertain, and judge it on the conversion the page exists for, not on clicks alone. The landing page headline test design sets out the hypothesis, metric, guardrails and duration for the most common case. Where traffic is too thin to test, ship the rewrite, record the date and watch the page's conversion rate and the questions customers still ask, read cautiously because other things change at the same time. No rewrite comes with a promised uplift.

Terms used on this page

Frameworks

Where this page is used

Part of these learning paths

Also related

Sources

  1. How Users Read on the Web — Nielsen Norman Group
  2. Know Your Customers' Jobs to Be Done — Harvard Business Review

Related