Copywriting That Converts · Editing, brand voice, AI-assisted copy and testing · lesson 19 of 19 · 8 min
Testing copy and the rewrite clinic
Opinions are free; evidence is valuable
Every team has opinions about copy. The founder prefers the clever headline, the designer prefers the short one, the sales lead wants more detail. Testing replaces opinion with evidence — but only when you test the right things, the right way, with enough people. This lesson gives you a toolkit for testing copy at every budget, then walks through a rewrite clinic: three real-world style pieces rewritten with the frameworks from this course, annotated so you can see each decision.
What is worth testing
Test big ideas before small words. Changes that alter what the reader understands or believes move results; cosmetic changes rarely do.
| Test this first (high impact) | Test later (usually small impact) | |---|---| | Value proposition and lead benefit | Synonyms in a headline | | Angle: problem vs outcome vs objection | Button colour | | Offer and risk reversal | Exclamation marks | | Proof type and placement | Minor word order | | Page length and structure for this audience | Emoji or no emoji |
Five ways to test copy (by traffic level)
| Method | When to use | What it tells you | Tools (examples) | |---|---|---|---| | Five-second / first-impression test | Before launch, any traffic | Is the value proposition clear? | Lyssna, Maze | | Preference and message tests | Before launch; B2B or niche audiences | Which message resonates, and why | Lyssna, Wynter, survey tools | | Customer interviews with drafts | Low traffic, high-value products | Objections and language gaps | Video calls + your notes | | Ad-platform experiments | Any business spending on ads | Which angle wins on cost per result | Google Ads experiments, Meta A/B tests | | On-site A/B tests | Enough traffic and conversions | Causal impact on conversion | VWO, Optimizely, AB Tasty, Convert, GrowthBook, PostHog |
Ad experiments are often the fastest way for small businesses to learn which angle works; then carry the winning angle to the landing page. On-site A/B testing needs enough conversions to detect realistic differences — the Conversion Rate Optimization course covers sample size in depth. As a rough orientation (illustrative): if a page converts at 3% and you hope to detect a relative lift of 10%, you typically need tens of thousands of visitors per variant. Low-traffic sites should test bigger changes, test upstream in ads, or rely on qualitative methods.
A copy test plan template
Test name: Hero angle — outcome vs objection (pricing page, UK traffic)
Insight: Exit survey: 41 of 120 responses (illustrative) mention "hidden costs" before signing up
Hypothesis: Because visitors fear hidden costs, leading with "One price. No setup fees." instead of
"Run payroll in minutes" will increase trial starts among first-time visitors.
Primary metric: Trial starts per visitor (new visitors)
Guardrail metrics: Refund requests in first 30 days; support tickets about pricing
Audience / split: 50/50, new visitors, all devices
Minimum run: 2 full weeks and the pre-calculated sample size, whichever is later
Decision rule: Ship B if primary metric improves at the planned confidence and guardrails do not worsen
Learnings log: [link]
Rewrite clinic 1: SaaS pricing page hero
BEFORE
H1: "Powerful payroll. Simplified."
Sub: "The all-in-one platform trusted by businesses everywhere."
CTA: "Get started"
DIAGNOSIS
- Five-question test: what it is (payroll) is barely clear; for whom, outcome, proof and next step missing
- "Trusted by businesses everywhere" = unprovable, generic
- Research: small UAE and KSA employers fear WPS compliance errors and hidden fees
AFTER
H1: "Payroll for small teams in the UAE and KSA — WPS-ready, one monthly price"
Sub: "Run salaries, payslips and end-of-service calculations in minutes. No setup fees."
Proof: rating from a named review platform (only if current) + one customer quote with permission
CTA: "Start my free 14-day trial" Microcopy: "No card needed. Import staff from Excel."
Annotations: the audience and market are named; the main fear (compliance) is answered in the H1; the objection (hidden costs) is handled in the sub; the CTA is first-person with anxiety-reducing microcopy. "WPS-ready" must be accurate for the product.
Rewrite clinic 2: promotional email
BEFORE
Subject: "MEGA SALE!!! Don't miss out 🔥"
Body: "Huge discounts on everything. Shop now before it's too late!!!"
AFTER
Subject: "30% off winter jackets until Sunday midnight"
Preview: "Sizes XS-3XL still in stock. Free exchanges."
Body: "Cold snap coming? Our waterproof jackets are 30% off until Sunday at midnight (UK time).
Most sizes are still in stock, and exchanges are free if the fit isn't right.
[Shop winter jackets]"
Annotations: specific offer and genuine deadline; the preview answers the top objection (size availability); the discount must be calculated from a genuine previous price.
Rewrite clinic 3: product description
BEFORE
"Premium quality water bottle. Best in class. Perfect for everyone."
AFTER
"Keeps water cold for 24 hours in a hot car (our test: 45°C, 24h, ice remained).
750 ml — fits standard car cup holders and most backpack side pockets.
Leak-proof lid tested upside-down in a bag for 48 hours.
Dishwasher-safe lid; hand-wash the bottle."
Annotations: every claim is specific and tied to a test the brand actually ran; the fit detail answers a common objection; care instructions prevent returns.
Common mistakes
- Testing button colours while the value proposition is unclear.
- Declaring a winner after two days or a handful of conversions.
- Testing on a page with too little traffic instead of using ads or qualitative tests.
- Measuring clicks when the business cares about qualified customers.
- Not logging results, so the same debate happens again next quarter.
Video lecture: Testing copy and the rewrite clinic
Lecture coming soon · 12 chapters · about 9 minutes. Read the full transcript below.
- Testing copy and the rewrite clinic
- Why test copy
- What to test first
- Five ways to test copy
- The traffic reality
- A copy test plan
- Simple example: Lahore skincare (illustrative)
- Rewrite clinic 1: payroll pricing page
- Rewrite clinic 2: promotional email
- Rewrite clinic 3: product description
- Common mistakes
- Recap
Lecture transcript
Testing copy and the rewrite clinic
Every team meeting about copy eventually hits the same wall. I like this one. Well, I prefer that one. And whoever's most senior usually wins. There's a better way. In this lecture you'll learn what's actually worth testing, five ways to test copy at any traffic level, how to write a copy test plan, and why most small sites shouldn't start with on-site A/B tests at all. Then we'll run a rewrite clinic together: three real-world style pieces, a pricing page, an email and a product description, rewritten with the frameworks from this course, one decision at a time.
Why test copy
Why does this matter? Because copy is full of plausible choices. The clever headline and the clear headline both sound good in a meeting. Only readers can tell you which one works. Testing protects you from two expensive mistakes: shipping a change that quietly hurts results, and throwing away a good idea because the loudest person didn't like it. It also builds a learning library. Over time, you stop guessing about your audience and start knowing.
What to test first
First, what to test. Test big ideas before small words. The changes that move results alter what the reader understands or believes: the value proposition, the lead benefit, the angle, the offer, the proof, the page structure. Cosmetic changes like synonyms, button colours and exclamation marks rarely move much on their own. Think of it like cooking. Changing the main ingredient changes the dish. Changing the garnish rarely does. So if your value proposition is unclear, don't test button colours. Fix the main ingredient first.
Five ways to test copy
Now, five ways to test, matched to traffic. One, five-second or first-impression tests: show the first screen briefly, then ask what it is and who it's for. Great before launch, at any traffic level. Two, preference and message tests with panels of your target audience, which work well for niche and B2B products. Three, customer interviews using drafts. Four, ad-platform experiments, like Google Ads experiments or Meta A/B tests, which are often the fastest way for small businesses to learn which angle wins. And five, on-site A/B tests with tools like VWO, Optimizely, AB Tasty, Convert, GrowthBook or PostHog, which need enough conversions to detect realistic differences.
The traffic reality
Let's be honest about traffic. On-site A/B tests need a lot of people. As an illustrative rule of thumb, if a page converts at three percent and you want to detect a ten percent relative improvement, you typically need tens of thousands of visitors per version. Many small sites simply don't have that. So what do they do? Test bigger, bolder changes that could produce larger effects. Test angles upstream in ads, where you can buy the sample. And use qualitative methods to make better decisions even without statistical proof. The Conversion Rate Optimization course goes deep on sample size and power.
A copy test plan
Every test deserves a written plan. It includes the insight: what research told you. The hypothesis: because of this insight, changing this for this audience will improve this metric. The primary metric, like trial starts per new visitor. Guardrail metrics, like refund requests or support tickets, so a win doesn't hide harm. The audience and split. A minimum run time, at least full weeks to cover weekday and weekend behaviour, and the pre-calculated sample size. A decision rule written before you see results. And a link to the learnings log. Writing it first stops you moving the goalposts later.
Simple example: Lahore skincare (illustrative)
A simple example. A small skincare brand in Lahore can't run on-site A/B tests; there aren't enough orders. But they spend steadily on Meta ads. So they test two angles as separate ads with the same image and budget: an outcome angle, sunscreen that doesn't leave a white cast, and an objection angle, no stinging, even on sensitive skin. After enough spend, one angle wins clearly on cost per purchase. They move that angle into the product page headline. No complex statistics on-site, but a real signal from the market, carried to where it matters.
Rewrite clinic 1: payroll pricing page
Now let's start the rewrite clinic. Piece one is a payroll software pricing page for small employers in the UAE and Saudi Arabia. The headline says, powerful payroll, simplified. The sub says, the all-in-one platform trusted by businesses everywhere. The button says, get started. Run the five questions. What is it? Payroll, barely. For whom? Missing. Outcome? Missing. Proof? Trusted by businesses everywhere is unprovable. Next step? Vague. Research says small employers fear compliance errors with the wage protection system, and hidden fees. So the new headline reads: payroll for small teams in the UAE and KSA, WPS-ready, one monthly price. The sub: run salaries, payslips and end-of-service calculations in minutes, no setup fees. The button: start my free fourteen-day trial, with no card needed underneath. And WPS-ready must be accurate.
Rewrite clinic 2: promotional email
Piece two, a promotional email for a UK outdoor clothing brand. Subject: mega sale, don't miss out, with three exclamation marks and a fire emoji. Body: huge discounts on everything, shop now before it's too late. Let's apply what we know. The subject should state the specific offer and the genuine deadline: thirty percent off winter jackets until Sunday midnight. The preview answers the top objection, which from support tickets is availability: sizes XS to three XL still in stock, free exchanges. The body: cold snap coming? Our waterproof jackets are thirty percent off until Sunday at midnight, UK time. Most sizes are in stock, and exchanges are free if the fit isn't right. One button. And the thirty percent must come from a genuine previous price.
Rewrite clinic 3: product description
Piece three, a product description for an insulated water bottle. Before: premium quality water bottle, best in class, perfect for everyone. Three claims, zero information, and best in class is an unprovable superlative. After: keeps water cold for twenty-four hours in a hot car, then the brand's own test conditions in brackets. Seven hundred and fifty millilitres, fits standard car cup holders and most backpack side pockets. Leak-proof lid, tested upside down in a bag for forty-eight hours. Dishwasher-safe lid, hand-wash the bottle. Every line is specific. Every claim traces to a test the brand actually ran. And the fit and care details answer the questions that usually cause returns.
Common mistakes
Common mistakes. Testing button colours while the value proposition is unclear. Declaring a winner after two days, or after a handful of conversions. Running on-site tests on pages with too little traffic, instead of using ads or qualitative methods. Measuring clicks when the business cares about qualified customers. Forgetting guardrail metrics, so a winning headline quietly increases refunds. And not logging results, so the same debate returns next quarter. Your learnings log is how testing compounds.
Recap
Recap. Test big ideas before small words. Choose the method that fits your traffic: five-second tests, message tests, interviews, ad experiments, and on-site A/B tests when you have the volume. Write a plan before you start, with a hypothesis, primary and guardrail metrics, a minimum run and a decision rule. And use the rewrite clinic pattern: diagnose with the five questions and your research, rewrite with specifics, and verify every claim. Try this now: pick one piece of copy, diagnose it with the five questions, rewrite it, and write a one-page test plan using the template in the lesson text.
Key takeaways
- Test big ideas — value proposition, angle, offer, proof — before small wording changes.
- Match the testing method to your traffic: five-second and message tests, interviews, ad experiments, then on-site A/B tests.
- Write the test plan first: insight, hypothesis, primary and guardrail metrics, minimum run and decision rule.
- Rewrite by diagnosing with the five reader questions and research, then replacing vague claims with specific, verified ones.
Try it
Choose one live piece of copy, diagnose it with the five reader questions, rewrite it using specific verified claims, and write a one-page test plan with a primary metric, guardrails and a decision rule.