Message testing for B2B: make buyers explain your copy back to you
B2B message testing: write variants that make different claims, have buyers explain them back, run a monadic survey, and size any A/B test before you start.
Message testing means showing target buyers different ways of describing your product and finding out which one they understand and care about, before you spend money on a launch. In B2B, good message testing is usually small: a handful of conversations where real buyers read your copy and explain it back to you. That tells you more than a large survey of people who would never buy. Live A/B tests come later, and only if you have the traffic to read them.
Fix the positioning before you test the words
In April Dunford's definition, positioning is about what your product leads at delivering and which clearly defined customers care a lot about it, and it is not messaging, a tagline or a brand story (April Dunford). Dunford's method starts with competitive alternatives: what a customer would do if your product did not exist.
If every variant falls flat, the problem may be the positioning, and rewording will not fix that. Write each message against the alternative your buyers use today, which, as Dunford notes, can be a spreadsheet, a manual process or doing nothing at all. (Doing nothing is a tough competitor. It is free and needs no onboarding.)
A made-up example: three messages for an audit tool
Say a startup sells software that collects evidence for security audits such as SOC 2. Its buyers are heads of security and compliance managers at software companies with 100 to 1,000 employees, most of whom collect evidence by hand with spreadsheets and screenshots. The team writes three messages:
- A, time: "Get through your SOC 2 audit without weeks of screenshots."
- B, risk: "See which controls would fail today, before your auditor does."
- C, sales: "Answer a prospect's security questionnaire the day it arrives."
Make each variant a different claim
Each of those bets on a different main pain and, in practice, a different main buyer. Swapping synonyms tests very little; nobody's buying decision hinges on "streamline" versus "simplify." Keep the format identical so you test the idea and not the layout: a headline, a two-sentence explanation and one proof point each, at roughly the same length.
Method 1: buyer interviews (start here)
Talk to a few people of each buyer type, and add more until the reactions repeat. In each conversation:
- Ask about today first. How do they handle the problem now? You get their words and the alternative you are really competing with.
- Show one variant at a time. Ask them to say, in their own words, what the company does and who it is for. If they cannot, the message is unclear, however much they like it.
- Use a highlighter. In a content highlighter test, people mark text that is clear in one color and text that is confusing in another, then explain why (18F). For messaging, ask for convincing and hard-to-believe instead.
- Ask what they would need to see to believe it.
- Rotate the order between participants. SurveyMonkey cites studies showing that people tend to rate the first concept they see more favorably than later ones (SurveyMonkey).
Jakob Nielsen of Nielsen Norman Group points out that people bend their answers toward what they think you want to hear (NN/g). Stronger signals are an accurate restatement or a specific question about price or integration. "I like that one" is a weak signal. It is also what people say about a friend's new haircut.
Pay for participants' time, not for a favorable reaction; paying for participation without paying for positive feedback explains how to keep those separate.
Method 2: a monadic survey
For more than a handful of reactions, use a survey where each respondent sees one variant only. SurveyMonkey recommends this monadic design: it mirrors what buyers will see in the market, avoids order bias, and leaves room for more follow-up questions. The cost is a separate group of respondents for each variant.
Ask what the company does (open text, scored for accuracy) and how believable it is. The hard part in B2B is the sample: a panel of "IT decision makers" may include people who do not make the decision, so use research screener questions.
Method 3: live tests, only with enough traffic
A live A/B test shows variants to real prospects on ads, emails or pages. It is the most realistic test and needs a lot of traffic. Evan Miller's rule of thumb for sample size is n = 16σ²/δ², where σ² is p(1 − p) for a conversion rate p and δ is the smallest change you want to detect (Evan Miller). For a landing page that converts at 3%, detecting a lift to 3.6% works out to about 13,000 visitors for each variant by our calculation. Many B2B pages never reach that.
Miller also warns against checking results repeatedly and stopping as soon as a difference looks significant, which makes the reported significance meaningless. Decide the sample size in advance and wait for it, even on the day message B looks like a winner. If you run a live test at all, use it to confirm a message that already did well in conversations.
Read the results
| What you see | What it suggests |
|---|---|
| Buyers restate the message accurately and ask about price or setup | The message is clear and relevant |
| Buyers like it but describe it wrongly | Clarity problem: rewrite before judging appeal |
| The same objection comes up in several conversations | Add proof or change the claim |
| Different buyer types prefer different variants | A positioning decision about who the main buyer is |
| No variant lands with anyone | Revisit the positioning, starting with the competitive alternative |
If heads of security preferred message B and compliance managers message A, you have a decision about the primary buyer. It belongs in your buyer persona work as much as in your copy.
Your next step
Write three messages that each make a different claim, all in the same format. Then book five conversations with buyers who match your target, and ask each one to explain every message back to you.
If those buyers are outside your network, describe them, such as "heads of security at software companies with 100 to 1,000 employees," and Instant Expert finds people who match. Start with the directory pages for chief information security officers and marketing professionals in cybersecurity.
Your invitation note is a message too, so don't test your copy in it. Say what the call is for (feedback on three short product messages) and how long it runs, and save the messages for the call so each buyer reads them fresh. You set what each person receives, our fee is added on top (a $100 offer costs you $125), and you're charged only for calls that get booked. A security head who declines costs nothing, and one who wants more can reply with their price.