You have six ways to describe the same product, one launch budget and no proof of which line will make people buy. Message testing settles that with evidence from your target buyers before you commit to creative and media. It sits inside the wider practice of concept testing, but it isolates the words: what you promise, and whether people believe you. This guide covers the measures, the survey designs, claims testing and a worked example.
Key takeaways
- Message testing compares plain statements of what you promise, so you learn which idea persuades.
- Read motivation to buy and believability side by side. A message people want but do not believe will not work in market.
- Use monadic cells for a few messages, MaxDiff for long lists and TURF to choose a message set.
- Claims testing is message testing for specific product claims you can substantiate.
- Cut results by segment: the best message overall may not be the best for your core buyer.
What is message testing, and how is it different from concept and ad testing?
All three show stimulus to target buyers. They differ in what the stimulus contains, and so in the decision they support.
| Method | What respondents see | Decision it supports |
|---|---|---|
| Concept testing | The whole product idea: what it is, its benefits, often its price | Whether to develop or launch the product |
| Message testing | Short, plain-text statements of a single promise or reason to buy | What to say, and which promise to lead with |
| Ad or creative testing | Finished or near-finished executions: video, print, display | Whether this execution works, and how to fix it |
Message testing sits between the two. Once the product is set, it tells you which promise to lead with, and creative testing then checks whether an execution delivers it. Plain text keeps design, casting and music from deciding the result.
What should a message testing survey measure?
Rate each message on five measures, usually on a 5-point scale, and report the top-2-box share (the percentage who answer 4 or 5).
- Appeal: how much the respondent likes what the message says.
- Believability: whether they think the brand can deliver it.
- Relevance: whether it matters to them and their needs.
- Differentiation: whether it says something other brands do not.
- Motivation to buy: whether it makes them more likely to buy.
Motivation is usually the headline measure because it sits closest to the decision. Our concept testing questions give wording you can adapt.
Why does believability matter so much?
A message can score well on appeal simply because it promises a lot: “The last cleaner you’ll ever need” sounds great, and few people accept it. A promise that is not believed does not move purchase, and it can make your other claims look less credible. In launch work, we see teams pick the most motivating line and only later learn that buyers shrugged it off. Read motivation and believability together, never motivation alone.
What is claims testing?
Claims testing is message testing applied to specific, checkable product claims, such as “dries in 10 minutes” or “dermatologist tested”. The method is the same.
The caution is substantiation. Advertising rules in most markets expect you to hold evidence for an objective claim before you make it, so test only claims your product and legal teams confirm you can support. A survey shows which claims persuade. It does not prove a claim is true, and high believability is not evidence of product performance.
Which survey design should you use for message testing?
The right design depends mostly on how many messages you have.
- Monadic, for a handful of messages. Each respondent sees one message, so none is judged against another, but you need a sample cell per message. Our comparison of monadic and sequential monadic designs covers the cheaper option of rotating three or four messages per person.
- MaxDiff, for long lists. With 10 to 30 candidate messages, ratings blur together. MaxDiff analysis asks people to pick the most and least motivating message from small sets, which forces trade-offs. Use it to shortlist, then rate finalists for believability.
- TURF, to choose a message set. TURF (total unduplicated reach and frequency) finds the combination of messages that reaches the largest share of people, where “reached” means at least one message in the set motivates them. It needs each respondent to rate several messages, so it pairs with sequential monadic or MaxDiff data, not pure monadic cells.
How do you read message testing results?
Here is a worked example. A household brand launching a refillable surface cleaner showed each of six messages monadically to 200 category buyers, 1,200 respondents in total. The table shows top-2-box shares for motivation to buy and believability.
| Message | Motivating (top-2-box) | Believable (top-2-box) | Quadrant |
|---|---|---|---|
| A. Cleans as well as leading sprays, with plant-based ingredients | 58% | 44% | Fix the proof |
| B. One refill bottle replaces six plastic bottles a year | 46% | 71% | Support |
| C. Safe to use around kids and pets | 63% | 68% | Lead |
| D. Made in our own factory | 31% | 48% | Drop |
| E. The last cleaner you’ll ever need | 61% | 29% | Fix the proof |
| F. Saves you money on every refill | 52% | 60% | Lead |
Ranked on motivation alone, E comes second. Believability changes the picture. To build the quadrant, use the six-message average as the cut-off on each measure.
- Average motivation: (58 + 46 + 63 + 31 + 61 + 52) / 6 = 311 / 6 = 51.8%
- Average believability: (44 + 71 + 68 + 48 + 29 + 60) / 6 = 320 / 6 = 53.3%
That gives four groups.
- Lead (motivating and believed): C and F, your primary messages.
- Fix the proof (motivating, not believed): A and E. A could earn belief with evidence, such as a named performance test. E is an overpromise and hard to rescue.
- Support (believed, less motivating): B, a reason to believe under a lead message.
- Drop: D, which is neither.
Check the gaps before you call a winner. With 200 respondents per cell, the margin of error on a 50% score is about plus or minus 7 points (1.96 × √(0.5 × 0.5 / 200) = 0.069). C’s 11-point lead over F on motivation is significant at 95%, but its 2-point lead over E and 5-point lead over A are not. So C and F lead, with C the stronger.
How do you test messages by segment?
The best message overall may not be the best one for the buyers you most need, so cut each message’s scores by key segments, such as age or current brand, in a crosstab. In the example, message B is motivating to 60% of under-35s (48 of 80) and 37% of those 35 and over (44 of 120), which together give the 46% in the table (92 of 200). The 23-point gap is significant, so B could lead in media aimed at younger buyers.
Watch the base sizes: segments of 80 respondents carry wide margins of error, so plan the sample around your key segments. Crosstabs with significance testing and weighting show which segment gaps are real. A message for the top of the purchase funnel also has a different job from one meant to close the sale.
What are the common mistakes in message testing?
- Testing slogans instead of messages. A tagline is an execution. Test the plain promise behind it.
- Showing too many messages per person. After four or five ratings, people tire and scores drift together. Add cells or switch to MaxDiff.
- Ignoring believability. Leading with a motivating message that nobody believes wastes media and costs trust.
- Writing messages unevenly. If one message is twice as long as the rest, you are testing the writing, not the idea.
- Calling winners on small gaps. A 3-point lead on 200 per cell is noise. Test it before you brief the agency.
What should you do next?
Write your candidates as plain, single-promise statements of similar length, and decide which segments you must read before you set the sample. Field motivation and believability, build the quadrant and test the gaps. Then keep the results where the next team can find them. When you connect findings across studies and waves, the next launch starts from what this one learned, not from a blank page.