Concept Testing in B2B: Methods, Questions and a Real Example
How B2B teams test positioning angles, names, value propositions and campaign concepts with real buyers before committing to them.
Updated November 4, 2026
There's a line attributed to Jim Barksdale, the former Netscape CEO: "If we have data, let's look at the data. If all we have are opinions, let's go with mine."
That's how most B2B concepts get picked. A new homepage direction, a product name, a campaign idea, a positioning angle. The leadership team looks at 3 options, debates for an hour, and the highest-paid person's opinion wins.
Nothing wrong with leaders making calls. They own the results and often know the industry best. But you are not your target customer. And unless you put the options in front of the people they're meant to sway, you stay stuck in my-opinion-vs-your-opinion land.
Concept testing gets you out of it. Here's how to do it in B2B, where the guides written for consumer products don't really apply.
Concept testing is showing early versions of an idea (a positioning angle, a product name, a value proposition, a campaign or a product concept) to people in your target market before you commit to it, and measuring which one they understand, believe and prefer, and why.
It's been a formal research method since at least the 1970s. William L. Moore's review in the Journal of Business Research (1982) is still a good overview of the academic side. Most of the literature is about consumer packaged goods. B2B is different in 3 ways: your audience is small and hard to reach, purchases involve many people, and you're usually testing words and ideas, not physical products.
| What you're deciding | Example concepts to test |
|---|---|
| Positioning angle | "Revenue intelligence" vs "conversation analytics" vs "sales coaching" |
| Category or use case | Is this a new category, or a use case of an existing one? |
| Product or feature name | 3 candidate names for a new module |
| Value proposition | Lead with speed vs lead with cost vs lead with accuracy |
| Homepage direction | 2 or 3 above-the-fold concepts with different headlines |
| Campaign concept | Contrarian vs data-led vs customer-story creative |
| Content format | Narrative customer story vs classic case study |
| Packaging | Per-seat vs usage-based plan structure |
The category question is a big one. I was on a Zoom call with the leadership team of a software composition analysis company, CEO, CPO and CMO all there. They were debating an all-in positioning shift to an emerging category. Half the room was skeptical. The cost of being wrong was months, maybe years, of wasted time.
They ran a message test, a preference test and a survey with their ICP on Wynter. The data showed clearly that what they had was a use case, not a category. A category-level headline would have resonated with a small fraction of their ICP and confused the rest. The whole thing took 2 weeks, including writing the tests. Once they saw the results, there was no internal arguing. They changed course and moved on.
Pricing and packaging are the most under-tested concepts of all. When I asked 50 B2B SaaS VPs and CEOs how they test pricing before rolling it out, only 18% did anything rigorous. 34% "tested" by asking a few friendly customers, and 28% never tested at all. One VP said they'd always been skeptical it could be done, "given that asking customers what they want to pay doesn't work." They're right about asking directly. That's why you test packaging concepts side by side instead of asking for a number.
| Method | How it works | Answers | Sample size | Watch out for |
|---|---|---|---|---|
| Monadic | Each person sees one concept and rates it | How good is this concept on its own? | 30 to 50 per concept | Needs more people, since you split the sample |
| Sequential monadic | Each person sees all concepts, one at a time, in random order | How do concepts score on the same scales? | 30 to 50 total | Order effects; always randomize |
| Comparative (preference test) | Each person sees concepts side by side and picks one | Which do they prefer, and why? | 30 to 50 | A vote without reasons tells you little |
| Qualitative read-through | People react to one concept in detail, line by line | What's unclear, unbelievable or missing? | 15 to 30 | Not a vote; don't count it like one |
Most B2B teams should combine a comparative test (which one) with open-text follow-ups (why). If you only have budget for one, pick the one that includes the why.
Sample size depends on the question.
For qualitative work, the principle isn't statistical significance. It's saturation: the point where new responses stop producing new insights. A review of 23 peer-reviewed studies found that 9 to 17 participants are usually enough to reach saturation, especially with a homogeneous audience and a narrow question (Hennink and Kaiser, Social Science & Medicine, 2022). Jakob Nielsen made a similar point for usability back in 2000: a handful of users finds most of the problems. As I wrote in 2023: "Running qualitative research studies with more than 15 people provides little additional benefit but adds a lot to the cost. Spend the extra budget on more studies, not more participants."
For preference, you want numbers, so you need more people:
| Goal | Respondents |
|---|---|
| Find what's unclear or unbelievable in one concept | 15 to 30 |
| Pick between 2 or 3 concepts, with reasons | 30 to 50 |
| A number you can defend to the board | 100+ |
One warning. With 2 concepts and 30 people, a 60/40 split is still within the noise. If the vote is close, the open-text answers carry the decision, not the count. More on picking the right number in our guide to sample sizes for qualitative research.
1. Write down the decision first. "Which homepage headline do we ship?" or "Is this a category or a use case?" If you can't name the decision, you're not ready to test.
2. Make the concepts actually different. Testing 3 versions of the same idea with different adjectives teaches you nothing. In Wynter's Sameness Benchmark, only 12.1% of B2B SaaS homepage messaging was unique. Don't make your concepts as similar to each other as your market already is.
3. Keep the stimuli comparable. Same length, same format, same level of polish. If one concept has a nicer design, you're testing design.
4. Recruit your ICP, not whoever's available. Match job title, seniority, company size and industry. Your team, your customers and your friends on LinkedIn are not your market.
5. Always ask why. Every choice question gets an open-text follow-up.
6. Read the comments before you look at the counts. The numbers tell you what happened. The comments tell you what to do.
Pick the 5 to 8 that match your decision. Grouped by what they measure:
Clarity
1. In your own words, what is this offering?
2. Who do you think this is for?
3. What, if anything, is unclear or confusing?
Relevance
4. How relevant is this to the challenges you're facing right now? (1 to 5)
5. Which part of this speaks most to your priorities?
Value
6. How appealing is this? (1 to 5)
7. What would this help you achieve?
Believability
8. How believable are the claims made here? (1 to 5)
9. What would you need to see to believe it?
Differentiation
10. How is this different from what you use today?
11. Which other companies does this remind you of?
12. How unique does this feel compared to alternatives? (1 to 5)
Preference (for comparative tests)
13. Which concept do you prefer?
14. What made you choose that one over the other?
15. What would make the other concept better?
Intent and next step
16. How interested would you be in learning more? (1 to 5)
17. Would you share this with a colleague? Why or why not?
18. What questions would you need answered before taking a next step?
Language
19. What would you call this?
20. Is there a word or phrase here that put you off?
Note what's missing: "Would you buy this?" In B2B, the person answering rarely holds the budget alone, and as I wrote in 2023: "qualitative research cannot accurately answer 'would you buy' type of questions." Selling is also research, and it's the best method for questions about future behavior. Use concept tests to understand reactions, not to forecast revenue. The research agrees: a meta-analysis by Vicki Morwitz, Joel Steckel and Alok Gupta found stated purchase intentions predict sales far worse for new products than for existing ones, and better when collected comparatively (International Journal of Forecasting, 2007).
We wanted to know how to format Wynter's customer stories. Concept 1 was a classic case study: challenge, solution, results, with headers and a results summary. Concept 2 was a narrative customer story with a contrarian headline and pull quotes.
We ran a preference test with 50 product marketers: senior PMMs up to heads of product marketing and senior directors, 60% at companies with 1,000+ employees.
The result: 50% vs 50%. A perfect tie.
If we had only looked at the vote, we would have learned nothing. The open-text answers explained everything.
The people who picked the narrative version said it was relatable, felt less self-serving, and was something they'd share. A director of product marketing: "the narrative driven version would be something I'd be more likely to share organically." A senior PMM: "for middle/bottom of funnel? A social post? I would read this."
The people who picked the classic version wanted to skim. A director of product marketing: "For a B2B market, direct is key." A senior PMM: "I'd even suggest bringing the results up to the top."
Several people on both sides said each format has a job. Narrative for top of funnel and social. Structured for late-stage proof, when a buyer needs numbers for their business case. One director suggested combining the structure and skimmable formatting of one with the style and prose of the other.
So the decision wasn't "which one wins." It was "which one for which stage." A tie is not a failed test. A vote-only concept test would have called it inconclusive and moved on.
| Concept testing | Message testing | A/B testing | |
|---|---|---|---|
| Question | Is this the right idea? | Do these words land? | Which version converts better? |
| When | Before you build or write the final version | Before you publish | After you publish |
| Who | Your ICP, recruited | Your ICP, recruited | Your live traffic |
| Gives you the why | Yes | Yes | No |
| Traffic needed | None | None | A lot |
That last row matters in B2B. Most B2B websites don't have the volume to run even one valid A/B test per month. As I wrote in 2023: "So they run their A/B test, get 11 vs 19 signups, declare their tests a win, and tell their team what they learned from the experience. In reality they learned nothing at all, and are now running enthusiastically in the wrong direction."
Concept testing comes first. Once you've picked the concept, use message testing and copy testing to sharpen the words.
And what about test marketing? That's launching in a limited market to see real sales. It's the most reliable signal but also the most expensive and the slowest. Concept testing is how you decide what's worth test marketing.
And a note on synthetic respondents, since every research vendor now offers AI personas. I watched a demo of a synthetic users tool where "persona building" was essentially asking an LLM to roleplay as your ICP. That's not research. That's fan fiction.
As I wrote in 2026: "LLMs don't know what your target buyer thinks. They know what has been written about people like your buyer. When a real VP of Eng tells you 'we evaluated 5 vendors and compliance killed 3 of them,' that is ground truth. When an LLM says it, it is improv."
Research exists to create certainty. You do it because you're about to commit budget, reputation, roadmap and 3 to 6 months of team effort. A research company founder whose platform offers both told me customers try synthetic, then come back to human research, because they read the outputs and wonder whether it reflects reality. The only way to check is to run human research, and now you're paying twice.
Synthetic has its place: low-stakes decisions, broad B2C audiences, early brainstorming, and catching obvious problems. We built a free AI persona grader at Wynter for exactly that. It will reliably find a vague headline or an undifferentiated value proposition. What it can't have is "a real budget, a renewal coming up, or a boss asking about this quarter. For anything you're actually betting on, test with humans who are in-market."
Remember when Docusign launched "Intelligent Agreement Management" as its main homepage message? It got teased and publicly mocked for a long time. Surely people inside the company knew it was meaningless jargon and fought against it, and lost. If all you have is opinions, the company goes with the highest-paid person's opinion. They later switched to simple language. They could have gotten there before launch with a test.
What is concept testing? Concept testing is a research method where you show early versions of an idea, like a positioning angle, a name, a value proposition or a campaign, to your target market and measure which they understand, believe and prefer, and why.
What is an example of concept testing? Showing 50 product marketers two versions of a customer story, a narrative one and a classic case study, and asking which they'd read and share, and why. Wynter ran this test and got a 50/50 split, explained by the open-text answers.
How many people do you need for a concept test? 15 to 30 for qualitative feedback on one concept, 30 to 50 to choose between 2 or 3 concepts, and 100+ for a number you need to defend.
What questions should you ask in a concept test? Questions about clarity, relevance, value, believability and differentiation, plus which concept they prefer and why. Avoid "would you buy this?" because people can't reliably predict their own future purchases.
What is the difference between concept testing and test marketing? Concept testing shows an idea to a recruited sample before launch. Test marketing launches the real product in a limited market and measures actual sales.
What is the difference between concept testing and A/B testing? Concept testing happens before launch with a recruited audience and tells you why people react the way they do. A/B testing happens after launch with live traffic and only tells you which version performed better, if you have enough traffic.
Brand perception survey · Competitive analysis template · Voice of the customer for B2B · B2B buyer personas · Copy testing
.jpg)
.png)