Concept Screening Survey Designer

How to use it

1. 1Paste the prompt, then add your concept.
2. 2Tell it the target audience and the decision to make.
3. 3Answer its clarifying questions to tighten the design.
4. 4Field it in your survey tool; ask for variants to A/B the concept.

See it in action

Concept screening survey — OatJoy canned oat lattes

Screener (qualify the right respondents):
1. Which have you bought in the past month? [RTD coffee / oat milk / plant-based drinks / none] — qualify buyers of RTD coffee OR oat milk.
2. Age band — qualify 25-45.
3. (Quota) split the sample 50/50 across the two positioning versions.

Concept exposure (neutral):
Show the concept description + an image. Use the SAME format for both positioning versions, changing only the headline:
- Version A (convenience): "Barista-style oat lattes — no fridge, no machine, anywhere."
- Version B (taste): "The oat latte that finally tastes like the cafe."

Core measures (rate 1-5, standard scales):
- Purchase intent: "How likely would you be to buy this?" (Top-2-box is your headline metric.)
- Uniqueness: "How different is this from what's available today?"
- Relevance: "How well does this fit your needs?"
- Believability: "How believable is the 'barista taste' claim?"
- Value: "At [insert price], how would you rate the value?"

Open-ends (the why):
- "What, if anything, appeals most?"
- "What would stop you from buying?"

Scoring & benchmark plan:
- Headline = Top-2-box purchase intent; compare A vs B.
- Read uniqueness × relevance together (high on both = whitespace; high unique/low relevant = niche).
- Set an action standard BEFORE fielding (e.g. proceed if T2B intent ≥ category norm and believability ≥ 3.5).

Bias flags:
- Don't precede the intent question with benefit-heavy copy that primes a yes.
- Avoid "Wouldn't you agree…" phrasing; keep scales balanced.
- Test price in the value question, not before intent, so price doesn't anchor everything.

Want a shorter 5-question quick-screen version, or the full thing formatted for your survey tool?

The prompt

Here's the start of the prompt. Download the free bundle for the full, ready-to-paste version — plus the installable Claude Skill and Custom-GPT instructions.

# Role
You are a senior quantitative consumer-insights researcher for food & beverage. You design concept tests that give an honest read — neutral wording, standard comparable measures, and a pre-set action standard. You never write a survey that flatters the concept.

# Context I'll provide
- Concept (paste below): [CONCEPT]
- Target audience: [AUDIENCE]
- The decision I'm trying to make: [DECISION]
- Anything to compare (e.g. two positionings, a price): [VARIANTS / PRICE]

# Your task
1. If the concept, audience, or decision is missing, ask for them before designing.
2. Write screener questions that qualify the right respondents (and quotas if comparing versions).

Frequently asked questions

What makes a concept test reliable?

A reliable concept test uses neutral, non-leading wording, standard comparable measures like Top-2-box purchase intent, the right qualified respondents, and an action standard set before you field it. This skill builds all of those in, so you get an honest read rather than a survey that inflates the concept's scores and misleads the go/no-go decision.

Can it compare two positionings or prices?

Yes. Tell it what you want to compare and it sets up a clean A/B — splitting the sample with quotas and holding everything constant except the single variable you're testing, whether that's a positioning headline or a price point. That isolation is what lets you attribute any score difference to the variable rather than noise.

Will it avoid biased questions?

Yes — that's a core design goal. It uses balanced scales, keeps benefit-heavy priming away from the purchase-intent question, and tests price where it won't anchor the whole survey. It also flags wording that could skew the read, because a leading survey is worse than no survey: it gives you false confidence.

Does it set sample size or significance?

It won't fabricate a sample size or promise statistical significance it can't know, but it advises on what you'd need and recommends an action standard to judge results against. For the read itself, cross-checking stated purchase intent against real demand data helps confirm whether what people say matches what they actually buy.