Guide
What is synthetic research?
TL;DR
- Synthetic research answers market research questions with simulated respondents, used instead of or alongside real people.
- Neon Blue builds simulated audiences: AI models that answer the way a defined group of people would, grounded in Neon Blue's own paid studies and AI-moderated interviews with real people.
- Fund analysts, product and growth teams, B2B GTM teams and marketing leads use it for questions they can't test live or can't get enough people for.
- It's weak on brand-new categories and shouldn't make a final high-stakes call without a real-people check.
- Quality comes down to what the simulation is grounded in: a persona prompt is a long way from a model built on real interviews.
What do the terms in synthetic research mean?
| Term | What it means | Real people involved? |
|---|---|---|
| Synthetic respondent | One simulated person answering a question | Only in the data behind it |
| Synthetic panel or simulated audience | A group of simulated respondents matched to a target audience | Only in the data behind it |
| Digital twin | A simulation of a specific person or segment built from their own data | Yes, as the source |
| AI-moderated interview | An AI interviewer talking to a real person | Yes, live |
| Synthetic data | Generated rows that copy the statistics of real data, often for privacy or ML training | No |
Synthetic research usually means the first two. Neon Blue runs both sides: AI-moderated interviews with real people, and simulated audiences grounded in those interviews.
How is a synthetic research model built?
The cheapest version is a prompt-only persona: you describe a person to a general chatbot and ask it questions. Its answers tend to be agreeable, too similar to each other, and based on what the internet says about a group.
A step up is a model tuned on published surveys and reviews. The spread of answers is better, but it knows little about your category or your customers.
The version worth paying for is grounded in first-party interviews and studies, then asked new questions. It's the only version where you can measure how far a new question sits from what real people actually said.
Who uses synthetic research, and for what?
| Reader | Typical question | Why simulation fits |
|---|---|---|
| Fund analyst or PM | Will regular guests keep visiting after the price increase? | It feeds one model variable and is needed before the print |
| Product or growth team | Which onboarding flow or paywall will convert better? | Not enough traffic to reach significance, or the build takes a quarter |
| B2B GTM team | Which value prop lands with mid-market ops leaders? | Testing on a live list burns it |
| Enterprise marketing or product lead | Which of these 12 concepts deserves a real test? | Budget to field three, ideas for twenty |
The common thread is statistical power. Teams with lots of traffic should keep experimenting live. Teams without it, or with decisions they can't test live, are who simulation is for.
When should you skip synthetic research?
Skip synthetic research, or pair it with real people, for:
- A category so new that nobody has talked about anything like it.
- Sensitive topics where people answer differently in private.
- Final go/no-go calls on big launches.
- Anything where a regulator or a court needs real respondents.
What does a small synthetic study look like?
Hypothetical: a snack brand's marketing lead has 12 flavor concepts and budget for one real concept test.
- Define the audience: adults who bought salty snacks in the last month, with a boost on 18 to 34.
- Show each concept as a short description with price and pack size.
- Ask purchase intent on a 5-point scale, which flavor it would replace, and what would stop them buying it.
- Rank the 12 on intent and on how much they'd cannibalize the current line.
- Take the top 3 to a real concept test with the same questions, so the two reads can be compared.
Want this run on your own question? Neon Blue's first study is free: start a free first study.
How does Neon Blue run synthetic research?
Neon Blue grounds simulated audiences in its own paid studies and AI-moderated interviews with real people. Results are always labeled simulated or real and kept separate. Creative pre-testing and UI/UX testing are in build.
Common questions about synthetic research
Is synthetic research the same as synthetic data?
No. Synthetic data means generated rows that copy the statistics of real data. Synthetic research means simulated people answering questions.
Will synthetic research replace surveys and A/B tests?
For high-stakes calls, no. It covers the questions you'd otherwise skip and shrinks real tests by screening options first.
How do I know if it works for my questions?
Run a paired test: ask the same questions to a simulated audience and to real people, then compare rankings and the spread of answers as well as averages.
How do I start a free first study?
Bring the one question you'd most like answered before your next planning meeting. Neon Blue's first study is free. Start a free first study