Ask your market anything. First answers in minutes.
Put a question to simulated customers, modelled on real people, in the market you sell to. See who says yes, who says no, and why, before you spend a week or a budget finding out.
Illustrative example. Simulated customers, modelled on real people. Not real respondents.
What a simulated customer is.
A simulated customer is a detailed profile of one kind of person, modelled on real people: where they live, what they earn and spend, how they decide, who they ask. Ask a few hundred of them and you get a spread of answers, not one tidy average.
Modelled on real people
Each one has a life behind the answer: a household, a budget, habits and the way they actually shop. That’s why two people in the same city can disagree, and why the disagreement is often the useful part.
- Decides byResearching heavily, checking credentials
- TrustsIndependent ratings over brand claims
- ShopsWeekly online order, top-ups on foot
Simulated customer. Not a real person.
Ready-made in 12 countries, or built from your market
Ready-made in 12 countries, or built from your market
Start with an audience that already exists, or describe the people you sell to in a sentence and Simulatte builds the audience for you. Some countries start small and grow as we add people.
Find the groups inside a market
Find the groups inside a market
Not everyone in a market wants the same thing. Simulatte shows the groups that answer differently and how each one answers, so you know who you’re actually winning. Group shares follow how the audience was built (census or your customer mix), not a measured market size.
An agent that knows your business
An agent that knows your business
Before it asks anything, the agent learns your products, your past work and the files you share. It sweeps the web for market, competitor and price context, and shows every source. It remembers every study, so you can ask across all of them.
Six ways to ask.
Pick the one that fits the decision. The plan shows the cost before anything runs, with a credit ceiling you approve.
| Method | What you get | Typical time | Credits |
|---|---|---|---|
| Quick ask | One question. Themes, shares and quotes, with who disagrees. | Timeabout 2 min | Creditsfrom 300 |
| Survey | Your questions, answered by as many simulated customers as you choose, split by group. | Time10–30 min | Credits5–8 per answer100 people × 5 questions: 2,500–4,000 |
| Focus group | Six simulated customers talk it through. Ask up to 20 follow-ups. | Timesame day | Credits4,000 |
| Visual test | Show ads, packs or screens. See which one wins, and what people noticed. | Time3–4 min3 ads, 30 people | Creditspriced per studyshown in the plan |
| Price test | Directional: how price options rank, who might trade down, and why. Check price levels with real buyers. | Time10–30 min | Creditspriced per studyshown in the plan |
| Message test | Which line people believe, which makes them act, and who differs. | Time10–30 min | Creditspriced per studyshown in the plan |
Then check it with real people. AI interviews →
Then check it with real people.
When a decision matters, send the same questions to your own customers through an AI interview link. Both sets of answers sit side by side, so you see where they agree and where they differ.
Simulate first, then check Live
A first read in minutes. Then spend your customers’ time only on the questions still open.
Both at once Live
Run simulated customers and real interviews on the same study and watch both fill in.
| Question | Simulated customers | Real people | Read |
|---|---|---|---|
| Would you pay 10% more for the refill? | SimulatedYes 46% | Real peopleYes 9/20 | ReadAgree |
| What would stop you? | SimulatedDoubt about the claim | Real peopleDoubt about the claim | ReadAgree |
| Where would you buy it? | SimulatedOnline, with the weekly shop | Real people“In store, so I can see the pouch” | ReadDiffer |
Directional, and honest about it.
Simulated customers give you a fast, early read: which way people lean, who disagrees and why. Use them to narrow the options. Use your own customers to confirm what you’ll bet on.
Simulated customers are not real people.
How we test accuracy →Better than chance in all three. A model trained on each company's own data still does better in all three, and we publish that too.
Priced in credits. Quick asks from 300 credits. Surveys cost 5–8 credits per answer, so 100 people answering 5 questions is 2,500–4,000. A focus group is 4,000.
See pricing →How simulated customers are built
How simulated customers are built.
Every simulated customer starts from census and market data about who lives and buys in a market. AI then writes each one up as a full profile, so answers come from a specific person, not an average.
- Real population dataCensus-calibrated in the US. In other countries, built to each market’s population structure.
- Your customer dataGround an audience in your own customer data or interviews. Rolling out with design partners now, not yet on by default.
- Your contextThe agent adds what it knows about your business: your products, rivals and prices, with every source shown.
- PopulationDigital project manager in Hackney, partnered, no children
- Buyer voiceResearches heavily before a first purchase; checks sustainability claims
- Your contextSees your $34 coffee gift box and three rival prices
Asked about simulated customers.
Are simulated customers real people?
No. They are simulated customers, modelled on real people. Treat their answers as an early, directional read, and confirm what matters with your own customers.
How is this different from asking ChatGPT?
A chatbot gives you one plausible average answer. Simulatte asks hundreds of distinct simulated customers in your market, shows how the answers split, who disagrees and why, and lets you check the result with real people.
Which countries can I ask?
Ready-made audiences cover the US, the UK, the UAE, Saudi Arabia, Qatar, Germany, the Netherlands, France, Spain, Italy, Sweden and India. Some start small and grow as we add people. You can also build an audience from a description of your market.
How fast is it?
A quick ask takes about two minutes. Larger studies take around 10 to 30 minutes. Batch mode returns results in about an hour for 20% fewer credits.
How accurate is it?
Directional. We test against real outcomes sealed before we predict, compare with chance and with a model trained on the real data, and publish the code and results, including where the trained model does better. See how we test.
Can I use it from Claude?
Yes. With the Simulatte connector for Claude you can launch studies, get results and ask your simulated customers without leaving Claude.
Ask the question you’ve been putting off.
A first answer in minutes. Check it with your own customers this week.