Questions
We already know our customers. What would a simulated audience add?
TL;DR
- A simulated audience adds answers about things your customers haven't seen yet. Product analytics, CRM notes and transaction panels show what people did with what was offered, not how they'd react to a new price, feature, message or product.
- 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.
- It also covers segments too small to test and people who aren't your customers yet.
- If the decision can be A/B tested quickly on enough traffic, test it live. Simulation is for everything else.
What can your customer data answer, and what can't it?
| Question | Your data | Simulated audience |
|---|---|---|
| What did users or customers do last quarter? | Yes | Not needed |
| Why did they churn or go quiet? | Partly, from exit surveys and call notes | Yes, with reasons by segment |
| How would they react to a price or paywall change you haven't made? | No | Yes |
| Would they want a feature that would take a quarter to build? | No | Yes, labeled as a pre-read |
| What do your competitor's customers think? | No | Yes |
| What does a segment of a few hundred accounts prefer? | Too small to test | Yes, if the grounding data covers it |
| Which of 15 messages works with a new ICP? | Only by burning the list | Yes |
| How will diners react to a chain's price increase before it shows up in card data? | No, transaction data lags | Yes |
Behavioral data records how people responded to the options you offered. Simulation lets you ask about options you haven't offered.
What does a simulated audience add for each team?
Product and growth teams have the analytics but often not the traffic. A paywall test on a page with modest traffic can take months to reach significance. A simulated read narrows the variants first. An API for calling it from code is in build.
B2B GTM teams know their current customers well from CRM and call notes. The open questions are about the next segment, buyers they haven't sold to yet. A synthetic ICP panel lets them test positioning before outbound.
Fund analysts already buy transaction data, which shows what happened after the fact. A simulated consumer read covers intent around an event before it reaches the numbers.
Enterprise marketing and product leads have brand trackers and past studies. Those are snapshots, and a simulated audience lets you ask the follow-up question without fielding again.
How would a simulated audience help diagnose churn?
Hypothetical: a subscription meal-kit company sees churn rising, and the growth team has two theories: price, or menu fatigue. Analytics show who left but not why, and the last exit survey had too few responses from households subscribed for more than a year, the segment that matters most.
- Audience: current and lapsed subscribers, cut by tenure, household size and plan.
- Questions: the main reason they'd pause or cancel in the next 3 months; how likely they'd be to stay if the menu changed weekly at the same price; how likely if the price dropped and the menu stayed the same; which they'd pick among a lower price, more recipe variety or more flexible delivery; and what they'd switch to instead.
- Read: compare stay-likelihood under the price and menu scenarios, by tenure.
- Check: run the top fix as a live test on long-tenure subscribers before changing pricing for everyone.
The analytics identified which subscribers were leaving, and the simulated read gave the team a reason to test the menu fix before touching price.
Want this run on your own question? Neon Blue's first study is free: start a free first study.
When don't you need a simulated audience?
- You have enough traffic to A/B test the decision quickly and cheaply. Test it live.
- The question is about past behavior your data already records.
- The decision is small and easy to reverse.
Neon Blue grounds its simulated audiences in its own paid studies and AI-moderated interviews with real people.
Common questions about simulated audiences and customer data
Isn't our own data more accurate than a simulation?
For what people already did, yes. A simulation is for what they haven't done yet: new prices, features, messages and products.
Can a simulated audience include people who aren't our customers?
Yes. Competitor customers, prospects in a new ICP and people who churned are some of the main uses.
We have lots of traffic. Should we just A/B test?
If the test is fast, cheap and reversible, yes. Simulation fits decisions you can't test live or can't get enough traffic for.
How do I start a free first study?
Bring the question your data can't answer that keeps coming up in planning. Neon Blue's first study is free. Start a free first study