Custom Customer Simulation for Growth & Product Teams
Train a custom simulation on your actual customer and experiment data. Test ads, landing pages, pricing, and product variants against customers that behave like your own. See the likely outcome before you spend weeks and thousands on live traffic.
Chorus research · Accepted at ICML 2026 AgenticUQChorus in Practice
Chorus does not replace the real test. It shows which ideas deserve live evidence first.
Example: Ad Challenger Selection
Same ideas. Same goal. A different path to the decision.
Traditional Live Experiment
Weeks lost. Thousands spent. Unreleased ideas exposed.
Weak concepts reach real customers too.
Every candidate competes for traffic.
The decision stays blocked for weeks.
Clarity comes after spend, exposure, and delay.
With Chorus
Reach clarity before the expensive part begins.
Your trained customer model responds before launch.
Weak ideas drop. Strong ideas get refined.
Use real traffic to confirm, not to screen blindly.
Not a Generic Synthetic Panel
Generic synthetic panels approximate a market. Meshona trains a custom simulation on your historical customers, experiments, and observed outcomes. It moves closer to how your own customers actually decide.
Historical customer behavior, experiments, and observed outcomes shape your simulation.
Run new ads, offers, messages, and product variants against customers modeled on your own.
Compare the forecast with the real result, then use that outcome to train the next version.
Your data trains your simulation. Real outcomes make the next version more specific.
Accepted Paper
The paper shows how synthetic audiences can accumulate evidence, decide when warranted, and abstain when uncertainty remains.
Read the Accepted Paper
Custom Simulation Training
Bring your historical ads, variants, customer context, and observed outcomes. Meshona trains a custom simulation for your business, validates it against reality, and retrains it as new outcomes arrive.
Train Your Simulation