Study
from CHF 3,000
One decision, two versions or more, your simulated customers in their world. The report in 48 hours.
From November 2026
Test your decisions before you launch them.
A product, a price, a message: thousands of simulated customers react, in their own world. In 48 hours, you know who buys, who walks away, and why.
One dot, one simulated personIn red, those who react
Today, decisions are tested on real customers, after launch.
You decide, you launch, then you discover in the sales figures that customers are missing, without knowing which ones or why.
- Market researchSeveral weeks and tens of thousands of francs.
- A/B testAfter launch, with traffic, on real customers.
- AI panelFast, but nobody knows when to trust it: depending on the model, the same question gives 27% or 98%1.
- SeuratBefore launch. In 48 hours, from CHF 3,000, without exposing a single real customer, with two AIs checking each other.
1. Same profiles, same question, three models: 27, 83 and 98%. MatrAIx paper, arXiv 2608.04205, 2026.
Every dot is a person.
Georges Seurat painted an entire Sunday with thousands of dots. Up close, you only see dots. From afar, a crowd, a park, a world.
Seurat does the same with the people your decisions affect: customers, policyholders, voters, users. Every dot is a simulated person, calibrated on the Swiss population or on your own data, with a biography, a budget, habits and a memory.
Georges Seurat, A Sunday on La Grande Jatte, 1884–1886, Art Institute of Chicago. Here in 11,000 dots: move your cursor closer.
A question, customers, a world. A report in 48 hours.
- 1
The question
The decision to test, in two versions or more. Here: free delivery from CHF 50, or from CHF 70?
- 2
The customers
2,000 simulated customers in three segments, calibrated on the Swiss population or on your customer base, with no personal data.
- 3
The world
Your real website, a competitor offering free delivery from CHF 40, the Christmas peak.
- 4
The report
Who changes behaviour, segment by segment, why, and whether the AIs agree. Every figure links back to the answers behind it.
AFree delivery from CHF 50
BFrom CHF 70
Price-sensitive800
38%
46%
Loyal700
7%
9%
Occasional500
22%
31%
- Your real website
- A competitor delivers free from CHF 40
- The Christmas peak
Order less often
Gap between the AIs: 3 points at most. They agree.
Fictional example. Hover overTap a dot to read a customer’s answer.
Others simulate customers. Seurat simulates the world around the decision.
Every person sees your competitors, the season, the media, their neighbours. Seurat tests your real product in that world.
- The decision to test
- 1People: customers, policyholders, voters
- 2Your product: website, app, chatbot
- 3Competitors
- 4Media and neighbours
- 5The season and events
- 6The rules: prices, stock, lead times
- What if the leading brand cuts its price by 20% in week 3?
- What if an article comes out the day before launch?
- What if a competitor delivers free from CHF 40?
Don’t take an AI’s word for it. Neither do we.
- 1
At least two AIs, from two providers.
Their disagreements are shown, never silently averaged. You know when to trust the result.
- 2
Always A against B, on the same population.
Versions are compared with each other rather than as absolute rates, unless calibration is documented.
- 3
Every prediction is sealed before the facts.
Then it is compared with reality. The gap is published, whether it flatters us or not.
First public proof: the Swiss federal vote of 29 November 2026.
Simulated voters, calibrated canton by canton, vote on the four proposals. The prediction is sealed and timestamped before the vote, and its fingerprint is published here. The details are revealed after the vote, with the gap, canton by canton.
- Vote
- 29 Nov 2026 · 4 proposals · 26 cantons
- Fingerprint
- published before the vote
- Comparison
- 30 Nov 2026, public
56days to the vote
Start with a study.
Pilot
CHF 10,000 to 25,000
Three months, three to four studies, and access to the tool for your teams.
Early 2027
Subscription
from CHF 490 a month
Your simulated customers and your world, kept up to date. Your teams run their own studies.
During 2027
The simulated people and the world stay yours: every new study costs less and aims truer. Your data stays aggregated, never personal, in line with the Swiss FADP and the GDPR.
No more decisions launched blind.
Launching a product, changing a price or planning a campaign this winter? Let’s test it first. We are opening three first studies to retailers and brands in Switzerland.