BETA CUSTOMER LOYALTY & CHURN

Pre-test the intervention. Don't just predict the churn.

Your churn model scores who is likely to leave. It can't tell you what happens when you raise the price or redesign the tier structure, because none of that has happened yet.

Join the waitlist → See how it works

WHAT YOU CAN SCREEN Price increase · Save-flow offer design · Loyalty tier redesign · Silent-disengagement triggers · Win-back targeting · Replenishment cadence

CUSTOM STUDIES ONLY — NOT SELF-SERVE.
SYNTHETIC POPULATIONS EXPLORE AND SCREEN. REAL DATA VALIDATES.

EMERGENT COHORTS · KIDS' NUTRITION STUDYRESOLVING

AUTHORITY-RESPONSIVE · 60% CONVENIENCE-LED · 40%

Nobody assigned these groups. They separated by how each simulated customer answered the same four retention offers.

Behaviour you didn't program, from customers who don't exist.

Most segmentation draws the boxes first — age, spend tier, RFM decile — then sorts people into them. Simulatte does the opposite.

A calibrated population, not an average

Each persona carries its own price sensitivity, trust, switching threshold and objections, and reasons from them. Built from your ICP, from real interviews, or seeded from your CRM states.

Emergent cohorts, not assigned segments

Personas cluster by how they responded, not by who they are. Two customers with identical demographics can land in opposite groups because they leave for different reasons.

Churn as a response, not a score

Each cohort drifts for a different reason, so each answers to a different intervention — including the one that pushes it further away.

CHURN SCORING · WHAT YOU HAVE

INTERVENTION SIMULATION · WHAT THIS ADDS

Answers who will churn

Answers what happens if we do this

Averages hide the segment that reacts badly

Surfaces where an intervention backfires

Scores the book you already have

Screens the roadmap you haven't shipped

THE SAME OFFER, TWO COHORTS

One lever wins a cohort. Another loses it.

Doctor-backed programme

A+0.358

B+0.124

Subscribe & save

A+0.097

B+0.403

Discount wallet · incumbent

A+0.088

B+0.235

A = AUTHORITY-RESPONSIVE · B = CONVENIENCE-LED. THE INCUMBENT PROGRAMME NEVER WINS EITHER COHORT.

We do not out-rank a purpose-built ML churn model, and we say so on the benchmark page. The persona's edge is that it also gives you the why and the intervention.

Hands paused over a laptop keyboard at a kitchen table at night, mid-decision. — THE SAVE-FLOW IS WHERE RETENTION IS ACTUALLY WON OR LOST — AND THE ONE PLACE YOU CAN'T AFFORD TO GUESS.

WORKED EXAMPLE · ANONYMISED CLIENT · KIDS' NUTRITION

The discount wallet was never the winning lever.

A children's nutrition brand wanted to know whether its existing discount programme was holding its base. 90 simulated existing buyers — parents of 2–6 year olds seeded into existing-customer states — were put through four candidate retention programmes.

THE BASE, BEFORE ANY INTERVENTION

9% firmly loyal 70% on the fence 21% effectively gone

A parent crouching beside a young child in a supermarket aisle, comparing two nutrition products.

COHORTS CLUSTERED BY RESPONSE — NOT BY DEMOGRAPHICS OR LAPSE STATUS
Cohort	Share	Drifts because	Best lever	Uplift
Authority-Responsive	60% (n=53)	family-legitimacy doubt + efficacy doubt	Doctor-backed growth programme	+0.358
Convenience-Led	40% (n=36)	reorder friction + price hesitation	Subscribe & save	+0.403

← SCROLL THE TABLE SIDEWAYS TO SEE ALL COLUMNS

MODELLED FORWARD MONTHLY REVENUE ON THE AT-RISK BASE

Matched lever per cohort

+138%

Blanket discount for everyone

+54%

Revenue figures are modelled forward monthly revenue on the at-risk base, with intent mapped to a fixed baseline price. A design partner calibrates them against real data. These are modelled, not measured outcomes. Taste, the real #1 driver in this category, was deliberately out of scope: it was neither seeded nor elicited, and it needs a real product trial.

Tell us the retention decision you're facing.

Loyalty & Churn is in beta and runs as custom studies only. We scope each one against a specific decision rather than handing over a self-serve tool, and we're taking a limited number of engagements while the methodology is being hardened.

BETA · CUSTOM STUDIES ONLY · NOT SELF-SERVE.
SYNTHETIC POPULATIONS SCREEN AND EXPLORE. REAL DATA VALIDATES.
WE'LL REPLY FROM A PERSON, NOT A SEQUENCE.

Got it. We'll come back to you from a person, usually within one working day.