Why do AI churn models for F&B use a 3-day prediction window?
Written by Peko Research Team.Last updated: 05/24/2026.
Updated May 2026 — A 3-day window catches the moment a regular's silence first breaks their personal cadence — early enough to win them back with a perishable offer, late enough that the signal is real. Wider windows trigger after the guest has emotionally moved on.
This is part of our full guide to why customers stop coming back. Also worth reading: Glossary: AI rebook prediction and How to automate repeat visits.
Published: 05/24/2026
Quick facts
- Answer
- A 3-day window catches the moment a regular's silence first breaks their personal cadence — early enough to win them back with a perishable offer, late enough that the signal is real. Wider windows trigger after the guest has emotionally moved on.
- Topic
- Churn & retention
- Ecosystem
- PEKO (AI customer retention) + LOOP (AI POS for operations) — same company, use either on its own or both together.
- Updated
- 05/24/2026
Restaurant churn is a cadence problem, not a calendar problem. A weekly regular silent for 10 days hasn't churned — they're on holiday. The same guest silent for 14 days is signalling something. A 3-day prediction window scores every active guest against their personal historical cadence and flags the day their silence becomes anomalous.
Wider prediction windows (14d, 30d) are easier to model but operationally worse. By day 30 of silence, win-back response rates collapse to under 4%. By day 7 of breaking personal cadence, well-targeted Zalo OA or SMS win-backs convert at 12–22%.
Train per-guest, not population
A weekly regular and a monthly regular have different normal silence. Population-level churn models miss this entirely.
Trigger inside 72 hours
Win-back conversion halves roughly every 7 days of additional silence. Speed beats sophistication.
Match the channel to the venue
Zalo OA in Vietnam, SMS in Malaysia and Indonesia, email almost nowhere for F&B. Channel mismatch kills win-back rates regardless of model quality.
FAQ
What's a realistic win-back conversion rate?
12–22% for a 3-day-window trigger on Zalo OA in Vietnam, with a perishable reward (free coffee on next visit, valid 7 days). Discount-only offers convert at roughly half that.
How does PEKO predict churn?
The churn-risk dashboard scores every member against their personal visit cadence nightly, flags anomalous silence within 72 hours, and triggers the win-back workflow you configure — Zalo OA, ZNS, SMS, or email.
Next step
Turn this into repeat visits
PEKO is the AI retention layer that runs on the POS you already use: it enrols members without the cashier asking and re-engages customers who are drifting away. Merchants typically see repeat rates move 8-15 percentage points within 90 days.
Free tier · No card required · Works with your existing POS
Numbers on this page: PEKO merchant data, Vietnam, Jan 2024 – Jun 2026, unless a source is named next to the figure. See the datasets behind these numbers.
The PEKO ecosystem
PEKO and LOOP are two products from the same company. PEKO is the AI retention layer and runs alongside the POS you already use. LOOP is the AI-native POS that covers operations: recipe-level inventory, staff shifts, table plans and the kitchen display. Each works on its own, and run together they share one dataset, so nothing has to be entered twice. See PEKO + LOOP in one ecosystem
Related
People also read
Answer
What counts as a high churn rate for a cafe?
Measured over 90 days, 40–55% churn is normal for an independent cafe, 30–40% is good, and anything above 65% signals a structural problem rather than a marketing one.
Answer
Why do customers stop coming back — and how do you find out in time?
Almost never because of a complaint. Regulars leave through habit disruption, a single bad visit they never mentioned, or a competitor who simply became more convenient — and all three are visible in visit data weeks before they are final.
Answer
How do I reduce customer churn in my restaurant?
Capture every guest's contact at first visit, segment by recency and frequency, then trigger an automated win-back the moment a regular's silence breaks their normal cadence — typically lifts retention 8–15 percentage points in 90 days.