What does an F&B CRM need in 2026?

Most systems sold as F&B CRM are contact databases with a reporting tab. They store phone numbers, count visits and produce a monthly export nobody opens. A CRM earns the name when it changes what happens to a specific customer next week, and that requires four layers working together.

Layer one — unified identity. Customer data arrives from the POS, the messaging channel, the booking system, delivery marketplaces and review platforms, and each source has its own idea of who the customer is. Pick one identity key — phone number in most markets, platform ID where a messaging super-app dominates — and resolve everything to it, including deduplication of the same person entered three ways. Skip this and every subsequent number is quietly wrong: your best customer looks like three average ones and your churn rate is computed on fragments.

Layer two — dynamic segmentation. Recency, frequency and monetary value scored together, refreshed nightly, producing a small set of segments an operator can act on. The critical property is not the number of segments; it is the refresh rate. A monthly export means a customer who churned three weeks ago is still classified as a champion and still receiving thank-you messages, which is worse than sending nothing.

Layer three — predicted churn per customer. A shared threshold cannot work across a base containing daily commuters and monthly diners. The CRM should hold a per-customer expected gap and surface deviation from it, ideally as a ranked action list rather than a chart. 'These eleven regulars are overdue' is a CRM output; 'churn rate is 47%' is a report.

Layer four — triggered messaging. Segment movement, not the calendar, causes sends: moved from Loyal to At Risk, crossed a reward threshold, has a birthday in five days, first visit 48 hours ago. Without this layer the first three are analysis, and analysis does not change revenue.

Implementation order matters as much as feature selection. Weeks 1–2: audit and export existing customer data, choose the identity key, write down your consent and data-retention policy. Weeks 3–4: connect sources, migrate and deduplicate. Weeks 5–6: stand up nightly segmentation and give managers a one-screen daily view. Weeks 7–8: turn on welcome, birthday and a rule-based win-back. Weeks 9–12: switch win-back from rules to predicted cadence once you have enough history, then run the first cohort review.

Venues that get all four layers live typically report repeat-rate improvements in the 15–25 point range and CLV gains of 35–55% over a year, alongside ten to eighteen hours a month of reclaimed marketing time. Venues that stop after layer one report a tidier contact list.