#RestaurantBrain — feasibility study
A technical and commercial feasibility study for a CRM for restaurants that includes its own POS — the apps, the hardware, and the connections to the hardware and systems restaurants already run.
There is no product code in this repository yet. That is the point of the study.
#The verdict in five lines
- Technically feasible — nothing in it is unsolved. 125 engineer-months to a sellable v1, 190 through a second market.
- The brief's shape doesn't fit the work — 3 engineers over 24 months on ~EUR 900k becomes 5 engineers over 36 months on EUR 1.8–2.4M.
- Own the POS, sell it CRM-first, in Czechia, to counter-service chains of 2–8 sites. Market choice is a compliance decision; Czechia has had no fiscalisation obligation since 1 Jan 2023.
- Two commercial facts decide it, and both are testable in four months for under EUR 40k, before any code: will operators pay ≥EUR 89/site/month, and will an acquirer put ≥25 bps of payment residual in writing?
- The bootstrapped end state is an option, not a return — ~220 sites and ~EUR 475k ARR at month 47, still burning, against breakeven at ~800 sites. Capital to breakeven is EUR 3.0–3.5M over 5–7 years.
Start here: 00 — The whole thing in plain words — the whole thing in plain words, two pages, no jargon. Then 01 — Feasibility verdict for the go/no-go and the decisions it asks for.
#Contents
| Document | What it covers |
|---|---|
| 00 — Plain words | What you're building, what's quick, what's slow, and why — no jargon |
| 01 — Feasibility verdict | The go/no-go, the six decisions being asked for, the next 120 days |
| 02 — Architecture | Offline-first design, event-sourced order domain, data model, stack, ADRs |
| 03 — Hardware and devices | Certified SKU matrix, three site BOMs, peripheral protocols, fleet operations |
| 04 — Payments, fiscal, compliance | Card-present architecture, PCI scope, 17-market fiscalisation table, GDPR design |
| 05 — CRM and product scope | Guest identity, capture playbook, offline-redeemable vouchers, feature map by tier |
| 06 — Integrations | Incumbent POS reality, delivery aggregators, connector strategy and its true cost |
| 07 — Roadmap, team, budget | Phases, dates, headcount, P&L for both scenarios. All money and calendar numbers live here |
| 08 — Risks, tradeoffs, difficulties | Case for, case against, hardest problems, 28-entry risk register, kill criteria |
| Canon | The binding decisions every document obeys: stack, phases, scope tiers, numbers, terminology, 24 owned assumptions with named owners and due dates |
#Appendices
docs/research/ holds the raw material: nine domain appendices (r01–r09), three adversarial critiques that revised the estimates upward (c01–c03), and the cross-document consistency sweep (c04).
#How this study was produced
Twenty-one agents in five phases: nine domain researchers writing appendices in parallel, three adversarial critics attacking their conclusions, one architect fixing a binding canon so the outputs could not contradict each other, seven writers authoring the deliverable against that canon, and one editor sweeping for cross-document contradictions. A second pass applied the editor's 24 resolutions plus two owner directives.
Every figure in these documents carries an evidence label — [verified] (checked against a source), [known] (from model training data to ~May 2026), or [estimate] (reasoned guess) — and legal claims additionally carry a confidence grade and a named counsel-verification owner. Nothing here substitutes for local legal advice in a fiscalised market.