Demand forecasting
Seasonal, boosted-tree, LSTM and attention forecasts trained on each store's own sales — with a held-out accuracy score beside every one.
Retail operations platform
Every aisle, every shelf, predicted.
RetailMind forecasts demand, flags shrink, routes procurement and keeps every store team a step ahead — for the people who run the floor and the people who shop it.
- Forecasting methods
- 4
- Payment tenders
- 6
- Roles
- 4
Platform
One system for forecasting, shelves and people
A stockout in produce or a slow rack of bread gets caught before it costs a sale — because forecasting, stock, procurement and the register all read the same data.
Procurement
Purchase orders drafted from forecasts, supplier scorecards, onboarding records and outreach that's logged on every click.
Shelf & loss alerts
A review queue for stock, quality and loss-prevention flags, routed to the right person and closed by a human decision.
Register & customer app
Six tender types, loyalty points earned and redeemed, digital receipts, a shopping list and recommendations from real history.
Warehouse
Zone utilisation from live stock, a replenishment pick route by aisle, and Saturday staffing scaled from real traffic.
Profit & loss
Net sales, cost of goods at landed cost, shrinkage and loyalty liability — every figure computed from orders, to gross margin.
Built for every role
The right screens for the right people
Admin
Store analytics and P&L, team and access, purchase approvals and every store in enterprise mode.
Manager
Inventory, alerts and forecasts for the floor, with the responsibilities an admin assigns.
Staff
Tasks, the register and the alerts routed to them — nothing they aren't allowed to change.
Customer
Find products, build a list, check out, earn points and see offers and receipts.
Computed, not canned
Every number traces back to a record
Forecasts train on sales history each time you ask. Stock moves on every sale. The P&L is arithmetic over orders and costs. Switch stores and watch it change.
Editions
Run it anywhere
The same interface and the same API, with two ways to host it.
Nothing to install
- Runs entirely in your browser — no server, no sign-up
- Workspace saved on your device and works offline
- Hosted free on GitHub Pages and Hugging Face
Your own infrastructure
- FastAPI and PostgreSQL behind the same interface
- Shared by every store and user, with real accounts
- One command with Docker Compose
FAQ
Questions people ask first
Where does my data live?
In your browser. This edition runs the whole platform — database, API and forecasting — inside your tab, and stores your workspace on your device (IndexedDB). Nothing is uploaded; clearing site data resets it.
Is the forecasting real?
Yes. A seasonal-trend regression, gradient-boosted trees, an LSTM and an attention network are trained on the selected store's sales history each time, and every result shows its held-out accuracy (MAPE) so you can see which one to trust.
What happens when a sale is rung up?
A real order is created, loyalty points are earned or redeemed, and stock is drawn from real batches, first-expired-first-out — so shelf fill, reorder alerts and the P&L all move with the sale.
Is the Profit & Loss screen real?
Net sales, cost of goods sold, gross margin, shrinkage and loyalty liability are all computed from orders and each product's landed cost. It stops at gross margin because rent and payroll aren't modelled.
Can I try every role?
Yes. Sign in as an admin, manager, staff member or customer from the workspace accounts on the sign-in screen. Each role sees a genuinely different set of screens and permissions.
Can I run it on my own servers?
Yes. The server edition replaces the in-browser engine with the FastAPI and PostgreSQL backend in this repository, run with Docker Compose, for multi-user deployments.
Ready to see it run?
Open the console and pick a role — every screen is live.