The problem
Coffee roasters carry inventory that generic tools don't model: green beans by origin lot, roasted stock with short shelf lives, and supplier lead times measured in weeks. Run out of a core origin and you're pulling products off the shelf; over-order and cash sits in a warehouse. Most roasters manage this in a spreadsheet updated from memory.
What we built
- Roaster-shaped inventory: stock tracking that models how a roastery actually works, not a generic SKU table.
- Run-out prediction: consumption-based projections of when each item runs dry, so ordering is driven by data instead of a warehouse walk-through.
- A reorder loop that closes: the reorder note isn't a dead end — mark a lot as ordered and it's tracked through to delivery, with the dashboard showing "ordered, arriving in 4 days" instead of re-flagging a lot you already handled.
- KPI bar: lots at risk, open orders, and total on-hand value, visible the moment the dashboard loads — no digging into individual lots to know if today needs action.
- Real persistence and auth: session login and a PostgreSQL backend — deployed, tested, production-grade.
How it was shipped
Built to full MVP depth on our standard production pattern: Node/Express over PostgreSQL, session auth, end-to-end verification before deploy. The purchase-order lifecycle also replaced a per-lot forecast query with a single grouped query across the whole book — the dashboard stays fast regardless of how many lots a roaster is tracking. RoastRadar is the proof-case for our micro-vertical thesis: pick one sharp workflow for one specific persona, and software that a horizontal tool can't match becomes buildable in days.
Try it
Open the live demo — no login required. Log usage against a lot, mark a reorder as ordered, and watch it move through to received. Demo data resets automatically.