All case studies
Shelves that count themselves — from monthly stock take to live inventory
Specialty coffee chain, Malaysia · 12 pilot outlets · 4 SKU classes · 5 months
Measured outcomes
- ±1.9%
- Count accuracy
- -68%
- End-of-day count hours
- -36%
- Fast-mover stock-outs
- 89%
- Variances resolved in 24h
Starting point
Stock was counted once per shift when staff were tired, with a 5% error band absorbing real losses. POS and shelf counts diverged daily, reorder fired at a fixed floor regardless of draw-down rate, and month-end investigations meant retracing days of paperwork and CCTV.
“The shelf tells us what we have before the shift ends — and we finally know why the numbers didn't match.”
What we deployed
- Flush-mounted load cells per SKU bay with rack concentrators and store gateway
- Cloud time-series stock history with demand modelling by outlet, daypart and day of week
- Dynamic reorder points and variance ledger splitting wastage, theft and miscount in near real time
- ERP API integration with canonical unit mapping — grams on shelf to stock units in the ledger
What changed
- Replenishment moved from fixed reorder floors to demand curves per outlet
- Supply-chain and store teams share one stock picture instead of parallel count sheets
- Pilot playbook reused for regional rollout without changing the rack kit or gateway profile
Client names are withheld under commercial confidentiality; figures are drawn from post-deployment reviews agreed with each client.
