Home/Work/ Retail Inventory
Retail · India · Operations Research

Stock control that
matches the floor.

Real-time inventory across 9 warehouses with a heuristic reorder engine. Cut dead stock by nearly half, increased throughput by 22% and went live in 14 days.

ClientRetail chain, India (NDA)
SectorRetail · India
Duration14 days to production
PracticeOR & Product Engineering
−44%
Dead stock
+22%
Throughput
14d
Time to production
The challenge

Nine warehouses. No single source of truth.

A retail chain operating across 9 warehouses and 40+ stores in India had been running inventory on a legacy ERP system that updated stock levels nightly. By morning, the numbers were already wrong.

Stock-outs on fast-moving SKUs coexisted with dead stock on slow movers filling valuable floor space. The reorder process was manual — purchasing managers received weekly Excel reports and placed orders based on gut feel and supplier relationships rather than data.

The client needed real-time inventory visibility across all locations and an automated reorder system that would eliminate both stock-outs and the accumulation of dead stock.

Our approach

Real-time event stream + heuristic reorder logic.

Phase 01 · Days 1–4

Event Stream Integration

Connected barcode scanners at all 9 warehouse gates via MQTT to a Kafka event stream. Every goods-in, goods-out and transfer event updates stock in under 500ms — replacing the nightly batch job.

Phase 02 · Days 3–8

Reorder Heuristic Engine

Demand-weighted safety-stock calculation per SKU per location. Dynamic reorder points that account for lead time variability and seasonal velocity. Auto-generates draft POs for purchasing manager approval.

Phase 03 · Days 7–12

Dashboards & Alerts

Operations dashboard showing real-time stock, velocity ABC classification, stock-out risk signals and dead-stock ageing. Mobile app for warehouse staff receiving and transfer confirmation.

Phase 04 · Days 12–14

Go-Live & Handoff

Parallel run for 2 days alongside legacy system. Cutover on day 14. Training for purchasing team on interpreting reorder recommendations and overriding system decisions with reason codes.

Similar inventory problem? Let's talk.

Book a call →
Tech used
Python Apache Kafka MQTT FastAPI React Native PostgreSQL Redis Metabase
← All case studies