// work / software development

Multi-warehouse inventory and order platform

A web platform that replaced spreadsheets and a legacy desktop app across three warehouses, with barcode receiving, low-stock alerts and role-based access.

Client
Regional distribution company
Industry
Wholesale distribution
Year
2025
Role
Lead developer
Duration
11 weeks
Rows of tall industrial shelving in a large warehouse
Photo: @carmendis on Unsplash

Context

A wholesale distributor was running three warehouses on a mix of an ageing desktop stock system, shared spreadsheets and WhatsApp messages. Stock counts took a full day per site, orders were keyed in twice, and nobody could answer “how much of this SKU do we have, right now, everywhere?” without three phone calls.

The problem

The business did not need a generic ERP. It needed one source of truth for stock and orders that warehouse staff would actually use on the floor, with the office getting reliable reporting out of the same data.

Three constraints shaped the build:

  • Handheld first. Receiving and picking happen on Android scanners with gloves on, so those screens had to be large, fast and forgiving.
  • Existing data had to come across. Eight years of SKUs, suppliers and price history lived in the old system’s Access-era database.
  • No downtime for sales. Orders are placed throughout the day; the switch-over could not stop them.

Approach

I ran a two-day discovery on site, walking receiving, put-away, picking and dispatch at the largest warehouse and mapping every place a spreadsheet filled a gap. That produced a short written scope and a data model we agreed before any code was written.

The system is a React front-end over a .NET 8 API on SQL Server, hosted on Azure App Service. Stock movements are an append-only ledger, so every quantity on screen can be traced back to a scan, an order or a correction with a user and a timestamp. Redis caches the hot stock-level reads the dashboard and the sales screens hit constantly.

Receiving uses the scanner’s camera and barcode gun: scan the purchase order, scan each line, confirm quantity, done. Low-stock alerts run every fifteen minutes and go to email and a dashboard tile, with thresholds per SKU per warehouse rather than one global number.

Migration was rehearsed three times against copies of the old database. On cutover day the old system went read-only at 6 p.m., the final import ran in forty minutes, and the morning shift started on the new platform.

Outcome

Stock counts dropped from a full day to under four hours per site, because counts are now scan-and-confirm rather than clipboard-and-retype. Double entry of orders disappeared, and the sales team can quote delivery dates from live stock across all three warehouses. The operations lead described the biggest change as “we stopped arguing about numbers.”

The platform has run for a year with monthly feature releases and no unplanned downtime.

Need something like this?

This is the kind of work I do under software development. Tell me about your version of the problem.