Multi-store inventory optimization

The business need

A retailer needed to bring together sales and stock records from different POS systems, along with incoming purchase orders. The application had to support decisions about where to move existing stock and what to order next.

The requirement was more than a view of inventory. Recommendations needed to follow the retailer’s rules and retain enough detail for users to understand how each calculation was made.

What I built

I built an application where authorized users can upload files, manage source connections and calculation settings, review recommendations and download Excel reports. Source adapters normalize stock and sales records from spreadsheets and APIs, matching products by their codes.

Each calculation retains its settings, supporting data and warnings. Users can return to previous runs to review how a recommendation was produced.

How the logic works

Transfers follow stock coverage and movement rules

Demand estimates give different weights to recent and older sales periods. The transfer calculation considers expected stock coverage and projected stockout timing, while respecting protected stock, allowed routes, quantity increments and minimum transfer sizes.

Purchasing accounts for stock already on its way

Purchase recommendations are calculated after proposed transfers and incoming supply. Lead times and expected arrival dates affect when that stock becomes available. Target coverage, minimum order quantities and a minimum calculated-need rule then shape the suggested purchase.

Challenges handled

An incomplete stock source is not the same as zero stock. The application distinguishes these cases, keeps unresolved product identities out of calculations and explains the data issues. Required sources that are stale or unavailable can block a complete run.

Incoming deliveries may arrive on different dates, so they cannot all be treated as immediately available. Duplicate incoming-order uploads are rejected, and calculation history remains reproducible as settings change.

What the system provides

Real-data runs produced stock-transfer recommendations, manufacturer-purchase recommendations and downloadable reports. Users can review those outputs alongside the calculation history and data warnings.

The application supports decisions; it does not execute stock transfers or place purchase orders. Some source data is supplied through file uploads.