Dutch fashion retailer Shoeby boosts revenue 2.96% and increases inventory turnover 4% with AI replenishment
Shoeby replaced manual min/max replenishment across 240 stores with WAIR's AI Replenisher, lifting forecast accuracy from 78% to 95%, cutting overstock from 22% to 5%, and generating €2.9 million in added revenue in the first year — with just 6 merchandisers managing the full network.
Background
With only six merchandisers managing replenishment across 240 stores and 48,000 SKUs, Shoeby could manually optimize only the top and bottom 5% of performing products. The traditional system's static stocking levels could not account for location-specific demand variation, and limited collections suffered from inventory stranded in the wrong stores. Shoeby needed a scalable, automated solution that a small team could operate with confidence in the AI's output.
What Was Implemented
- WAIR AI Replenisher: deep-learning model producing SKU-level, store-level demand predictions 14 days ahead
- Dynamic adjustment of min/max stocking limits directly in Microsoft Dynamics 365 ERP via ACA Fashion Software's XPRT integration
- Input signals: sales data, product information, local demand, individual store performance, and external data sources
- Merchandiser override capability and configurable business rules per store, category, or SKU
- 10-day implementation timeline
- Applied to 30% of replenishment stock in first year
Results
After implementing WAIR's AI Replenisher, Shoeby achieved: 4% increase in inventory turnover , 2% reduction in end-stock , and 2.96% growth in total revenue (approximately €2.9 million in added value in the first year). Forecast accuracy improved from 78.1% to 95.4% and overprediction dropped from 21.8% to 4.4% . These results are vendor-reported by WAIR.
Lessons
- A small merchandising team (six people) can manage replenishment at 240-store scale when AI handles the analytical workload, freeing humans to focus on exceptions and strategic overrides
- SKU-level, store-level predictions are essential in fashion; aggregate or category-level forecasts are insufficient for limited collections
- Rapid ERP integration (10 days) is achievable with pre-built connectors, lowering the implementation barrier
- Vendor-reported metrics should be verified against independent data before citing as definitive