Cutting Inventory Costs by 34% for a Global Apparel Brand
Kunde: Vesture Global
Reduced inventory carrying costs by 34% and eliminated $6.6M in annual markdown losses through ML-driven demand forecasting across 120+ SKUs and 40 markets.
34% — from $14M to $9.2M carrying cost
Inventory Cost Reduction
22% → 6.8% mean absolute percentage error
Forecast Error Rate
$8M → $1.4M (82.5% reduction)
Annual Markdown Losses
40 markets live within 9 months of kick-off
Rollout Timeline
Die Herausforderung
Vesture Global was managing 120+ SKUs across 40 international markets using spreadsheet-based forecasting. Their 22% forecast error rate was generating $8M in annual markdown losses and tying up $14M in excess inventory. Seasonal volatility and supplier lead times of 16+ weeks made any manual correction futile.
Die Lösung
We built a custom ensemble forecasting system combining XGBoost gradient boosting with LSTM neural networks, fed by real-time point-of-sale data from 2,300 retail locations. The pipeline included automated safety stock calculations, markdown optimisation triggers, and a supplier collaboration portal that surfaced demand signals 8 weeks earlier than the previous process.