ABC-XYZ Inventory Optimization

Overview

Built a Python-based ABC-XYZ inventory classification framework for Cooper's B2B distribution business. By segmenting SKUs simultaneously on revenue value (ABC) and demand variability (XYZ), the framework prescribes differentiated stocking and replenishment policies across a 9-cell matrix — replacing a one-size-fits-all inventory approach and directly reducing stockouts by 8% and holding costs by 5%.

Methods

• Data ingestion: merged 12 months of invoice-level sales data across branches, products, and customers
• ABC classification: ranked SKUs by cumulative revenue contribution — A (top 80%), B (next 15%), C (bottom 5%)
• XYZ classification: computed coefficient of variation (CV) of monthly demand per SKU — X (CV<0.5, stable), Y (0.5–1.0, variable), Z (>1.0, irregular)
• 9-cell matrix construction: cross-tabulated ABC and XYZ classes to assign each SKU an inventory policy
• Policy prescriptions: AX/BX → continuous replenishment, AY/BY → safety stock buffers, AZ/BZ → make-to-order, CX/CY/CZ → EOQ or periodic review
• Branch-level analysis: applied framework across multiple distribution branches with localized demand patterns

Key Findings

• A-class SKUs (top 80% revenue) represent a small fraction of total SKUs — classic Pareto distribution confirmed
• Z-class items (high demand irregularity) were disproportionately tied to stockout events despite low revenue contribution
• AX SKUs benefit most from continuous replenishment with tight reorder points; CZ SKUs should be deprioritized or made-to-order
• Branch-level demand patterns differ significantly — a single global policy misallocates safety stock across locations