Target improves on-shelf availability 150 bps for top 5,000 items with machine learning forecasting

Target operates thousands of stores across the U.S. with an assortment spanning essentials and discretionary categories — a combination that creates structurally different inventory risks under one roof. Inaccurate demand forecasting led to chronic overstock in discretionary goods and out-of-stocks in high-velocity essentials. Without real-time visibility into product flow from supplier to shelf, replenishment decisions lagged actual demand signals. The downstream effect was compounding: excess inventory in slow-moving categories required markdowns that pressured gross margin, while empty shelves in essentials eroded guest trust and comparable sales. For a retailer already navigating declining comparable sales and a stock underperforming peers like Walmart, inventory misalignment was a direct drag on operating income.

Target deployed machine learning models to overhaul demand forecasting and inventory placement across its supply chain network. The ML system monitors product flow continuously from supplier to shelf, replacing static replenishment cycles with dynamic, signal-driven restocking logic. Predictive analytics models analyze demand patterns at the SKU and store level, enabling the system to differentiate between discretionary and essential category behavior and adjust positioning accordingly. This capability was integrated into Target's broader AI-driven operating model — alongside GenAI merchandising tools and market-based fulfillment orchestration — so that forecasting signals inform not just replenishment but also fulfillment routing and in-store labor allocation. The rollout reflects an enterprise-wide infrastructure investment, with Target committing roughly $4 billion in CapEx to underpin technology, store, and fulfillment upgrades.

On-shelf availability for Target's top 5,000 items improved by more than 150 basis points year over year — a meaningful gain at the SKU level where velocity and margin concentration are highest. The improvement was reported in the context of a challenging quarter, making the inventory gain a rare operational bright spot against declining comparable sales.

• On-shelf availability: +150 bps YoY for top 5,000 items
• Markdown exposure: Reduced, reversing a significant gross margin headwind from prior periods
• P&L stability: Improved inventory positioning contributed to a more predictable cost structure across the store network

The results validated that ML-driven forecasting can deliver measurable outcomes even while broader financial performance remains under pressure.

• Prioritize ML forecasting at the top SKU tier first — the highest-velocity items carry disproportionate margin and availability impact, making gains faster to measure and more defensible to the P&L.
• Inventory accuracy compounds when forecasting is connected to fulfillment routing and labor planning, not siloed as a standalone merchandising tool.
• Markdown reduction is a measurable ROI signal for forecasting investments — track it explicitly alongside availability metrics to capture the full margin case.
• Differentiating replenishment logic by category type (essentials vs. discretionary) is essential at retail scale; a single model applied uniformly will underperform.