Days of Supply: The Forward-Looking Inventory KPI Most Retailers Skip

On-hand units is a backward-looking number

Most retail inventory scorecards center on units on hand. The store has 240 of SKU X. The DC has 4,800. The region has 18,000. These numbers are precise, easy to report, and almost useless for operating decisions.

The reason: units on hand is meaningful only in relation to demand. 240 units of a SKU selling 8 per week is 30 weeks of supply — a severe overstock. 240 units of a SKU selling 80 per week is 3 weeks of supply — likely an understock given typical lead times. Same on-hand number, opposite operating implications.

Days of supply (DOS) is the demand-weighted version: on-hand divided by expected daily demand. It converts a unit count into a time horizon, which is what operators actually need to know. "We have 4 days of supply" is actionable. "We have 240 units" requires three more queries to interpret.

The retailers operating on DOS as the primary inventory KPI make better replenishment, allocation, and markdown decisions because the metric integrates demand into the conversation from the start.

DOS variance across stores is the real signal

Chain-level DOS is an average that hides everything important. A chain at 42 DOS on a category averages stores at 28 DOS (running tight, risk of stockout) and stores at 65 DOS (overstocked, working capital trapped, markdown risk). Same category, same chain, two opposite problems happening simultaneously.

The variance distribution is where the operating decisions live. A typical mid-market retailer carries 12-18 categories where store-level DOS variance exceeds 50% of the chain average. Those are the categories where central allocation is producing structural mis-matches between supply and demand at the store level.

For each high-variance category, the operating opportunity is the same: rebalance inventory from high-DOS stores to low-DOS stores. The total chain inventory stays the same. Working capital doesn't change. But the cells most at risk of stockouts get supplied, and the cells most at risk of markdowns get relieved. The net result is higher fill rate, lower markdown exposure, and better turn at the chain level — all without adding inventory.

The rebalancing math

Consider a category at 42 DOS chain average, with $4.8M of inventory across 80 stores. The variance distribution: 20 stores at 28 DOS, 40 stores at 42 DOS, 20 stores at 65 DOS.

The 20 stores at 65 DOS hold roughly 30% of category inventory ($1.4M) against 19% of category demand. The 20 stores at 28 DOS hold 13% of inventory ($620K) against 21% of demand. The imbalance is structural and runs continuously because central allocation rules don't adjust quickly enough to local demand patterns.

Rebalancing 8 DOS worth of supply from the high-DOS cluster to the low-DOS cluster moves approximately $290K of inventory geographically. No new inventory. No reduced inventory. Just relocated inventory. The fill rate impact on the low-DOS cluster is a 12-18% reduction in stockout incidents. The markdown impact on the high-DOS cluster is a 22-30% reduction in terminal clearance volume.

For one category, the annual value is $200-400K in recovered margin. Multiply across 12-18 high-variance categories: $2.5-7M annually. The chain didn't grow inventory. They moved it to where it should have been.

DOS as the leading indicator for markdowns

DOS at the SKU-store level is the strongest single predictor of future markdown exposure. The relationship is mechanical: if a SKU at a store has 80 DOS in week 4 of a 12-week season, that SKU will not sell through at full price by end of season. The math is irrefutable. 80 DOS × residual season weeks of demand exceeds available season weeks. Markdown is mathematically required.

Most retailers don't run this forward-looking check until weeks 6-8 of the season, by which point intervention requires deeper markdowns. The retailers that run weekly DOS-against-residual-season checks identify the markdown candidates in week 3-4, when smaller interventions still recover meaningful margin.

The classification logic is straightforward. For each SKU-store cell, compute current DOS and compare against weeks remaining in the season:

• DOS < weeks remaining × 0.7: healthy, may need replenishment
• DOS within ±30% of weeks remaining: on track, monitor
• DOS > weeks remaining × 1.4: structural overstock, markdown likely needed
• DOS > weeks remaining × 2.0: severe overstock, accelerate intervention

The categorization changes weekly as DOS evolves with sell-through. A SKU-store cell drifting from "monitor" to "markdown likely" triggers an alert. The merchandising team gets early signal in weeks 3-5 rather than reactive signal in weeks 8-10.

Forward DOS vs. trailing DOS

Most retailers calculate DOS using trailing demand: on-hand divided by average daily demand over the last 30 days. The trailing number is acceptable for stable categories. For categories with momentum, seasonal shifts, or new launches, it's actively misleading.

Consider a category trending up 15% week-over-week through early season. Trailing 30-day average underweights recent demand and inflates the calculated DOS. A SKU showing 42 DOS on trailing demand might be at 28 DOS on forward-projected demand. The operator sees "healthy supply" when the actual position is "approaching tight."

Forward DOS uses projected demand: applying the trend or forecast forward rather than averaging the past. The number is more useful for replenishment and allocation decisions because it reflects what's likely to happen, not what just happened.

The retailers using forward DOS catch trending categories earlier — both up-trends (replenishment increases ahead of stockouts) and down-trends (allocation reductions ahead of overstock). The accuracy of forward DOS depends on the forecasting model underneath, which is why tier-decomposed WMAPE matters: forward DOS is only as good as the demand forecast that feeds it.

Monitoring DOS continuously

DOS monitoring at the SKU-store-category cell weekly is computationally trivial but data-engineering-heavy at scale. A chain with 5,000 SKUs and 80 stores has 400,000 SKU-store cells. Daily on-hand plus daily demand plus forward projection means 400K records updated multiple times per week.

This is exactly the kind of monitoring that lives or dies on infrastructure. The retailers running it well have automated pipelines that compute DOS at the cell level, classify against thresholds, and surface variance alerts to the right operators. The retailers running it poorly have a quarterly DOS report that takes a data analyst 2 weeks to produce and is stale by the time it lands.

The financial impact of continuous DOS monitoring at a mid-market retailer is consistent across the operations we've worked with: 1.5-2.5% improvement in turn, 80-150 bps reduction in markdown rate, 1.0-1.8% improvement in fill rate. Combined annual value runs $3-7M for a $400M chain.

None of that improvement requires more inventory, more headcount, or new vendors. It requires seeing DOS variance at cell level and acting on it before the markdown season makes the variance permanent.

Key takeaways

• Units on hand is a backward-looking number with no operational meaning without demand context. Days of supply integrates demand into the metric and becomes immediately actionable.
• Chain-level DOS averages hide store-level variance. Most mid-market retailers have 12-18 categories where store-level DOS variance exceeds 50% of the chain average — exactly the categories where rebalancing creates value without adding inventory.
• DOS rebalancing across high-variance categories typically recovers $2.5-7M annually for a $400M chain by reducing stockouts and markdown exposure simultaneously, without changing total inventory.
• DOS-against-residual-season is the strongest leading indicator for markdown exposure. The math identifies overstock cells in weeks 3-4 of a season, when smaller interventions still preserve margin.
• Forward DOS (projected demand) outperforms trailing DOS (historical demand) for trending categories. Trailing averages misclassify accelerating categories as "healthy" when they're approaching tight.
• Continuous DOS monitoring at SKU-store-week cadence is the operational baseline. Quarterly DOS reports are stale on arrival and produce no operational value.
• Annual impact of continuous DOS monitoring at a mid-market retailer: 1.5-2.5% turn improvement, 80-150 bps markdown reduction, 1.0-1.8% fill rate improvement. Combined $3-7M annual value for a $400M chain.