Luca Barberis

How often are we losing sales to out of stock?

№ 059 · Stockout rate

01

Definition

MetricShare of SKUs or SKU-days unavailable to purchase

Unit% of listed SKUs out of stock

Percent of SKU-days where inventory hit zero. Spikes during peak. Build by counting (SKU times day) pairs with zero inventory and dividing by total SKU-day combinations. Lost-sales signal. Example: 4.2% baseline, 12% during Black Friday week (revenue gone).

SKU
Stock keeping unit. One distinct sellable variant. Counted at the variant level, not the product level.
02

Benchmarks

Bottom 30%MedianTop 30%
10%+, and roughly double the base rate on promoted items8% to 8.6%under 3% to 4% (on-shelf availability 96% to 97%+)

Retail and ecommerce blended, global. The 8.3% anchor is the GMA global study of 71,000+ consumers across 29 countries and has been re-confirmed at 8.6% in Europe by ECR Retail Loss. Definition used here: SKU-level unavailability as a share of listed SKUs. · The two most-cited anchors are old. The 8.3% global figure dates from the GMA study of 2002 and the promoted-items-are-twice-as-bad finding from ECR Europe in 2003, both carried forward in current literature because later reviews keep landing in the same range rather than because they have been re-run at that scale. Treat them as durable order-of-magnitude anchors in a 2026 reference, not as current measurements. Definition mismatch is the larger practical problem: event-based rates, time-based rates (share of selling time lost) and demand-based rates (unfulfilled demand) are different metrics and are routinely compared as if they were one. Online adds a specific distortion, since an item can show as unavailable when available-to-promise hits zero while units exist elsewhere in the network, and can show as available after the last unit is committed if sync lags.

By category
CategoryBottom 30%MedianTop 30%
Grocery and food12%+7% to 9.5%5%
General retail and ecommerce blended10%+8% to 8.6%3% to 4%
Promoted and discounted lines20%+10% to 16%6%
03

When it looks bad

The stockout line spikes precisely on promotion weeks and on the top-velocity SKUs rather than spreading evenly across the tail, so the business is going out of stock exactly on the demand it paid to create.

Blended stockout runs 6% but the promoted subset hits 17% in week two of each campaign, and the ten highest-velocity SKUs account for 40% of lost availability days. At a 4%-of-sales cost of stockouts, that concentration is worth more than the promotion's incremental margin.

04

What to do about it

  • Measure availability on your top-velocity decile separately from the blended rate, daily rather than periodically. A blended number that includes a long tail of low-velocity SKUs will not move when the SKUs that carry revenue go dark.
  • Fix inventory-record accuracy before buying forecasting software. Controlled tests show correcting inventory records alone has produced 4% to 8% sales uplift, because system stock and actual available stock diverge long before the forecast is the constraint.
  • Uplift safety stock on promoted lines specifically, since promoted items go out of stock at roughly twice the base rate. Promotions concentrate demand into a window narrower than the replenishment cycle, so the base service level is structurally wrong for them.
  • Instrument the substitution path rather than only the availability rate. Roughly 31% of shoppers buy the missing item elsewhere and about 9% abandon the purchase entirely, so a back-in-stock capture and a ranked substitute recommendation recover demand the availability fix cannot reach in time.
05

Sources

  1. Pygmalios (compiling GMA, ECR Retail Loss, IHL, Corsten and Gruen)State of Shelf Availability and Store Foot Traffic 2026 · 2026 · Compilation; underlying GMA study covers 71,000+ consumers across 29 countries Global average out-of-stock rate 8.3% (GMA, 2002) with ECR Retail Loss reporting about 8.6% across Europe; world-class performers run on-shelf availability of 96% to 97%+, meaning out-of-stock at or below 3% to 4% pygmalios.com ↗
  2. Pygmalios (compiling ECR Europe and Corsten and Gruen)State of Shelf Availability and Store Foot Traffic (report PDF) · 2026 · Compilation of ECR Europe 2003 Optimal Shelf Availability study and Corsten and Gruen HBR 2004 Promoted items go out of stock at roughly twice the rate of non-promoted items; stock-outs cost about 4% of sales, roughly $40M a year for a billion-dollar retailer; when an item is missing about 31% of shoppers buy it elsewhere and about 9% do not buy at all pygmalios.com ↗
  3. Xorosoft (citing IHL Group and Purdue University)Stockout Statistics Every Retailer Should Know · 2026 · IHL Group 2026 research; Purdue survey of 1,200 US consumers IHL puts global inventory distortion at $1.7 trillion, 6.2% of global retail sales, with 65.6% attributed to out-of-stocks; US food out-of-stock rate 9.5% in 2024, down from 12.3% in 2023 and 19.3% in 2022 xorosoft.com ↗
  4. Opensend29 Inventory Stock-out Rate Statistics for eCommerce Stores · 2025 · Ecommerce benchmark compilation Typical ecommerce out-of-stock rate 2% to 5% under normal conditions with about 8% used as a practical blended benchmark, rising to 10% for promoted or discounted items opensend.com ↗
  5. Opensend29 Product Availability Statistics for eCommerce Stores · 2025 · Ecommerce benchmark compilation Industry leaders maintain rates below 5%; 60% of online brands face stockout situations weekly; stockouts are implicated in roughly 20% of cart abandonments opensend.com ↗
  6. PazoOn-Shelf Availability in Retail: Definition, Formula, Benchmarks · 2026 · Global FMCG and retail benchmark reference Strong on-shelf availability falls between 95% and 98%, with measurable revenue leakage beginning below 95%, and OSA must be measured daily or weekly rather than periodically gopazo.com ↗
  7. Pygmalios (citing ECR Retail Loss and IHL/Scandit)State of Shelf Availability and Store Foot Traffic 2026 · 2026 · ECR Retail Loss 2024 retailer survey; IHL and Scandit 2025 shelf intelligence study Correcting inventory-record accuracy alone has produced 4% to 8% sales uplifts in controlled tests, and only 7% of surveyed retailers use any form of video analytics for shelf monitoring pygmalios.com ↗