By Elie Dufeu, CTO & Co-Founder, Metreecs. Published 17 November 2026.
To reduce stockouts in retail, measure them by the demand they cost rather than by how many products are empty, find which cause is behind each one (forecast, ordering rules, inventory records, shelf execution, allocation, or supplier lead time), and apply the fix that matches that cause. Adding safety stock across the board treats every stockout the same way, and most of them don't share a cause.
A stockout looks like a single event on the shelf, yet it is the end of a chain of decisions made weeks earlier. A forecast missed a promotion, a reorder point was set two seasons ago, the system believed three units were in the backroom when there were none, or the last units went to the wrong store. Each of those calls for a different fix, and a retailer that applies one generic remedy usually ends up with more stock and roughly the same gaps.
This guide covers how to measure stockouts so the numbers point at money, the six causes that produce most of them, and the specific fix for each. It closes with a 30-day plan you can run with the data you already have. If you'd rather start from your own store and product data, you can review your stockout patterns with our team.
Key Takeaways
- About 47% of retail out-of-stocks are caused by inadequate store ordering and forecasting and 25% by poor shelf management, according to a study by FMI, GMA, and CIES that pooled 52 published reports.
- When a product is missing, 31% of shoppers go to another store, 26% switch to a different brand, and 9% buy nothing at all, per the same study.
- DeHoratius and Raman found that 65% of nearly 370,000 inventory records across 37 stores of one retailer were inaccurate, which means many replenishment systems reorder from numbers that are wrong.
- McKinsey reports that AI-driven forecasting can cut forecast errors by 20 to 50% and reduce lost sales and product unavailability by up to 65%.
What is a stockout, and how should you measure it?
A stockout (or out-of-stock) happens when a customer wants a product at a given location and time and it isn't available to buy: the shelf is empty, the size is missing, or the website shows "out of stock". It is measured per product and location, because a unit sitting in another store or in the warehouse doesn't make the sale.
How you count stockouts shapes which ones you fix. Most retailers start with a simple stockout rate: the share of products with zero stock on a given day. It is easy to compute, and it treats a slow product that sells twice a year the same as a bestseller that sells 40 units a week.
| Metric | How to calculate it | What it tells you | Limit |
|---|---|---|---|
| Stockout rate (count) | Products at zero stock / products in range | How widespread empty slots are | Ignores how much each product sells |
| Demand-weighted availability | Forecast demand of products in stock / total forecast demand | Share of expected sales you could serve | Needs a forecast per product and location |
| Estimated lost sales | Forecast daily demand x days out of stock | The money a stockout costs | Undercounts lost baskets and lost customers |
| Fill rate | Units shipped on time / units ordered | How well the warehouse or supplier serves stores | Says nothing about the shelf |
Here is a worked example with our own illustrative numbers. A store carries 400 products and 20 of them are at zero stock this week, so the count-based stockout rate is 5%. Those 20 products, though, include four bestsellers. Together the 20 empty products would have sold 540 units this week, out of 3,000 units of forecast demand for the whole range. The demand-weighted view says the store could serve only 82% of expected sales, an 18% availability gap hiding behind a "5%" headline.
Lost sales are the next step. If one of those bestsellers sells 12 units a day and stays empty for six days, the estimate is 72 units, before counting the shoppers who left without the rest of their basket. Ranking stockouts by estimated lost sales, rather than by count, is the first change that makes the work to reduce stockouts pay off, because it puts the effort where the revenue is.
The six causes behind most stockouts in retail
The FMI, GMA, and CIES study found that 70 to 75% of out-of-stocks are triggered at the retail level, with store ordering and forecasting the largest single cause. That finding still holds up in practice: a stockout is more often a planning or execution gap inside the retailer than a supplier failure. The table below maps each cause to the symptom you will see in the data and the first check to run.
| Cause | Typical symptom | First check |
|---|---|---|
| Forecast error | Repeated stockouts on the same products during promotions, weather shifts, or seasonal peaks | Compare forecast with actual sales for the weeks before each stockout |
| Ordering parameters | Reorder points and order quantities that no longer match current sales | Date of the last update to reorder points and safety stock per product |
| Inventory record errors | System shows stock, shelf is empty, no reorder triggered | Products with positive stock and zero sales for several days |
| Shelf execution | Stock is in the backroom while the shelf is empty | Gap between store stock and shelf capacity, replenishment frequency |
| Allocation | Same product empty in some stores and overstocked in others | Stock and sales by store for products in the stockout list |
| Supplier lead time | Stockouts cluster after late or partial deliveries | Actual versus planned lead time, and its variability, per supplier |
Most retailers have all six, in different proportions by category. A fashion chain will see more allocation stockouts on sizes, a beauty retailer more promotion-driven forecast misses, and a home goods chain more record errors on bulky items that are hard to count. The point of the diagnosis is to know which ones dominate in your own data before you spend money trying to reduce stockouts across the board.
One mistake we repeatedly see is a retailer answering every stockout complaint with a higher safety stock setting. It does reduce some stockouts, at a real cost in working capital, and it does nothing for the products that are empty because the record is wrong or the stock sits in the wrong store.
How to reduce stockouts caused by forecasting and ordering
Forecasting and ordering sit together because they fail together. A good forecast feeding stale reorder points still produces empty shelves, and fresh reorder points built on a weak forecast do the same.
Start with the forecast level. A forecast built per category or per region spreads the error across products and stores, so it looks accurate in aggregate while individual products in individual stores run out. Forecasting each product in each location, with promotions, price changes, holidays, and local events as explicit inputs, is what allows replenishment to anticipate the peak instead of reacting to it. McKinsey's work on AI-driven forecasting puts the potential at 20 to 50% fewer forecast errors, which feeds directly into availability.
Then refresh the ordering parameters. Reorder points and safety stock are often set once, when a product is listed, and revisited only when someone complains. Sales speed, lead time, and the service level you want all change over a season. Our guide to dynamic safety stock optimization explains how to recalculate buffers by product segment instead of applying one rule to the whole range.
Hana runs replenishment for a 45-store beauty retailer. Her team noticed that the same six skincare products ran out every time they featured in the monthly promotion, while their reorder points were built on average weekly sales. She added the promotion calendar to the forecast and raised the order for those products two weeks before each campaign. In the next promotion, four of the six stayed in stock throughout, and the two that didn't were limited by supplier allocation, which is a different conversation. (Hana is an illustrative example.)
Review frequency matters as well. A product reviewed once a week needs stock to cover a week of demand plus lead time, while the same product reviewed daily can run leaner and still stay available. Moving fast sellers to a shorter review cycle is often cheaper than raising their buffer.
How to reduce stockouts from inventory record errors and shelf gaps
Inventory record errors create the stockouts that systems can't see. When the record says three units and the shelf holds none, the reorder point never triggers, and the product can stay empty for weeks while every report shows it in stock. This is often called phantom inventory. DeHoratius and Raman's study of a large US retailer found 65% of the inventory records they examined to be inaccurate, and they linked the errors to store complexity and distribution structure, with regular auditing as the main mitigating factor.
The most useful detection rule is simple: flag any product that has positive stock in the system, a forecast of steady sales, and zero sales for longer than its normal gap between sales. A product that usually sells every day and has sold nothing for four days is probably not on the shelf, whatever the record says. Sending those products to a targeted count, rather than counting the whole store on a calendar, puts staff time where the errors are, and it is often the cheapest way to reduce stockouts that no forecast can see.
Priya leads inventory control for a 60-store home goods chain. Her stores ran full counts twice a year and her replenishment system trusted the records in between. She set up a daily list of products with stock on record and no sales for longer than expected, capped at 15 lines per store, and asked store teams to check them before opening. Within the first month the list surfaced damaged stock never written off, units stuck in the backroom, and a product scanned under the wrong variant at receipt. (Priya is an illustrative example.)
Shelf execution is the other half. The FMI, GMA, and CIES study attributed about 25% of out-of-stocks to poor shelf management: the stock is in the store but not where the customer looks. Matching delivery quantities to shelf capacity, so that cases don't sit in the backroom because they don't fit, and timing shelf fills around peak hours both reduce stockouts without adding a single unit of inventory.
In our experience deploying anomaly detection at Metreecs, store and planning teams often underestimate how much of their reported availability depends on record accuracy, because the error is invisible in every dashboard built on those records.
How to reduce stockouts caused by allocation and supplier delays
Allocation stockouts are frustrating because the company owns the stock, just not in the right place. A chain can hold enough units of a product in total while three stores are empty and five hold weeks of cover. This usually comes from an initial allocation based on store size or last year's sales rather than on each store's demand for that product, and from size curves applied evenly when sizes sell differently by region.
The fix starts upstream: allocate from a forecast per product and store, and hold back a share of the buy in the warehouse to replenish the stores that sell fastest. When stock is already in the wrong place, a transfer can make sense if the margin on the expected sale beats the transfer cost. Our article on why stockouts and overstock happen at the same time works through when a store-to-store transfer is worth it.
Marco plans footwear for a 30-store chain. Every season the same pattern returned: sizes 42 and 43 ran out in the city stores by week three, while the suburban stores still held them at the end of the season. He moved from one national size curve to curves by store cluster and kept 20% of each style in the warehouse for the first replenishment. The city stores still ran low on the core sizes, but later in the season and with stock already on its way. (Marco is an illustrative example.)
Supplier lead time is the last cause, and the one retailers control least. What you can control is the planning input: measure actual lead time per supplier, including its variability, and use that rather than the contractual number in your reorder points. A supplier that promises 21 days and delivers anywhere between 18 and 35 needs a larger buffer than one that delivers reliably in 28. When a customer can still order an item that's temporarily out, the way you handle it matters too, and our guide to what a backorder means for retailers covers how to manage and communicate it.
A 30-day plan to reduce stockouts in retail
You don't need a new system to start. This plan uses sales, stock, and order data that most retailers already have, and it shows within a month which causes dominate.
- Week 1: measure the right thing. Calculate demand-weighted availability and estimated lost sales per product and store for the last 13 weeks. Rank stockouts by lost sales, not by count.
- Week 1: pick the top 50. Take the 50 product and store combinations with the highest estimated lost sales. These are your test list.
- Week 2: tag the cause. For each one, use the cause table above to tag forecast, ordering parameters, record error, shelf execution, allocation, or supplier lead time. Some will have two causes.
- Week 2: start the phantom inventory list. Run the "stock on record, no sales" rule every day and send the results to stores for a check.
- Week 3: fix by cause. Refresh reorder points and safety stock for the products tagged with ordering issues, add promotions and events to the forecast for the forecast misses, and rebalance or transfer where allocation is the problem.
- Week 3: talk to suppliers with data. Share measured lead times and their variability with the suppliers behind recurring stockouts.
- Week 4: measure again. Recalculate availability and lost sales for the 50 combinations and compare with week 1. Extend the approach to the next 200.
Across Metreecs' work with multi-store retailers, steps 3 and 4 are usually where the first surprises come from, because tagging causes one by one shows how many stockouts were never a forecasting problem at all. Running this cycle every week across thousands of products and stores is where AI-driven inventory optimization earns its place, since it keeps forecasts, buffers, and anomaly checks current without a weekly spreadsheet rebuild.
FAQ
What causes stockouts in retail?
Most stockouts in retail come from inside the retailer rather than from suppliers. The FMI, GMA, and CIES study found that about 47% are caused by store ordering and forecasting and 25% by shelf management. Other common causes are inventory record errors, stock allocated to the wrong stores, and supplier lead times that are longer or less reliable than planned.
How do you calculate stockout rate?
The simplest stockout rate divides the number of products at zero stock by the number of products in the range. A more useful version weights each product by its forecast demand, so a missing bestseller counts for more than a missing slow seller. Track estimated lost sales alongside it (forecast daily demand multiplied by days out of stock) to see the cost.
Does more safety stock prevent stockouts?
More safety stock reduces stockouts caused by normal demand and lead time variation, at the cost of more inventory. It doesn't help when the record is wrong, the stock is in the backroom, or the units sit in another store. Set safety stock by product segment and service level, and fix the other causes separately.
What is phantom inventory?
Phantom inventory is stock that the system records but that isn't available to sell, because it was stolen, damaged, misplaced, or scanned under the wrong product. It causes stockouts that replenishment never corrects, since the system thinks the shelf is full. Flagging products with stock on record and no recent sales is a practical way to find it.
How can AI help reduce stockouts?
AI forecasting estimates demand for each product in each store using promotions, prices, seasonality, and local events, which reduces the forecast errors behind many stockouts. McKinsey reports up to 65% less lost sales and product unavailability from AI-driven forecasting. It works best combined with accurate stock records and refreshed ordering rules.
Conclusion
The fastest way to reduce stockouts in retail is to stop treating them as one problem: rank them by lost sales, tag the cause behind each one, and apply the fix for that cause, whether it's the forecast, the ordering rules, the stock record, the shelf, the allocation, or the supplier. To see what this diagnosis looks like on your own stores and products, book a demo built around your data.
Sources
- Food Marketing Institute, New industry study explores root causes of global out-of-stock dilemma (14 October 2002): study by FMI, GMA, and CIES (Gruen, Corsten, and Bharadwaj), 52 reports pooled; causes of out-of-stocks (47% store ordering and forecasting, 25% shelf management, 70 to 75% at retail level) and consumer responses (31% another store, 26% another brand, 9% no purchase).
- DeHoratius, N. and Raman, A., "Inventory Record Inaccuracy: An Empirical Analysis", Management Science 54(4), 2008: 65% of nearly 370,000 inventory records across 37 stores of one retailer inaccurate.
- McKinsey, AI-driven operations forecasting in data-light environments (February 2022): forecast errors reduced by 20 to 50%, lost sales and product unavailability reduced by up to 65%.








































































