Size Curve Demand Forecasting: Fix Phantom Stockouts

Most retailers know what a stockout looks like. An item sells through, inventory hits zero, and customers move on to something else. Phantom stockouts are harder to spot because inventory is still sitting in the system.

A sweatshirt may show 120 units available, but if Medium and Large disappeared two weeks ago, those remaining units in XS and XXL are doing very little for today's customer. The ERP says the style is in stock. The shopper sees the opposite. That's the gap.

This happens every season across apparel, footwear, and any category where size matters. The issue is not always how much inventory was bought. More often, it's whether the right sizes were bought in the first place. Once the size curve breaks, planners start reading distorted signals. Sales appear to slow, weeks of supply look healthy, and replenishment decisions become less reliable because demand is no longer visible.

The result is a costly cycle. Core sizes sell out too early, fringe sizes linger until markdown, and future buys are based on incomplete demand history. Retailers that plan at the size level instead of treating every style as a single inventory bucket are far better positioned to protect full-price sales and avoid repeating the same mistakes.

Why Broken Size Curves Create "Phantom Stockouts" Instead of Healthy Inventory

A size curve is simply the expected distribution of demand across the sizes within a style. It answers a practical question before inventory is ever purchased: how many Smalls, Mediums, Larges, and other sizes should be available to match customer demand?

That sounds straightforward. In practice, it rarely stays balanced for long.

Unlike many retail categories, apparel inventory isn't interchangeable. Ten units of XXL cannot replace ten units of Medium, even though both belong to the same style. Once a customer's preferred size is gone, the sale often disappears with it. Research has consistently shown that customers have limited willingness to substitute into another size when faced with a size-specific stockout.

This is why style-level inventory reports can create a false sense of security. A planner might see 60 percent of the original inventory still available and conclude the style has weeks of selling left. The customer sees a rack missing every popular size.

Imagine a women's denim style entering the season with a balanced size run. After several weeks, sizes 6, 8, and 10 sell through quickly, while sizes 2 and 16 remain plentiful. The inventory report still shows healthy availability, but conversion drops because the majority of shoppers cannot find their size.

From an operational perspective, inventory exists.

From the customer's perspective, the product is effectively out of stock.

That difference matters because it changes how retailers interpret demand. Sales decline, not because customers suddenly lost interest, but because the inventory no longer matches what shoppers came to buy. Unless planners examine inventory at the size-SKU level, those lost purchase opportunities remain invisible.

The longer the broken size curve persists, the more misleading traditional performance metrics become. Style availability looks acceptable while customer availability steadily deteriorates. That's the essence of a phantom stockout.

The Financial Cost of Broken Size Curves: Lost Sales, Hidden Demand, and Margin Erosion

Broken size curves hurt profitability from both directions.

The first hit comes when core sizes disappear too early. Those are often the sizes customers are willing to buy at full price. Every missed purchase represents revenue that was available but never captured.

The second hit comes months later.

The remaining inventory typically consists of slower-moving sizes that require deeper markdowns to clear. Retailers end up sacrificing margin on inventory that never matched customer demand in the first place.

This creates an expensive contradiction. One style can experience stockouts and overstock simultaneously.

Consider a footwear launch. Demand concentrates in men's sizes 9, 10, and 11, but the initial buy overestimates demand for sizes 7 and 13. Within weeks, the core sizes disappear. Sales begin flattening, not because interest fades, but because customers stop finding their fit. By season's end, the remaining fringe sizes are marked down to free working capital.

Nothing about that outcome is efficient.

Inventory carrying costs continue accumulating while slow-moving units occupy valuable warehouse and store space. GMROI falls because inventory dollars remain tied up in products that no longer generate meaningful sales. Even if the retailer eventually reaches very high sell-through, much of that volume comes after discounting has eroded profitability.

This is why headline metrics can be deceptive. A style that achieves nearly complete sell-through isn't necessarily a successful buy if a significant portion required markdowns while earlier full-price demand went unmet.

Retailers often focus on total inventory investment without asking whether that investment was distributed correctly across sizes.

That's the real issue.

Margin isn't lost because inventory existed. It's lost because the inventory customers wanted disappeared first, while the inventory they didn't want stayed behind.

Why Traditional Forecasting Keeps Repeating the Same Size-Curve Mistakes

Historical Sales Are Not the Same as True Demand

Forecasts usually begin with historical sales. That's reasonable until inventory availability starts influencing those sales.

Once a popular size sells out, the sales history becomes incomplete. The system records zero additional sales for that size, even though customers may have continued looking for it every day afterward.

This is known as censored demand. Actual demand is hidden because inventory was unavailable.

If forecasting models treat those zero sales as declining interest, they underestimate future demand for the very sizes that were already underbought. The mistake then carries into the next buying cycle.

Many planning teams have experienced this without calling it censored demand. They review last season's reports, see lower unit sales in Medium after week six, and reduce future buys accordingly. The reality is Medium stopped selling because Medium stopped existing.

Sales data only tells the full story while inventory is available.

Static Size Curves Can't Keep Up with Modern Retail

Even accurate size curves don't stay accurate forever.

Customer demand changes across regions, store formats, digital channels, and product categories. A suburban location may consistently sell larger sizes than an urban flagship. E-commerce often behaves differently than physical stores because assortment is broader and inventory pools are shared.

Returns also influence future buying decisions. If one particular fit generates unusually high return rates in certain sizes, historical sales alone don't explain what actually happened.

Promotions complicate things further. A clearance event may temporarily distort size demand, while new product fits can shift purchasing patterns compared to previous collections.

Yet many retailers still allocate inventory using static size curves that change very little from season to season.

That approach reinforces forecasting errors.

The same size shortages repeat. The same excess inventory accumulates. The organization believes demand is stable because the planning process keeps feeding historical limitations back into future forecasts.

Improving forecasts isn't only about better algorithms. It's about separating genuine customer demand from inventory constraints before those constraints become tomorrow's assumptions.

Building Dynamic Size Curves That Protect Full-Price Sell-Through

Better size planning starts with cleaner data.

The most reliable demand signals usually come from periods when inventory was fully available and products sold at regular price. Once stockouts or heavy markdowns enter the picture, demand becomes distorted.

That means planners need to correct for stockout bias instead of accepting historical sales at face value.

Forecasting should also move beyond the style level. Demand exists at the size level, so planning should reflect that reality. Instead of deciding to buy 2,000 units of a jacket and applying a standard allocation, retailers can forecast expected demand for each size using recent selling behavior, regional differences, and channel performance.

Store-specific size curves are equally important. A single national allocation rarely reflects how customers actually shop across locations.

Returns deserve a seat at the planning table as well. If certain sizes consistently come back because of fit issues, that information should influence future buying decisions instead of being treated as a separate operational problem.

This is where machine learning has become genuinely useful.

Rather than replacing planners, modern demand planning systems continuously evaluate fresh sales, inventory positions, returns, and changing demand patterns to recommend adjustments throughout the season. The planner still applies judgment. The system simply surfaces patterns that are almost impossible to detect manually across thousands of size-SKUs.

For retailers managing large assortments, this also reduces reliance on spreadsheet-based analysis that quickly becomes unmanageable as products, stores, and channels multiply. Platforms such as Flagship support planners by identifying emerging size imbalances early, allowing allocation and replenishment decisions to happen before phantom stockouts become visible to customers.

The objective isn't perfect forecasting. Retail doesn't work that way.

The objective is continuously improving how inventory matches real customer demand while there's still time to act.

The Retail KPI That Matters: Measuring Inventory Availability the Way Customers Experience It

Most inventory KPIs were designed to measure operational efficiency.

Weeks of supply, style-level sell-through, total units on hand, and inventory value all have their place. None of them answer the customer's question.

Can I buy the size I need today?

Retailers should monitor complete size-run availability alongside traditional inventory metrics. A style with 70 percent of its inventory remaining but missing every core size is not healthy inventory, regardless of what the dashboard suggests.

Tracking stockouts at the size-SKU level also reveals recurring planning issues before they become expensive. If the same sizes consistently disappear first across similar product categories, that's a forecasting problem, not bad luck.

Forecast accuracy should be judged by how well retailers fulfill customer demand, not simply by how closely purchased units matched aggregate sales.

Protecting core sizes has a ripple effect across the business. Full-price sell-through improves. Markdown dependency falls. Inventory productivity increases because capital is invested in the inventory customers actually purchase. Customer satisfaction improves because shoppers find complete size runs more often instead of encountering partially stocked assortments.

The retailers making the biggest gains aren't necessarily buying more inventory. They're allocating it more intelligently.

As demand planning becomes more dynamic, the focus shifts away from style-level inventory counts toward customer-level availability. That shift helps retailers reduce phantom stockouts, preserve margin, and generate stronger returns from every inventory dollar they commit.