If you’ve ever been frustrated by how size curves misrepresent actual demand, you’re not alone.
Most retail planning systems calculate size curves directly from sales data. That sounds logical, until you realize how often sales don’t reflect intent. If a certain size is consistently out of stock, it’s not just underrepresented in your data — it’s invisible. And that invisibility leads to missed demand, misallocated inventory, and recurring stockouts, hindering effective retail inventory optimization.
Imagine a T-shirt that appears to sell 20% small, 60% medium, and 20% large. But large has been sold out half the time. That distribution isn’t demand — it’s a data artifact. The real demand for large may be closer to 40%. Yet most planning systems take the constrained curve at face value and use it for future decisions. The cycle repeats, season after season.
Over time, these distortions compound. Each new assortment, season, and store opening builds on incomplete information. Planners and allocators then spend valuable time manually correcting avoidable errors because the underlying size curves were never accurate to begin with. The key to ending this frustration lies in accurate demand signals — and the ability to apply them directly to size curve modeling. And with forecasting and data integration advances, retail's reliance on inaccurate size guidance can finally end.
Why traditional size curves fall short
There are several structural issues that make traditional size curves unreliable.
1. They inherit the limitations of sales data
If a size sells out early, the system interprets that as low demand rather than a failure to meet customer expectations. Out-of-stock events bias the distribution before planners even begin analysis.
2. They assume different products behave the same
Different articles that are the same size do not perform uniformly. One may peak in medium, while another skews toward large. Treating all items alike ignores unique and important article differences and customer preference.
3. They create circular logic
Bad data makes bad data. When constrained sales are used to generate size curves and those same curves drive future allocations, past inaccuracies will echo throughout your future planning cycles. The next distribution might look more like the last one.
As a result, planners and allocators are often left working with noisy, unstable size ratios, especially for low-volume SKUs. These bad signals demand more manual intervention and slow decision-making.
A smarter approach to size curves
Retailers need size curves that better reflect real customer demand. This requires moving beyond raw sales data and incorporating a broader view of product behavior, historical patterns, and inventory availability.
A more modern approach uses machine learning to analyze similarities across items. Instead of learning from sales alone, the model embeds both visual and textual attributes of a product: things like category, color, silhouette, and description. These signals help the system learn how comparable articles typically distribute across sizes.
Given a new article, the model looks for similar items in this embedding space and learns from their historical size distributions. Crucially, it only considers available sizes, so if the product is offered in medium, large, and extra-large, it predicts a curve within that scope. There’s no hallucinating demand for sizes that aren’t in the assortment.
Because the model learns from similar articles, it doesn’t need extensive sales history for the specific SKU. And when there is little to no data — such as in a new store — it falls back to more general curves built at the category or segment level. This means you still get an informed size distribution even in cold-start scenarios.
This approach offers several advantages:
- It corrects for stockout bias so high-demand sizes remain visible, even if they were not consistently available.
- It generalizes across the catalog to provide more reliable curves for low-volume or new styles.
- It adapts to the sizes offered for a specific article and does not predict demand for unavailable sizes.
- It supports new stores or new items by drawing from category or segment-level patterns.
This results in a more resilient, flexible system that captures real buying behavior without being restricted by the limitations of historical sales.
Real-world scenarios retailers face every day
Better size curves are most valuable when they support real merchandising decisions. Here are a few examples of how more accurate size distributions can change outcomes in the field.
Scenario 1: Increasing allocation for a high-performing store
A popular store is trending above plan, and the planner increases total units for a top style. With a smarter size curve, the distribution automatically scales in proportion to expected demand for each size. The store receives more depth in the sizes that matter without any manual recalculation.
Scenario 2: Reducing inventory for a lower-performing location
Another store is underperforming, so the total units for the same style are reduced. Intelligent size curves concentrate the remaining units in the core selling sizes and trim depth in fringe sizes. This helps maintain productivity even when volume drops.
Scenario 3: Allocating to a new store with limited or no history
A new location is opening with no size-level selling history. Instead of guessing, teams can use broader category or segment-level patterns to inform the curve. The result is a thoughtful, data-driven distribution that reflects how similar stores typically behave.
In each case, allocators spend less time correcting size-level allocations and more time making strategic decisions that increase inventory productivity.
Why this matters — and what’s next
When size curves reflect true demand, retailers gain a clearer picture of how customers actually shop. Accurate size distributions improve forecasting, reduce stockouts in core sizes, limit excess in less popular sizes, and strengthen overall margin performance. They also empower planning teams to move faster and spend less time correcting manual errors.
Traditional size curves often fall short because they rely too heavily on incomplete or biased data. Modern retail requires a more adaptive approach that accounts for product differences, inventory constraints, store behavior, and the realities of fast-moving demand. By strengthening these foundations, retailers create a path toward more accurate allocations, more responsive in-season adjustments, and a more resilient inventory strategy overall.
