Forecasting Demand by Size More Accurately

Total Demand Is the Easy Part

Most fashion brands forecast how many units of a style they'll sell reasonably well. Where they fall down is forecasting which sizes. A style that sells through overall can still be a disaster at the size level — stocked out on Medium by week two, drowning in XXL at markdown by end of season. The aggregate forecast hides this, because the totals can look fine even when the size distribution is badly wrong. Getting the size curve right is where real inventory efficiency lives, and it's where most brands are flying blind. This guide covers how to forecast demand by size more accurately, why it's harder than aggregate forecasting, and how better fit data sharpens the picture. It builds on our work on how fit data improves inventory planning.

Why Aggregate Forecasts Fail at the Size Level

A total-units forecast averages away the thing that matters: the shape of demand across sizes. Two styles with identical total sales can have completely different size curves, and applying a generic distribution to both guarantees you'll be wrong on at least one. The default many brands use — a standard bell curve across the size range — rarely matches any specific style or customer base.

The cost shows up as simultaneous stockouts and markdowns: lost sales on the sizes you under-bought and margin erosion on the sizes you over-bought, often within the same style. Both are expensive, and the aggregate forecast made them invisible until it was too late. Forecasting at the size level is what surfaces and prevents this, and it requires treating the size curve as a first-class output, not an afterthought applied to a total.

Your Historical Sales Curve Is the Starting Point

The best starting input is your own historical sell-through by size, for comparable styles. How did similar products actually distribute across sizes — not how you bought them, but how they sold and, crucially, how they sold at full price versus markdown? Buying data is contaminated by past forecasting errors; sell-through at full price is closer to true demand.

Segment this by category, fit type, and customer group, because a relaxed-fit style and a fitted style sell different curves even to the same customers. The richer your historical size-level data, the better your forecast. Most brands have this data but never analyze it at the size level, defaulting to a generic curve instead. Mining your own history is the highest-ROI first step, and it connects to building a better size curve.

Returns Distort the Picture — Correct for Them

Raw sales by size are misleading if you don't account for returns, because a size with high gross sales but high returns isn't really in the demand you think it is. A size that sells well and comes back constantly is a fit problem, not strong demand, and forecasting more of it repeats the mistake. Net sales by size — gross minus returns — is the truer demand signal.

This is where fit data and returns data feed forecasting directly. A size that returns heavily for fit reasons tells you the grading is off, not that customers want more of it. Correcting your size curve for returns prevents you from over-buying problem sizes. The connection between returns and planning runs throughout turning returns data into better-fitting products.

Customer Body Data Sharpens the Curve

The most direct input to size demand is your customers' actual bodies. If you know the distribution of body measurements across your customer base — from a fit tool, a sizing widget, or order history — you can forecast the size curve from the demand side rather than inferring it backward from contaminated sales data. This is a fundamentally better signal.

A brand whose customers skew toward certain sizes should buy that curve, not a generic bell. Most brands never collect this data and so never use it, but those that do can forecast size demand with far more confidence. Customer body data turns size forecasting from guesswork into something grounded in who your customers actually are, which is exactly what a fit tool can surface — see how fit data improves inventory planning.

Account for New Styles With No History

The hardest forecasting case is a brand-new style with no sales history, especially in a new silhouette or fit. Here you lean on analogs — the size curves of the most similar past styles — adjusted for known differences in fit. A new fitted dress should borrow from past fitted dresses, not from your loose tops.

Be explicit about the fit characteristics of the new style and match it to the right historical analogs rather than applying a house-default curve. The more your historical data is tagged by fit type and silhouette, the better your analogs, which is why the tagging discipline pays off. New styles will always carry more forecast risk, but disciplined analog selection narrows it considerably, and post-launch you can correct fast if you're watching size-level sell-through.

Plan for In-Season Correction

No size forecast is perfect, so the second half of the discipline is reacting fast when reality diverges. Watch size-level sell-through closely in the early weeks, and have a plan to reorder hot sizes and slow markdowns on cold ones before the problem compounds. A brand that spots a Medium running out by week two and reorders saves the sales a static plan would have lost.

This requires size-level reporting in near real time, not a season-end post-mortem. The brands that win on inventory aren't the ones with perfect forecasts — they're the ones that detect and correct fast. Building that monitoring is as important as the forecast itself, and it ties into the analytics discipline behind reducing markdowns with better fit data.

Tie Size Forecasting to the Bigger Picture

Size-level forecasting doesn't live in isolation — it connects to grading, assortment, and ultimately margin. A brand that forecasts sizes well buys closer to true demand, stocks out less, marks down less, and frees up working capital tied in dead inventory. Those are real financial outcomes, not operational nicety.

Make size-level performance a tracked metric and a recurring conversation, not a buried detail in a seasonal report. When the organization treats the size curve as something that drives margin, it invests in the data to get it right. This is part of treating fit and sizing as a strategic discipline, covered in making fit a company-wide priority. Better size forecasting is one of the clearest paths from fit data to bottom-line impact.

Where YourSizer Fits

YourSizer captures the actual body data of your shoppers and the fit signals behind your returns — exactly the demand-side inputs that sharpen a size curve. Instead of forecasting sizes from contaminated buying history or a generic bell curve, you can ground your plan in who your customers really are and which sizes truly perform after returns. See the underlying technology in our product overview, explore the merchant side on for brands, and check pricing.

Frequently Asked Questions

Why isn't a total-units forecast enough?

Because it averages away the size distribution, which is where the money is. Two styles with identical total sales can have completely different size curves, and a generic distribution will be wrong on at least one — producing stockouts on the sizes customers want and markdowns on the ones they don't, often within the same style. The aggregate number hides both.

Should I forecast from gross sales or net of returns?

Net of returns. A size with high gross sales but high returns isn't real demand — it's usually a fit or grading problem, and forecasting more of it repeats the mistake. Net sales by size gives a truer demand signal, which is why returns data and fit data feed directly into accurate size-level forecasting.

How do I forecast sizes for a brand-new style?

Use analogs — the size curves of your most similar past styles, matched by fit type and silhouette rather than a house-default bell curve. The better your historical data is tagged by fit characteristics, the better the analog. Then watch size-level sell-through closely after launch so you can reorder hot sizes and correct fast.