How to build a size and colour curve from sell-out data in a spreadsheet

KEY TAKEAWAYS Summary by the editors

1. A size curve is each size's share of total demand, derived from past sales and applied to a new buy or allocation.
2. Stockouts censor the data: a size that sold out looks less popular than it was, so the curve should use the sales rate on days the size was in stock.
3. Curves differ by fit, category, location and channel, so a single company-wide curve will misallocate sizes.
4. For new products without history, planners borrow the curve of comparable items, and accuracy depends on how well the attributes match.
5. Fixed pre-packs can clash with actual demand and leave orphan sizes, so compare pack ratios with the curve before committing.

To build a size curve in a spreadsheet, collect units sold and in-stock days per size, clean out distorted periods, correct each size for stockouts, then divide each size's adjusted demand by the total. The result is the percentage mix to apply to your buy. The same logic works for colour. The method is simple, but the quality of the curve depends on the data you feed it.

What data do you need?

Inventory Planner's method starts with size-level data: units sold, inventory levels, sell-through and returns for each size. Add the style or fit group, the channel or location, and the dates on which each size was in stock. Sell-out data (what retailers sold to consumers) is ideal for a brand planning sell-in, but it is often incomplete, so state clearly which accounts are covered.

How do you calculate the curve step by step?

1. Collect size-level sales and stock history for comparable styles.
2. Clean the data: remove or adjust out-of-stock periods, unusual promotions, inconsistent size labels and discontinued products.
3. Calculate each size's raw share: units sold in that size divided by total units sold across the size run.
4. Correct for stockouts: for each size, take the average daily sales rate on days it was in stock and multiply by the number of selling days in the season.
5. Normalise: divide each size's adjusted demand by the total adjusted demand and express it as a percentage. This is the curve.
6. Apply it: multiply the total forecast by each size's percentage to get unit targets.
7. Review at season end by comparing predicted size mix with actual sales and stockouts.

In this invented example, size L sold 180 units but was out of stock for 30 of 90 days, so its adjusted demand is higher than its raw sales. A naive curve would have given L only 20% of the buy, repeating the shortage.

Read also

What is demand sensing and how does it work in fashion?

What are the main pitfalls?

• Stockout censoring understates demand for popular sizes if not corrected.
• One curve for everything: fits, categories, locations and channels can differ.
• Thin history on new products: planners often rely on proxy items.
• Rigid pre-packs: fixed supplier ratios can leave sizes short and others long.
• Slow replenishment: an accurate curve loses value if purchase orders cannot be adjusted.

How do you handle new products and colours?

Blue Yonder describes borrowing the curve of like items with similar attributes such as fabric or fit, noting that a poor match yields a poor curve. The same approach works for colour mix, though colour demand is more volatile and trend-driven, so many teams apply curves for size and use judgement and test orders for colour. Keep the attributes used for matching consistent, and record which proxy was used so the result can be reviewed. Note that Blue Yonder is a vendor of size scaling software, so its benefit claims should be treated cautiously.

When is an algorithm needed instead of a spreadsheet?

A spreadsheet suffices for a limited number of styles and locations. It becomes cumbersome when curves must be segmented by store cluster, refreshed weekly or combined with allocation. At that point software helps, but the principles above still apply and the spreadsheet remains a useful check of any automated output.

What are the limits of sell-out data?

Sell-out data records what consumers bought, which is the right signal for size demand, but it has gaps. Not every retailer shares it, store coverage may be uneven, and the sample may favour larger accounts. A curve built from a few accounts may not describe the rest. Where sell-out is unavailable, sell-in history can serve as a rough proxy, with the understanding that it reflects what buyers chose to buy, not what consumers wanted.

Uphance notes that AI forecasting needs consistent style attributes in one system and accurate sales and return data, with returns posted to inventory within days. For size curves, returns data is useful because returns often cluster by size and fit. A size that sells strongly but is returned often should be interpreted carefully.

• State which accounts and periods the data covers.
• Treat a heavily returned size with caution.
• Keep size labels consistent across suppliers and seasons.
• Re-run the curve each season instead of carrying the old one forward.

Read also

Open-to-buy explained: how fashion retailers budget their buying

How do you apply the curve to a buy or an allocation?

Multiply the total forecast units for a style by each size's percentage to get a target per size, then round to match the case pack. If a supplier ships in fixed size ratios (pre-packs), compare the pack ratio with the curve: Blue Yonder notes that pre-packs save labour but can leave orphan sizes that need markdowns. A mixed approach, with pre-packs for the core sizes and open sizing for extremes, is often a compromise.

For wholesale, apply the curve to the retailer's order, not only to your own buy. Show the buyer a suggested size split per style, based on the retailer's own sell-out where available, and let them override it. A suggestion with a visible reason is more likely to be accepted than an unexplained one.

Finally, test the curve against the most recent season before using it: if it would have predicted the size mix poorly, investigate the data before trusting it.

Frequently asked questions

What is a size curve in fashion buying?

It is the percentage of demand falling to each size in a size run, derived from past sales and used to split a buy or allocation across sizes.

How do you correct a size curve for stockouts?

For each size, use the average daily sales rate on days it was in stock, multiply by the number of selling days, then normalise across sizes, as in Inventory Planner's method.

Should I use one size curve for all products?

No. Fit, category, location and channel can all shift the size pattern, so curves are usually segmented.

How do you make a size curve for a new style?

Use the curve of comparable items with similar attributes, then review after the first sales and adjust.

GuideThe complete guide to AI in fashion merchandising and buyingRead the complete guide