Size curve: how to split a buy across sizes
Uses sales by size to compute each size’s share and splits the quantity to buy across sizes. Compares with a flat split and shows how many units would end up in the wrong sizes.
At a glance
- What question does it answer?
- How to split the quantity to buy across sizes using historical sales.
- What data is needed?
- Sales by size for a period when all sizes were in stock and at full price.
- How to read the result?
- The table shows each size’s share and quantity by the curve against an equal split. “Units in the wrong sizes” estimates the price of a flat split: that many units will have to clear on markdown. Example on the default values: A buy of 1,200 units based on sales of six sizes: the leading size gets 388 units and the two leading sizes give 56.2% of sales. With an even split, 351 units would land in the wrong sizes. If the two leading sizes give over half of sales, keep them in depth and the ends of the run at a minimum. Next: “Size-run completeness and core sizes” and “Size curve by store cluster and pack composition”.
Why and how to reason
Demand for sizes is uneven: in the middle of the range people buy far more than at the edges. If you buy equally, M sells out quickly and XXL sits. So the buy is split by each size’s historical share of sales.
Share = size sales / total sales. Quantity to buy = share x total volume. But this gives fractions, and the sum after rounding may not match, so a special rounding is used that keeps the sum equal to the order. The minimum per size and the pack multiple reflect real factory and display constraints.
Example. Sales: XS 90, S 310, M 420, L 300, XL 140, XXL 40, total 1,300. A buy of 1,200: M = 1,200 x 420 / 1,300, about 388, S about 286, L about 277. The two leading sizes (M and S) give 56% of demand. With an even split of 200 units per size, about 351 units would land in the “wrong” sizes.
When to use it
It comes in handy when ordering a new run, when you have size sales history for a similar style. Suitable for apparel and footwear. It does not suit a new style with no analog: use the category curve then. Assumption: the demand structure will persist. Curves differ by region and channel: for an online store the edge sizes sell differently than in a store. Re-check the curve after every season.
How to read the result
Compare the buy by curve with an even split: the difference in “units in wrong sizes” shows how much you would lose without the curve (351 of 1,200). The leading size shows where to hold the most stock. But do not forget minimums: single sizes are needed so the display looks complete. Decision: order by the curve but with a minimum per size, and re-check it after each season. Also check the minimum per size: if it is above the edge share you overpay for rare sizes. It is worth comparing the result with the curve of the region or channel where the run is going.
Common mistakes
The usual mistakes: taking sales only from sizes that sold out (hidden shortages distort the curve); not noticing that M ran out earlier; using the curve of another category. Also: rounding each size separately so that the sum does not match. Also: transferring an apparel curve to footwear, where the run is longer.
What data is needed?
- Sales by size for a period when all sizes were in stock and at full price.
- The total quantity to buy and the multiple per size if the factory ships in packs.
- Where to take it: sales by size from the sales report of the model or a similar one for a period when all sizes were in stock and not discounted (1C, ERP, warehouse); the multiple from the factory’s terms (size runs or cartons). For a new fit use the nearest model’s data.
Inputs
| Field | Unit | Value in the example | Meaning |
|---|---|---|---|
| Total quantity to buy | units | 1,200 | How many units of the style you buy in total; the calculation splits them across sizes. The defaults are set for apparel. For footwear count the shares by foot length (half sizes), accessories have no size curve, for children’s use height: see the size reference. Take your own data. |
| Minimum per size (optional) | units | 0 | A size with no sales would get 0 units and the size run would break. Enter the minimum you keep per size, for example one pack; the rest is split by the curve. There is no single norm in Russia: use your own history. Optional. Empty: the field is ignored (counted as zero in sums). |
| Multiple per size | units | 1 | If the factory ships in packs, enter the pack size: quantities round up to a multiple, so the total can slightly exceed the buy. |
How it is calculated
| Step | Formula |
|---|---|
| Size share | Size sales ÷ sales of all sizes |
| Buy | Share × total, rounded so that the sum matches |
Worked numbers
Default values from the app; in the interactive version you enter your own.
- Top size
- 388 · MThe size with the largest share of the buy.
- Two leading sizes
- 56.2 %The share of sales in the two fastest sizes.
- Buy by the curve
- 1,200 units
- Units in the wrong sizes with a flat split
- 351How many units end up in slower sizes if you buy equally.
- By the sales curve
- Evenly
Data as a table
| Measure | By the sales curve | Evenly |
|---|---|---|
| XS | 83 | 200 |
| S | 286 | 200 |
| M | 388 | 200 |
| L | 277 | 200 |
| XL | 129 | 200 |
| XXL | 37 | 200 |
Split by size
| Size | Sales share | By curve, units | Flat, units | Difference |
|---|---|---|---|---|
| XS | 6.9 % | 83 | 200 | −117 |
| S | 23.8 % | 286 | 200 | +86 |
| M | 32.3 % | 388 | 200 | +188 |
| L | 23.1 % | 277 | 200 | +77 |
| XL | 10.8 % | 129 | 200 | −71 |
| XXL | 3.1 % | 37 | 200 | −163 |
The two leading sizes give 56 % of sales. Update the curve from the first 2–4 weeks of sales and keep separate curves for online and stores if the data allows.
Nuances and limits
- Update the curve from the first 2–4 weeks of sales and keep separate ones for online and stores.
- For a new fit use data of a similar style.
- Sales from a period when the core sizes sold out understate the end sizes: check availability by week.