Type what each size sold, what came back, and what is still on the shelf. You get the order per size, next to an even split and next to what you would order if you ignored returns.
The numbers it opens with are the Heavyweight Tee from the sample store on our homepage. That is a generated store, not a real merchant. Replace them with your own style.
Net, not gross. Each size’s demand is what it sold minus what came back. The sizes that sell most are often the sizes returned most, so splitting on gross sales buys more of what you are already refunding. The arithmetic, worked through.
The split follows the style. Each size gets its share of the style’s net demand. The style needs its daily net rate times the days until the order lands plus the days it should last. Each size’s order is its share of that, less what is on hand, and never below zero.
What is on the shelf counts. This is why an even split goes wrong: it buys more of the sizes that are still sitting there, and the sizes that sold out run out again. And a size that sold out weeks ago stopped selling because there was none, not because nobody wanted it, so each size is ordered by its share of the whole style rather than by its own last few weeks.
Thin history is flagged. Under 24 net units, the calculator says the curve is a guess, because it is.
This is the same calculation the Sizecurve Shopify app runs, and a test holds the two to each other. The app does it for every style at once from your real orders and refunds, and returns are counted against the month the item sold, not the month it came back. Install it from the Shopify App Store, or the free check will tell you which of your styles are missing their middle size right now.