How do I forecast demand and build a purchase plan for a seasonal Shopify brand?

Do not use trailing twelve months as if December were July. A seasonal brand forecasts by week against last year, then overlays what you know now: live velocity, inbound already on the water, the marketing calendar, and cash.

Build a weekly demand curve

Build a weekly demand curve from last year's shipments (not orders, not Shopify "available"). Align the calendar (4-5-4 if that is how you close). Lift or cut weeks that have a named event: BFCM, Mother's Day, a drop date. New SKUs inherit a sibling style's curve, not a flat weekly average. A 365-day ABC class will hide the season. Use a 90 to 180 day lookback for trend, and last year's same weeks for the peak. Plan at style or color, then explode to size with a size curve. Planning every variant as its own forecast is how you overbuy XL and stock out of M. Wholesale and DTC are not one curve. An 850 you already hold is committed. Do not buy it twice, and do not let Shopify velocity eat the hold you already promised.

Turn the curve into a purchase plan

Turn the curve into a purchase plan by walking backward from the in-stock date. Cover is demand during actual lead time, plus safety. Actual lead time is receipt date minus PO date, on the last few POs, not the supplier's promise. Safety is the miss you will not survive (a slipped boat, a BFCM week). It is not a second forecast. Subtract inbound already purchased and already named. Ignore a named container and the system will recommend a second factory PO the week Shopify goes sold-out. Sold-out on the storefront is not the signal if that container is already on the water. Round up to MOQ and case-pack. Then stop and look at cash: the deposit, the balance when the container clears, and what is already committed to wholesale holds. A mathematically perfect PO you cannot pay is not a plan. This is a buy you will send. It is not a min/max policy that recommends a weekly top-up.

Freeze the buy and exception-manage

Freeze the buy. Exception-manage after. Weekly: SKUs that will stock out before inbound, POs that slipped, DTC velocity that is eating a wholesale hold. Draft the next PO from those exceptions. A human sends it. Do not let a storefront assistant draft a buy it cannot see (3PL counts, FBA reserved, landed cost, open 850s). Receive the container into the hold or pool that funded it, or the remainder math lies the week it docks. The forecast is a calendar you will be wrong about in pieces. The purchase plan is the cash and the dock dates you can still move.