Trailing 7-day moves with the latest signal; trailing 28-day reads steadier but lags. Normalize per store or per visitor when comparing across channels.
SKU velocity is the per-SKU rate of sale. The window is a choice: a trailing 7-day velocity moves with the latest signal, a trailing 28-day reads steadier but lags. Omnichannel brands usually normalize further to units per day per store (so a 200-store rollout doesn’t masquerade as demand growth); DTC analysts often normalize to units per day per visitor or per session to strip out traffic shifts.
Velocity is a rate; sell-through rate is a cumulative percentage of starting inventory sold over a window. A SKU can have a high velocity and a low sell-through if the buy was huge, or low velocity and a high sell-through if the buy was thin. Planners read them together — velocity tells you how fast it moves, sell-through tells you how much of the bet has paid out.
Velocity does three jobs at once. It sets the reorder point: roughly <Term slug="lead-time" text="lead time" /> × daily velocity + safety stock, the textbook form that real replenishment tools refine with service-level targets and demand-variance terms. It drives assortment decisions — ranking SKUs by velocity descending exposes the long tail that most DTC brands defer cuts on, and that ranking is the input to ABC analysis. And it informs marketing prioritization: high-velocity SKUs earn hero placement on the PDP grid, paid creative, and email features because incremental traffic converts more reliably on proven sellers than on speculative ones.
The number breaks in three predictable ways. New SKUs have no baseline — the first two weeks are small-N noise, and treating them as a velocity read locks in a forecast off a coin flip. Stockouts make velocity look weak when demand was actually fine: a SKU at zero inventory shows zero sales, which is censored data, not low demand. The correction is to compute velocity only over in-stock days, or treat stockout days as missing rather than zero. Seasonality is the third trap. A swimwear SKU’s July velocity is not its January velocity, and a trailing-28 window read in March will badly under-forecast May. Seasonal categories need like-for-like comparisons — same week year-over-year — not raw trailing windows.
Read velocity alongside in-stock rate, stockout rate, and weeks-of-supply (on-hand units / daily velocity) to separate demand signal from inventory mistakes. Inventory turnover is the portfolio cousin: turnover is the blended figure across the catalog, velocity is the per-SKU read underneath it. The slice that pays back the time it takes to build is velocity-ranked-descending, because the long tail is where DTC brands defer cuts they already have the evidence for.