Sales velocity: units per day, and why the window matters
Sales velocity is how many units of a SKU you sell per day, averaged over a window of recent days. The window you pick, 7, 14 or 30 days, decides which version of your demand that number describes, and every reorder decision inherits it.
What sales velocity means
Sales velocity is units sold per day for a single SKU, calculated by taking the units sold across a fixed window of recent days and dividing by the number of days in that window.
That is the whole formula. Units sold, divided by days. What makes it interesting is the window, because the window is a judgement call and nobody makes it for you.
A short window says what is happening now. A long window says what usually happens. Both are true. They are rarely the same number.
A worked example
Take one SKU. Over the last 30 days it sold 60 units. Over the last 7 days it sold 21.
- 30-day window: 60 divided by 30 is 2.0 units per day
- 7-day window: 21 divided by 7 is 3.0 units per day
Now put a 14-day supplier lead time against each. At 2.0 per day, you expect to sell 28 units while you wait for the shipment. At 3.0 per day, you expect 42. Same SKU, same day, same supplier, a gap of 14 units. That gap is the window, and nothing else.
Picking a window: 7, 14 or 30 days
| Window | What it catches | Where it misleads |
|---|---|---|
| 7 days | A sudden lift from a promotion, a feature, a post that landed. Fast-moving staples where demand is steady week to week. | One unusual day carries a seventh of the weight. A single bulk order or a quiet bank holiday swings the number hard. |
| 14 days | A middle setting. Two full weekly cycles, so weekday and weekend patterns balance out without burying a recent trend. | Still reacts to a two-week promotion as though it were the new normal. |
| 30 days | The underlying run rate on slower movers and long tail SKUs, where a week of data is mostly noise. | Slow to notice a real acceleration. A SKU that took off ten days ago still looks calm. |
Why the window is a real decision, not a default
Merchants often set one window for the whole store and forget it. That is fine as a starting point, but the right setting usually differs by how a SKU behaves.
Match the window to the SKU, not the store
A hero SKU that sells every day gives you enough data in 7 days to trust a short window. A SKU that sells three units a fortnight does not. Averaging 3 units over 7 days produces a velocity that looks like noise, because it is noise.
Watch for events inside the window
If a discount ran for four days inside your 7-day window, that velocity is a promotion number, not a baseline number. Reordering against it means buying to a demand level you deliberately engineered and then stopped.
Short history is a special case
A SKU listed three weeks ago has no 30 days of history. Divide by 30 anyway and you understate velocity badly, because you are averaging real sales across days the product did not exist.
The window compounds downstream
Velocity feeds lead time demand, lead time demand feeds the reorder point, and the reorder point decides when an alert fires. Change the window and you move the alert date. That is the honest reason to think about it for more than five seconds.
Related terms
Sales velocity rarely travels alone. These are the terms it sits next to.
- Lead time. The days between placing an order with your supplier and the stock being sellable. You supply this number. Nobody can infer it from your sales.
- Lead time demand. Sales velocity multiplied by lead time. The units you expect to sell while you wait.
- Reorder point. The stock level at which you place the order. At minimum it is lead time demand, plus whatever buffer you hold.
- Safety stock. The buffer above lead time demand that absorbs a late shipment or a better week than expected.
- Days of inventory remaining. Current stock divided by sales velocity. The same window applies, so this figure moves when the window moves.
- Stockout. The SKU hits zero and stays there. The cost is the sales you cannot make and the customer who goes elsewhere.
For how these combine into an actual order date, see the guide on turning sales velocity and lead time into an order.
Questions
How do you calculate sales velocity?
Take the units sold for one SKU across a fixed window of recent days and divide by the number of days in that window. Sixty units over 30 days is 2.0 units per day.
Which sales velocity window should I use?
It depends on the SKU. Fast, steady sellers tolerate a 7-day window and benefit from its speed. Slow movers need 30 days, because a week of their data is mostly noise. Fourteen days is a reasonable middle setting when you are not sure.
Does a shorter window make my reorder point more accurate?
It makes it more responsive, which is not the same thing. A short window reacts quickly to a genuine change in demand, and just as quickly to a one-off spike that will not repeat.
What happens to velocity if a SKU has less history than the window?
Dividing real sales by days the product was not on sale understates velocity. Treat a new SKU's number as provisional until it has at least a full window of history behind it.
Is sales velocity the same as days of inventory remaining?
No. Velocity is units per day. Days of inventory remaining is current stock divided by that velocity. Change the velocity window and the days figure changes with it.
Can I change the window later?
The engine is built so the velocity window is a setting of 7, 14 or 30 days feeding the reorder point calculation, which means changing the window recalculates the reorder points. Restoc is not open to merchants yet, so the waitlist is the way in.