After Black Friday: why the reorder report overstates December demand

Black Friday is 27 November 2026 and Cyber Monday is 30 November. For four days most stores sell at some multiple of their normal rate, and then on Tuesday 1 December the rate goes back to normal while the reorder report does not. It keeps suggesting Black Friday quantities for most of December, because the sales velocity it uses is a trailing average and those four days are still inside the window.

This post is about that specific problem: what the Restock Ops reorder report calculates, exactly how long a weekend spike distorts it for each window length, and the three ways to correct for it without turning the report off. The holiday ordering post covers the run-up to the season; this one starts the morning after.

What the velocity is

The report reads orders created in the last N days, where N is the velocity window in Settings (7 to 365 days, 30 by default, one value for the whole shop), counts the units of each variant in those orders, and divides by N. Cancelled orders are excluded; the what counts as a sale post has the finer points on refunds, draft orders and locations. The result is units per day, shown in the Velocity column with the raw unit count beside it in brackets.

Everything else in the row follows from that number. Days of cover is position (on hand plus incoming) divided by velocity. The target is velocity multiplied by lead time plus days of cover, rounded up. The suggestion is target minus position, rounded up again to the supplier's case size and minimum. Double the velocity and you double the target; the suggested quantity more than doubles, because position stays the same.

A weekend inside a window

Take a product that sells two a day in an ordinary week, with a supplier lead time of 14 days and 14 days of cover, so the report aims to hold 28 days of stock. Normally that is 56 units. Over the four days from Black Friday to Cyber Monday it sells 80 units, ten times its usual rate, which is unremarkable for a promoted item.

On 1 December, with the default 30-day window, the report looks back to 1 November. It finds 26 ordinary days at 2 a day (52 units) plus the 80 from the weekend: 132 units in 30 days, a velocity of 4.4 a day. The target becomes 124 units instead of 56. With 40 on hand, the suggestion is 84 units where a normal week would suggest 16. Days of cover reads 9 instead of 20, so the row also looks urgent.

Nothing is wrong with the arithmetic. 132 units really were sold in those 30 days. The report cannot know that 80 of them happened in a four-day burst that is not coming back, and it does not try to guess, for the same reason it does not guess after a stockout: a trailing average that quietly edits its own inputs is one you cannot check.

How long it lasts, by window length

The spike leaves the window exactly N days after it happened. With a 30-day window the Black Friday units drop out on 27 December and the Cyber Monday units on 30 December. With a 90-day window they stay until the end of February. With a 7-day window they are gone by 8 December, but for the week they are inside it the distortion is savage, because four loud days are being averaged with only three quiet ones.

December is not normal either

There is a second effect that the table hides. Once the weekend leaves a 30-day window at the end of December, the window now contains December itself, and for anything giftable December runs above the yearly average even without a promotion. So the velocity steps down at the end of the month but does not land on the ordinary rate until late January, when the window is finally made of January days. For a store that reorders every two weeks, that is three or four reports in a row that run hot to different degrees.

The reverse case exists too. Products that sold out on the Friday afternoon and stayed out until the restock landed have a velocity that is too low, not too high, because nothing can be sold from an empty shelf. Those rows need the opposite correction, and the stockout post deals with them.

Three corrections, and when each fits

The first is to read the brackets. The Velocity cell shows the unit count next to the rate, and you know roughly what you sold over the weekend. Subtract it in your head, divide by the window, and you have the underlying rate for that row. For a store with twenty rules this is quicker than any setting, and the draft purchase order the report creates has editable quantities, so correct the line there before marking it ordered. The draft purchase order post shows where.

The second is to shorten the window in Settings from about 8 December until the end of the month. Seven days is the minimum. For the week the weekend is inside a seven-day window the report is useless, so make the change after the weekend has aged out of it, not on 1 December. The cost is noise: a seven-day average swings with every quiet Tuesday, and the setting applies to every rule in the shop, including the slow sellers for which seven days of sales is two units and a coin toss. Set it back to 30 in the first week of January.

The third is a manual velocity on the handful of rules that matter. A rule with a manual rate ignores the orders entirely and uses the number you typed, and the Velocity cell says so. It is the right tool for the five hero products that did most of the weekend's volume and whose ordinary rate you know, and the wrong tool for a long tail, because every manual rate is a number somebody has to remember to clear. Put the clearing in the January calendar now; a manual rate left over from Christmas is how a store orders Christmas quantities in March.

What to do this month

• Before the weekend: note each hero product's ordinary units per day. You will not be able to see it again until late January.
• 1 to 7 December: do not act on suggestions for promoted items without subtracting the weekend's units. Expect the days of cover column to look alarming and check on hand yourself.
• 8 December: if you reorder weekly, shorten the window to 7 days; if you reorder every two weeks or less often, leave it at 30 and correct the few rows by hand or with a manual rate.
• Through December: watch for the opposite problem on anything that sold out, and remember that December's own rate is high for gifts.
• First week of January: window back to 30, clear every manual rate, and write down the multiplier each hero product actually hit. Next November that number replaces the guess.

A trailing average is a blunt instrument, and the point of a report built on one is that it stays blunt in a way you can see and correct, rather than clever in a way you cannot. The four days after Thanksgiving are the loudest test of that each year. Know which window you are on, know when the weekend leaves it, and the report goes back to being right on its own.

FAQ