Black Friday Inventory Planning: How Much Stock to Order

Every other buying decision in the year has a correction built into it. Get the quantity wrong and you reorder. The Black Friday buy does not work that way: the demand arrives inside four days, and by the time you can see it, the last order that could have covered it left three weeks ago.

That is the whole planning problem. Not forecasting a spike, which is hard but tractable, but committing to a number early enough that the commitment is still useful. This post covers sizing that commitment and clearing the date. It does not cover estimating promotional uplift in general, which how promotions affect inventory forecasting owns, or the wider Q4 calendar, which is in holiday inventory planning for Shopify.

Why BFCM is not a season

A holiday season is ten weeks long and forgiving. If week three runs hot you can usually get one more delivery in before the peak. BFCM is four days. There is no mid-event replenishment at any lead time anyone actually has, so the quantity you decided on weeks earlier is the quantity you sell from, and the only levers left during the weekend are merchandising levers.

It is also the wrong shape for a seasonal index. An index built month by month, the method in how to forecast seasonal inventory, describes how November compares to a normal month. It cannot see a four-day event inside November, and averaging the weekend into the month it sits in is exactly how a store ends up ordering a November's worth of stock spread evenly across thirty days.

The dates move, which matters more than it sounds like it should, because a store that plans against the previous year's calendar date rather than the previous year's weekday can be several days out on a deadline that has no slack in it. Thanksgiving is the fourth Thursday in November, Black Friday is the day after, and Cyber Monday is the Monday following.

Derive those from the federal holiday listing for Thanksgiving rather than copying them from an evergreen "national day" calendar page. One such page read in September 2026 listed Cyber Monday as December 1, which is a Tuesday. The dates above are the ones this post uses.

On the size of the thing, this site quotes exactly one source, because it is the only one that says how it counted. For BFCM 2025, Shopify reported $14.6 billion in sales by its merchants across the weekend, with a peak of $5.1 million a minute at 12:01 p.m. EST on the Friday. Shopify describes those figures as approximate and unaudited, says they are based on gross merchandise volume across its platform, and states that its methodology may vary year on year so prior years are not directly comparable to current ones. Read the peak-per-minute number as evidence that the weekend concentrates demand into hours rather than spreading it over four days. Read all of it as evidence of nothing whatsoever about your own store, which is the part most BFCM statistics are used to imply.

Estimating the spike

Your own past BFCM weekends are the only input with your catalogue, your customers and your traffic in them. The method is the baseline-and-uplift split covered in the promotions post, applied to four specific days and then rescaled, because the store you are estimating for is not the store that generated the history.

Cedar & Fig, 250g sold 140 units across the four BFCM days last time it ran. Its baseline back then was 5 units a day, so 20 of those units would have sold anyway and the uplift the weekend actually created was 120. The baseline is now 6 units a day, twenty per cent higher, so the uplift estimate scales with it: 120 × 1.2 = 144 units, plus the 24 baseline units the four days will produce regardless. That is 168 units of expected demand for the weekend.

Rescaling matters because an uplift measured in units is tied to the level of the store that produced it. Reusing the raw 140 on a store that has grown since will under-buy, and the shortfall lands on the four highest-traffic days of the year. The same reasoning applies in reverse if traffic has fallen.

Two things this estimate is not. It is not a multiplier you can lift from anywhere else: there is no defensible "expect three to five times normal volume" figure, and any post that gives you one has not told you which stores, which offers or which discount depths it came from. And it is not transferable between SKUs, because the uplift depends heavily on which product carried the offer.

If you have never run BFCM, you cannot estimate it from history you do not have. Use the analogous-product approach in forecasting inventory for new products with no sales history, keep the buy small enough that being wrong is survivable, and treat the first run as the one that generates the data.

Hero SKUs vs. the long tail

The spike does not land evenly across a catalogue. It lands on whatever you put in the email, the ad and the top of the collection page. That is a choice you make, which means the demand distribution is partly yours to set, and it is the single biggest reason a uniform uplift assumption fails.

Buying twenty per cent extra of everything converts one promotion into two hundred small overbuys, most of them on products that were never going to be featured. The alternative is deliberate concentration: pick the three to five SKUs that will actually carry the weekend, buy depth on those, and leave everything else on its normal cover.

Two consequences follow. A hero SKU that runs out mid-weekend does more damage than its own margin, because it is carrying the traffic the whole event was built to attract, and there is nothing behind it. And a hero SKU needs a nominated substitute decided in advance, so that swapping the featured product is a two-minute job rather than an argument at nine on a Friday night.

Keep the long tail out of it. If a product is not in the campaign, it is having an ordinary weekend with a bit of extra site traffic, and its normal reorder point handles that.

Sizing the order

The 168-unit estimate is a projection, not a measured average, so the buffer goes on top of it and it goes on wider than usual. How much wider is a judgment about this SKU, not a rule anyone has sourced. Twenty per cent on Cedar & Fig gives 168 × 1.2 = 201.6, call it 202 units, and the supplier's case pack of 24 rounds that up to nine packs, 216 units. At $7 a unit that is $1,512 committed against $3,888 of retail value. Rounding an order up to a supplier minimum is its own trade-off, covered in how much to order.

Then the date, which is the part that actually decides whether any of it happens. Counting backwards from the Friday: 3 days to have the stock live and sellable before the weekend opens, 5 days to receive it, 4 days of buffer for lead-time variance, and the supplier's 12-day lead time. 3 + 5 + 4 + 12 = 24 days. The purchase order leaves 24 days before Black Friday or it does not matter what quantity is on it.

The deadline moves with your supplier, not with your ambition.

The 12 days of buffer are Cedar & Fig's: 3 on the shelf, 5 receiving, 4 for lead-time variance derived from a 2-day standard deviation. A supplier with a wider spread needs a wider buffer, which pushes every date in that table further left. Measuring your own spread rather than trusting a quoted lead time is covered in how supplier lead times affect your forecast, and it is the input this entire calculation is most sensitive to.

The cost of both errors

Compressing a spike into four days compresses both failure modes, and it is worth being clear that this is a trade rather than a rule pointing one way.

Selling out early costs the sale and the traffic that was bought to produce it. On an ordinary Tuesday a stockout costs you a customer who might come back. On the Friday it costs you the highest-intent traffic of the year, arriving at a page that cannot convert, after you have already spent to get it there. Working out what that is worth for your own store, rather than quoting a figure at it, is covered in how to calculate the cost of a stockout.

Over-ordering costs less on the day and more afterwards. The stock is still there on the Tuesday, but the willingness to pay for it has changed: a product the market has just seen discounted for four days is harder to sell at full price than it was the week before, and the obvious remedy is another discount. Ten ways to sell excess inventory without losing margin covers the ways out.

Which error to lean towards depends on the SKU rather than on a principle. A high-margin item with a long shelf life that sells all year tolerates an overbuy easily, because the surplus is just next quarter's stock bought early. A low-margin item, a perishable one, or something whose demand genuinely evaporates in December does not, and on those the honest position is to buy tighter and accept selling out. Decide that per hero SKU, before the order goes out, and write down which way you chose.

During the weekend

No lever available between Friday and Monday changes how many units exist. All of them change where the demand goes.

Watching cumulative sell-through per SKU across a four-day window is the kind of thing that is easy for five products and impossible for two hundred at once. StockCue lets you flag a promotion window on a variant with an expected uplift multiplier so the buying quantity accounts for it directly, and keeps those days out of the baseline it recalculates everywhere else. Forecasting is on every plan including Free; purchase orders and receiving start at Starter.

After the weekend

The four days will sit in your sales history looking exactly like ordinary demand unless you mark them otherwise. Left unflagged they raise every rolling average that includes them, and the store spends the following quarter reordering against a rate it will not see again until the next November. Pulling those days back out of the baseline is a separate job with its own method, covered in the promotions post. Do it in the first week of December, while it is still obvious which days were the promotion.

Then record the weekend properly, because this is the input that turns the next estimate from a guess into a measurement. Per hero SKU: units sold each of the four days, the baseline rate going in, the buffer you added, whether it sold out and on which day, and what you paid for the traffic. That last one is what turns a stockout from a story into a number.

Whatever is left over is a decision, not a leftover. If the SKU sells all year, it is simply stock and it will clear at its normal rate. If it does not, the discount that clears it in the first half of December is shallower than the one that clears it in February, and the calendar is not on your side.

Frequently Asked Questions

How much stock should I order for Black Friday?

Start from what the same SKU actually did over your own past BFCM weekends, rescaled to your current baseline sales rate, rather than from a multiplier borrowed from an article. Take the uplift your own history shows, scale it by however much your baseline has grown since, add back the baseline units the weekend would have sold anyway, then add a buffer wider than you would use on an ordinary reorder because the whole figure is a projection. If you have never run BFCM before, the honest method is comparing the product to something similar you already sell rather than inventing a multiplier.

When is the deadline to place a Black Friday purchase order?

Count backwards from the Friday: the days you want the stock live on the site beforehand, plus your receiving time, plus a buffer for your supplier's lead-time variability, plus the supplier's lead time. On a twelve-day lead time with three days of pre-weekend cover, five days of receiving and a four-day variance buffer, that is twenty-four days before the Friday. On a sixty-day lead time the same arithmetic gives about ten weeks, which means committing before your own run-up has produced any signal.

Should I plan Black Friday and Cyber Monday as one event?

For inventory purposes, yes. You cannot realistically receive stock between the Friday and the Monday, so the whole weekend draws down one pool of stock and has to be bought as one quantity. Commercially they can still be different events with different offers and different featured products, but that changes which SKUs the demand lands on, not whether you get a second chance to order.

What should I do if I sell out on the first day?

Accept that no purchase order placed on the Friday will arrive before the Monday, and treat it as a merchandising problem instead of a buying one. Pull the sold-out SKU out of any paid traffic so you are not paying for clicks on an unavailable product, move the featured offer onto something you do have depth in, and decide deliberately whether to take backorders with a stated ship date or turn the product off. Then write down the number it sold out at, because that is the most useful input you will have the next time you size this buy.