Did Black Friday Really Win, or Did It Borrow From December? A PPC Reality Check
Your Black Friday PPC report closes at the peak, so it can't see the orders you borrowed from December. British retail data and promotion research show how big that loan can be. Here are three checks on last year's data and one test to set up before the sale.
Most Black Friday reports are written too early.
You know the moment. The report arrives the day after Cyber Monday. Orders are up. Revenue is up. ROAS looks like a record. In my talk at SMX Advanced Europe in Berlin in September 2026, I described it like this: "It looks amazing. Wow, we've done it. We met our KPIs."
Now look at the same week from the point of view of the shop owner. In my experience, in about one case in three, a sale like this was unprofitable once we counted the discount and the margin.
I use a simple rule for any promotion: a salary advance is not a pay rise. You get more money this month and less money next month. A Black Friday discount can do the same with orders. You get more orders this week and fewer in December.
In Google Ads, Microsoft Ads or any other PPC platform, the promotion week will always look like the win. The platform reports what happened, not what would have happened without the discount. The borrowing only shows up in your order history and your margin.
What frustrates me most is the next step. In January, the same team looks at a soft December and blames the economy.
Does Black Friday steal sales from December?
In Britain, the official numbers say yes, at least in part. Black Friday became a big event in British retail around 2013 and 2014, and the Office for National Statistics tracked what it did to the shopping calendar.
Before seasonal adjustment, here's how the quantity of goods bought in Great Britain grew month on month:
Growth into November: 8.7% in 2013, 13.1% in 2018.
Growth into December: 17.6% in 2013, 11.6% in 2017, 9.8% in 2018.
The figures come from the ONS bulletins for November 2018(1) and December 2018(2).
The ONS December 2017(3) bulletin put it plainly: "Consumers continue to move Christmas purchases earlier, with higher spending in November and lower spending in December than seen in previous years." The same bulletin describes "Black Friday sales in November encouraging spending to be brought forward from December".
The pattern didn't stop. In December 2023(4), non-food sales volumes in Great Britain fell 3.9% on the month, after a 2.7% rise in November. Both figures are seasonally adjusted, so this is more than the normal Christmas pattern. Retailers told the ONS that part of the fall was "because of consumers purchasing gifts earlier than usual, in November".
That's the dip after Black Friday. There's also a dip before it.
Do shoppers wait for Black Friday?
Yes. Shoppers know the date months in advance, so some of them hold back. Helen Dickinson, Chief Executive of the British Retail Consortium, described October 2025(6) like this: "Many delayed spending, waiting for Black Friday deals and cooler temperatures before buying toys, electronics and clothing."
Research shows the same dip in store data. Sandrine Macé and Scott Neslin made 39,441 estimates of pre- and post-promotion dips in 83 stores and ten product categories(7). Both dips were stronger for high-priced, frequently promoted, mature products with a high market share. Many Black Friday bestsellers fit that description, so check your sale list against it before you choose what to discount.
So the spike has two dips next to it:
One before it, while people wait for the deal.
One after it, when people already have what they bought.
How much of a Black Friday spike is borrowed?
The best research on this uses store sales data, not Black Friday press releases. Harald van Heerde, Peter Leeflang and Dick Wittink split promotion spikes(8) from four store-level datasets into three parts of about one third each:
Switching from other brands.
Growth of the category.
In their words, "primary demand borrowed from other time periods".
In their data, discounts supported by feature advertising, the printed flyers of that time, were strongly linked with borrowing from other weeks. With deeper discounts, the share of category growth went up and the borrowed share went down. My reading for paid search: the discount you advertise most to your existing customers is the one most likely to borrow from your own December.
If you sell your own brand, switching is a real gain, because those orders came from a competitor. In a shop with many brands, a move from one of your brands to another isn't a gain. Category growth is real. The borrowed third is different. Your report counts it as a Black Friday win, and then the same orders are missing in December.
Do Black Friday customers come back?
A deeper discount does different things for different customers. Eric Anderson and Duncan Simester ran three large field experiments(9) with a catalogue retailer that sold durable goods by mail:
Deeper discounts increased future purchases by first-time customers.
The same deeper discounts reduced future purchases by established customers.
One of the reasons they name is forward buying: customers buy now what they would have bought later.
For your regular customers, Black Friday works like a loan. They buy earlier, at a lower price, and then buy less later. For new customers, it's an introduction.
The question is whether the introduction lasts. Hello Retail(5), which sells personalisation tools to webshops, looked at about 1.9 million Black Week orders from 2025 across webshops in the Nordics and Benelux. 68% of Black Friday buyers hadn't bought from the same store in the previous 12 months, and they brought 67% of the weekend's revenue. But only 4% of a store's Black Friday buyers had also bought there on the Black Friday before. Their own summary is short: "Very few of them come back."
New to the store doesn't always mean new demand. Some of these people would have bought from you in December anyway, and some are a competitor's customers who came for the price. The week report can't tell you which.
How do you measure the real Black Friday lift?
You can't measure this year's Black Friday before it happens. You can measure last year's, and you can do it before the sale starts.
1. Count six weeks against a normal week
Take daily orders from your shop platform or CRM, not from Google Ads. A discount code or an order tag lets you separate promotion orders from full-price orders. Count returns too. A Black Friday order that comes back in January isn't a win.
Don't compare with the same weeks of last year. Last year also had a Black Friday, so the borrowing is in both years and the comparison hides it. Compare with a normal week instead.
Here are the dates for last year:
Black Friday 2025 was on 28 November, so week 0 is 24 to 30 November 2025.
The six weeks from week -2 to week +3 run from 10 November to 21 December 2025.
Your normal week is the average week from 29 September to 2 November 2025.
If some shoppers were already waiting for the deal in October, this normal week is a little low, and the check will show less borrowing than there was.
December has its own Christmas rise, so the normal week has to move with the season. Products that weren't in the sale show you that rise.
Express each week as a share of its normal week, for sale products and for non-sale products separately. The difference between the two lines is what the promotion did. It isn't a perfect control, because some shoppers switch from non-sale to sale products, but it's far better than a comparison with last year's Black Friday.
In this example, the promotion week brings 900 extra orders. The two weeks before it lose 120, and the three weeks after it lose 180. The real result is 600 extra orders, not 900. That's still growth, just smaller than the report says. Real lift is the spike minus the dips around it. Step 3 tells you whether it was a win.
2. Split the buyers
Mark every Black Friday order as a new or an existing customer, using purchase history in your CRM. For existing customers, compare the date of this order with their usual time between orders. Count the ones who ordered earlier than usual and then placed no order in December. That part of the spike was borrowed.
For new customers, put one check in your calendar for 31 January. Divide the number of second orders by the number of new Black Friday customers. This number tells you whether the discount bought customers or one-off orders.
3. Count margin, not revenue
Take the existing customers' orders that step 2 marked as borrowed. Multiply them by the average discount on those orders. That sum is the discount you gave on orders you were going to get anyway. Set it against the gross profit of the extra orders from step 1, after returns and ad spend. None of this appears in Google Ads. This is the step where a "record" sale can turn out to be unprofitable.
How to set up a Black Friday holdout test
The only clean answer is a group of customers who didn't get the promotion. Set it up before your first Black Friday email goes out.
The simplest version starts in your email tool:
Take a random 10% of your customer list and leave them out of all Black Friday emails and offers.
Upload the same people as a customer list to Google Ads and Meta.
Exclude that list from every sale campaign, otherwise your ads still show them the offer.
Compare orders per customer with the other 90% until 31 January.
The price on your website is still the same for everyone. So this test measures what your emails, PPC campaigns and social ads add for existing customers, not the whole promotion. That's the part you control and the part you pay for. If the difference is small, your best customers would have bought without the emails and ads.
If you sell in several regions, a geo test is the next step. Google describes a go-dark geo experiment as one where you "completely shut off spend in test geos"(10).
There's one trap. If the campaign is limited by budget, removing the test regions makes Google move the unspent money into the other regions, and the baseline goes up. Google's advice is to cut the daily budget to the run rate of the control regions and to use an individual campaign budget, not a shared one.
Conversion Lift(11) based on geography does a similar job, but it isn't available for all accounts and you need your Google account representative to get it.
What else should PPC teams set up before the sale?
Set the report dates now. Black Friday 2026 is on 27 November, so week +3 ends on 20 December. Agree with the client or the finance team that Black Friday is judged on 21 December 2026 and again on 31 January 2027, not the day after Cyber Monday.
Give the deepest discount to people who aren't yet customers, for example with a code that works only on a first order. In Google Ads, customer lifecycle goals have a "New Customer Only" mode(12) for Search, Performance Max, Shopping and Demand Gen. Google says you should set up a separate campaign to reach existing customers when you use it. Give regular customers a different reason to buy, such as a full-price bundle or free delivery.
Don't set a seasonality adjustment from memory. It tells Google's AI bidding to expect a jump in conversion rate. Google says to use seasonality adjustments "only if you expect major changes to conversion rates, because Smart Bidding already manages seasonal events" (13). They're "ideal for short events of 1–7 days" and "may not work as well" for periods "more than 14 days at a time".
Optmyzr's data(14) shows why this matters:
Among advertisers without adjustments, the Black Friday conversion rate lift fell from 17.5% in 2022 to 11.9% in 2023 and 7.5% in 2024.
In 2024, across 5,886 accounts, advertisers who used adjustments saw CPCs rise 34% and ROAS fall 15.7%.
Advertisers who left Smart Bidding alone saw ROAS rise 5.7%.
The adjusters did buy more volume, but their extra growth in conversion value shrank from about 25 points above everyone else in 2022 to about 6 points in 2024.
So if this year's offer isn't much bigger than last year's, leave the adjustment off. If you set one, set it only for the days of the offer. Volume bought at a worse ROAS in the peak week can be the same December demand, bought earlier and at a higher price.
When does Black Friday really win?
Black Friday can be a real win. Not every product is bought in advance. A product that isn't a gift, and that people don't buy in bulk, has less future demand to borrow. And a big share of Black Friday buyers are new to the store.
The answer to "did Black Friday win?" comes from three checks:
Extra orders for weeks -2 to +3 against a normal week, ready on 21 December.
Gross profit on those extra orders minus the discount on borrowed orders, also ready on 21 December.
The share of new Black Friday customers who bought again, ready on 31 January.
If all three are good, Black Friday won, and you have a win you can defend in front of any finance team. If the extra orders are much smaller than the peak and almost no new customers came back, Black Friday borrowed from December. It paid for the loan with your margin.
Conclusion: judge Black Friday in January, not on Cyber Monday
The report you get the day after Cyber Monday is written at the peak. Before your first Black Friday email or PPC campaign goes live:
Agree the report dates: 21 December and 31 January.
Set up a holdout group and exclude it from your sale campaigns.
Tag every order as new or existing.
In January you'll know whether your Black Friday won or borrowed.
Comments