What we got wrong: we first wrote parts of this from the abstracts of four studies. After reading the full papers we could get, we changed what we said about them. Two remain abstract only, as their source notes say. The conclusion about the break-even line stands, because that’s arithmetic. What changed is the evidence for how much gets borrowed, which is wider than we said, and the advice on who gets the deepest offer.
CVS: At CVS, about 45% of promotional lift was incremental and more than half of promotions lost money. About 45% was incremental, 45% was switching inside the store, and only 10% came from later purchases, in non-holiday weeks. More than half earned less than a no-promotion baseline. [20]
Borrowed share: Set that part aside and, by our reading, borrowing is closer to half of a store’s extra sales. We default the calculators to a third as a kind starting assumption, so run a half as well. The evidence runs from about 18% to about half, by our arithmetic, and a third is an assumption near the middle.
Deepest offer: The deep offer belongs with people who’ve never bought from you. It has the most support with new or lapsed customers. The fall among established buyers in that study was mostly selection. [9]
- The usual sum says 20% off at a 50% margin needs 67% more orders during the sale. It treats every extra order as new.
- If a third of the extra orders were borrowed from the weeks around the sale, the same discount needs 150%.
- There’s a line where no volume pays: a borrowed share of 1 minus discount divided by margin. At a 50% margin and 25% off, that’s 50%.
- The evidence on borrowing runs from about 18% at CVS to about 36% in Dutch grocery to about half in a rounded decomposition, all by our arithmetic. We default to a third, which is an assumption. [20][4][8]
- In all retail across 26 European countries, the median November index rose 3.0 points between 2010-13 and 2022-25 and December’s fell 4.0, so the two together show no net gain. The timing doesn’t prove Black Friday did it. [1]
- Online-focused retailers’ November rose 10.9 points, netting +3.5 across October to January. In US online sales, November’s index stayed flat while December’s shrank. [1][2]
- At CVS, in non-holiday weeks, about 45% of promotional lift was incremental, 45% was switching inside the store, and 10% came from later purchases. More than half of promotions earned less than a no-promotion baseline. [20] Across 44 Dutch grocery events, net spending over 12 weeks was about zero. [4]
- Among 11 listed US retailers, the median holiday-quarter gross margin sat 0.9 points below the first nine months. Williams-Sonoma, which said in 2024 that it runs no sitewide promotions, went up. [5][6]
- Run the sum by category, try less depth first, aim the deepest offer at new or lapsed customers, ask suppliers to fund it, and read the result in January.
Black Friday is four weeks away, and someone in your company wants 25% off everything. Someone else will open a spreadsheet and work out how many extra orders that takes. The spreadsheet will be right, and it’ll still be missing the term that decides the whole thing.
Here’s the fuller sum, what official sales data says about the missing term, and what to do before November 27.
The sum everyone does
Say a product sells at 100 and you keep 50 after its costs. Take 20% off and you keep 30 on every order, including the ones you’d have had anyway. To end up where you started you need 50 / 30 = 1.67 times the orders, or 67% more.
The general form is discount / (margin - discount). Use contribution margin after shipping, payment fees, and returns, and compare the sale days with what you’d have sold on those same days with no discount. That second number may be lower than a normal week, because shoppers wait and rivals are loud, so October isn’t the right baseline.
Here’s how that looks across the depths retailers actually run, using contribution margin. Adobe’s 2025 US figures had discounts peaking at 30.9% off list in electronics, 25.1% in apparel, and 18.8% in furniture. [7] Switch tabs to see what changes.
| Margin | 10% off | 20% off | 25% off | 30% off |
|---|---|---|---|---|
| 30% | 50% | 200% | 500% | Never |
| 40% | 33% | 100% | 167% | 300% |
| 50% | 25% | 67% | 100% | 150% |
| 60% | 20% | 50% | 71% | 100% |
| 70% | 17% | 40% | 56% | 75% |
| Margin | 10% off | 20% off | 25% off | 30% off |
|---|---|---|---|---|
| 30% | 100% | Never | Never | Never |
| 40% | 60% | 300% | 1,500% | Never |
| 50% | 43% | 150% | 300% | 900% |
| 60% | 33% | 100% | 167% | 300% |
| 70% | 27% | 75% | 115% | 180% |
| Margin | 10% off | 20% off | 25% off | 30% off |
|---|---|---|---|---|
| 30% | 200% | Never | Never | Never |
| 40% | 100% | Never | Never | Never |
| 50% | 67% | 400% | Never | Never |
| 60% | 50% | 200% | 500% | Never |
| 70% | 40% | 133% | 250% | 600% |
Each cell is the extra orders you need during the sale just to match the profit you’d have made without it. Our arithmetic, before ad costs.
The order you borrowed
The usual sum treats every extra order as new. Some aren’t. A customer who’d have bought from you on December 9 at full price buys on November 27 at 20% off. You didn’t gain an order. You paid to move one.
So an extra order earns margin minus discount now, and costs margin times the borrowed share later. With a 50% margin, 20% off, and a third borrowed, that’s 30 minus 16.7, or 13.3 per extra order, and you need 20 / 13.3 = 150% more orders to recover the discount you gave everyone. That’s the jump between the first two tabs.
Keep going and the sum runs out. At 25% off with half the extra orders borrowed, each one earns 25 now and costs 25 later. It adds nothing, and the discount on your ordinary orders never gets paid back. The line sits at a borrowed share of 1 minus discount divided by margin. Past it, each extra order adds to the loss, and a bigger sale day only makes the loss harder to see.
At a 50% margin and 25% off, if half the extra orders were coming anyway, no volume pays.
Mediovsky arithmeticWhat each extra order really earns
Run this per category at least, and by brand where you can, because a sitewide average hides the ones past the line.
Model: margin - discount - borrowed share × margin. It assumes a borrowed order would otherwise have sold at full margin, which overstates the cost if you’d have discounted it later anyway. It also leaves out halo, extra sales of other products in the store. CVS put that at a median 0.16 units per unit of gross lift, though it counted items, and your average order already includes the rest of the basket. [20]
How much gets borrowed
Nobody has published this number for online Black Friday, as far as we can find. The nearest evidence comes from studies that could see the weeks around a promotion. Most of it is grocery and drugstore, and each study measures something slightly different. Here’s what each found, and what it implies for a store once you set aside shoppers swapping between products on its own shelves.
| Study | Borrowed share, our arithmetic | Setting | What it found |
|---|---|---|---|
| Ailawadi and colleagues, 2006 [20] | About 18% | CVS, 2003, 177 categories, holiday weeks excluded | About 45% of gross lift was switching inside the store, 10% came from later purchases at CVS, and 45% was incremental |
| Guyt and Gijsbrechts, 2020 [4] | About 36% | 44 themed events at Dutch grocers, modeled from a household panel | Volume fell 7.84% the week before, rose 8.47% over the event weeks, and was down a non-significant 0.58% over the eight weeks after. Net over 12 weeks: volume +1.08%, spending about zero |
| van Heerde, Leeflang, and Wittink, 2004 [8] | About half, from rounded averages | Four store-level scanner datasets, single-item promotions | Cross-brand, borrowed, and category-expansion effects each made up about a third of a brand’s bump |
The arithmetic: 10 / (10 + 45) at CVS, dips of (38.03 + 8 × 2.08) against a lift of 3 × 50.15 from the Dutch Table 7, reading its levels as weekly averages, which is what reproduces its net row, and 1/3 / (1/3 + 1/3) for the decomposition. The Dutch week before alone gives 25%, because the post-event dip isn’t statistically significant.
So the range is wide, and the line moves with it. At a 50% margin, a borrowed share of 18% puts the line at 41% off, 36% puts it at 32% off, and a half puts it at 25% off. We default the calculators to a third because it sits near the middle, and show a half as a test case. Those are assumptions, and neither is a finding about your store.
Each number leans its own way. CVS counted only later purchases in its own stores, and its authors expect much of the stockpiling came from other stores, which counts as a gain for CVS. [20] The Dutch model’s look-back is one week, so longer waiting isn’t captured. [4] The half is a ratio of rounded averages across single-item promotions. Its abstract says borrowing’s share tends to fall at deeper discounts, which pulls it down. The part we set against borrowing can include shoppers taken from rival stores, who may not come over when rivals discount too, which pushes it up. [8][22] Stockpiling dish soap isn’t the same as buying a coat early either. Researchers have argued that buying ahead should be more common for durable goods. Across one catalog firm’s three experiments, the authors found some evidence of it, but among established buyers it explained only about 2.5 points of a 10.8% fall in later units, and that difference wasn’t statistically significant. [9]
The one Black Friday study we found covers existing store shoppers at a Brazilian pet retailer whose first online purchase came in a one-day, online-only flash sale in 2019. Measured against the gap between the two groups in the month before the sale, they spent 128.6 masked currency units more than comparable shoppers who moved online on their own in the sale month and 134.3 less the month after, on pre-sale spending of about 490 a month. The gap stayed negative for all 11 months tracked, significantly in 10. Its authors read that “as evidence consistent with forward buying, rather than as a definitive mechanism test,” and a gap that long, running into the pandemic, may also reflect who these shoppers are. [15]
Borrowing also isn’t the only thing that sinks a promotion. More than half of CVS promotions earned less than a baseline at regular margin, which its authors call an inflated benchmark, since a lower one raised net profit. They point at margin: “all promotional units earn the lower margin, not just the incremental units.” [20] A later report on the same chain says a field test showed that dropping promotions chainwide in 15 of the worst categories would cost about $7.8 million in sales and add about $52.6 million in profit. [3] A study of 25 categories at one Midwestern grocer found total category margin fell for 55% of the brands, and its authors blame slow price recovery more than borrowed volume. [10] CVS’s point is the first term in the sum, the discount on orders you’d have had anyway. Slow price recovery is a further cost the sum leaves out. The usual sum isn’t wrong, only incomplete.
The Dutch events are the closest analog we found: mass-advertised, unusually deep, and under one theme. They also differ in ways that matter. The authors class a supermarket’s Black Friday deals as a different, seasonal kind of event, warn that stockpiling may not carry over outside grocery, and report that only about 24% of revenue sold on promotion in event weeks, against 21% otherwise, so these weren’t sitewide sales. They describe the result as “typically leading to a null effect in terms of profit.” Their back-of-the-envelope model assumes a 25% margin with retailers bearing 10% to 30% of the discount, and its averages ran from -0.44% to +1.47%. A store funding its whole discount isn’t modeled at all. [4]
What happened to December
If Black Friday added sales across the board, you’d expect the peak as a whole to take a larger share of the year. We checked, using Eurostat’s unadjusted monthly turnover for every country with complete data. [1] We expressed each month against the 12 months around it, with 100 as an ordinary month, then compared 2010-13 with 2022-25, skipping the pandemic.
Across all retail, November and December together moved by a median of -0.4 points, so the peak didn’t take a larger share of the year. Online-focused retailers look different. November gained far more, and the four months from October to January netted +3.5 points, so the season took a larger share of that channel’s year. An index can’t say whether those sales were new or came from stores.
Across all retail, the timing doesn’t fit a simple transfer either. By 2017-19 November had already gained 2.8 points, while December had lost only 1.2, so most of December’s fall came later. Read this as no sign of a net gain in the all-retail peak, rather than proof that November took December’s sales.
Here are the two months in larger markets, from the early window to the late one. Switch tabs to compare the two channels.
| Market | Seasonal index, 2010-13 to 2022-25 |
|---|---|
| Germany | Nov 103 to 107 Dec 115 to 111 |
| France | Nov 98 to 101 Dec 126 to 121 |
| Italy | Nov 99 to 102 Dec 127 to 121 |
| Spain | Nov 96 to 101 Dec 119 to 116 |
| Poland | Nov 96 to 100 Dec 117 to 113 |
| Netherlands | Nov 101 to 105 Dec 115 to 110 |
| Norway | Nov 103 to 109 Dec 126 to 115 |
| Romania | Nov 103 to 105 Dec 110 to 109 |
| United States | Nov 103 to 104 Dec 120 to 114 |
| Market | Seasonal index, 2010-13 to 2022-25 |
|---|---|
| Germany | Nov 115 to 124 Dec 114 to 116 |
| France | Nov 118 to 128 Dec 142 to 129 |
| Italy | Nov 114 to 143 Dec 117 to 142 |
| Spain | Nov 109 to 125 Dec 121 to 130 |
| Poland | Nov 104 to 116 Dec 116 to 117 |
| Netherlands | Nov 108 to 117 Dec 112 to 112 |
| Norway | Nov 118 to 147 Dec 116 to 109 |
| Romania | Nov 118 to 167 Dec 150 to 119 |
| Sweden | Nov 112 to 132 Dec 128 to 115 |
| United States | Nov 113 to 112 Dec 143 to 124 |
In the online tab, countries split. Romania and Sweden look like demand moved from December into November. Italy and Spain grew both months. So there’s no single borrowed share to look up, which is the point: a number you assumed is doing a lot of work in your plan.
Method: net turnover, unadjusted, for NACE G47 and G47.91, with each month divided by a centered 12-month average, then rescaled so a year sums to 1,200. Medians are of 26 countries (EU members with data, excluding Ireland and Sweden, plus Norway) and 18 for online. Volume instead of value gives +3.6 and -3.1.
A check on the index: an index only shows share of the year, so any month that grows pushes the others down, and direction alone proves little. If November had grown on purely new sales, the others would shrink evenly, so October, December, and January would give up about 3/11 of the gain, or 27%. Pooled across countries where November rose, they gave up 51% online and 149% in all retail. Above 100%, something besides November is shrinking those months.
Limits: nothing here isolates Black Friday. Stop the comparison at 2017-19 and December’s fall shrinks to 1.2 points. Start from 2012-14 and it’s 1.9, from 2008-12 and it’s 4.9. Across all retail it stayed negative in every window we tried, though online it reaches +0.2 from 2008-12. Cyber Monday fell in December in two of the later years and one of the earlier ones, which works against the pattern. US online shows December’s index can shrink without November’s growing: November went from 113 to 112 while December fell from 143 to 124. [2] Everyday goods moving online may flatten the calendar too.
The US rows are worth a second look, because that’s a mature Black Friday market. November barely moved in either tab while December fell, and online most other months rose, with July going from 91 to 99 and May and October from 96 to 101. A taller peak isn’t what the US shows.
None of this says your store can’t win. National totals can stay flat while you take share from a rival, and that’s the real case for joining in. It does mean you shouldn’t assume the extra orders came from nowhere. Our comparison of four holiday forecasts covers what those headline growth numbers do and don’t count.
What big retailers’ margins show
Listed retailers report enough for a rough read. We took SEC filings for the fiscal year ended around January 31, 2026, and subtracted nine months from the full year to isolate the holiday quarter. These are gross margins, with the gap in points. [5]
| Retailer | Gap | First nine months | Holiday quarter |
|---|---|---|---|
| Macy’s | -4.3 | 39.5% | 35.2% |
| Gap | -3.7 | 41.8% | 38.1% |
| Best Buy | -2.4 | 23.3% | 20.9% |
| Target | -1.9 | 28.5% | 26.6% |
| Ulta | -1.5 | 39.6% | 38.1% |
| American Eagle | -0.9 | 36.8% | 35.9% |
| Ross | -0.7 | 27.9% | 27.2% |
| Chewy | -0.5 | 29.9% | 29.4% |
| Burlington | -0.2 | 43.9% | 43.7% |
| TJX | -0.1 | 31.0% | 30.9% |
| Williams-Sonoma | +1.1 | 45.8% | 46.9% |
The median gap is 0.9 points. Compare each company only with itself, since cost lines differ. Gap’s includes occupancy, for instance, and a busy quarter spreads that thinner, so merchandise margin alone probably falls further than this shows. [11] Other costs move individual rows too: Macy’s put about 60 basis points of tariff cost on its holiday quarter. [12] These retailers also discount the rest of the year, so this measures promotion against promotion.
Even so, the largest gap on this list is about four points, and off-price chains barely moved. For scale, 25% off everything at a 40% margin takes the margin rate on those sales from 40% to 20%. A quarter blends sale days with all the others, so a gap this size doesn’t rule out deep sale days. It does show that the discount a store advertises and the margin a quarter reports are very different numbers.
Williams-Sonoma is the row to notice. Its CEO said in 2024 that the company “maintained our commitment to not offering sitewide promotions” and that this and fewer offers “have improved margins.” [6] Its brands and mix are its own, occupancy helps its margin in a busy quarter too, and those remarks predate this one, so one row proves nothing. It’s still a large retailer whose margin didn’t fall in the peak.
The other bills
Ads. The complaint is CPMs, and they do rise. In one vendor’s sample of $22.6 million in Meta spend, the median brand paid 59% more per thousand impressions in Black Friday week than in September, and 84% more on the day. But conversion rate rose 48% and cost per purchase fell 20%. [13] That’s a lone source using platform-reported purchases, and a cheap purchase is only cheap if it was going to be new. Our pieces on incrementality testing and new-customer ROAS cover how to check.
Returns. Retailers told the NRF they expected 17% of holiday sales to come back in 2025, and 19.3% of online sales across the year. [14] Those are survey estimates. Take returns out of your margin before you run any of these sums, or your margin is fiction.
Customers. The usual defense is that Black Friday brings in customers who pay back later. The evidence is thin and mixed. In one catalog firm’s experiments, about 60% off won more first-time buyers than about 30% off, and those buyers later bought more units. Established buyers later bought about 10% less, but that gap nearly vanished once the authors controlled for purchase history, and lapsed, infrequent buyers bought more. So if you can target, the deepest offer has the most support with new or lapsed customers. That’s one firm with no profit measure. [9] In the Dutch events, occasional shoppers responded most and regulars produced no net volume gain. [4]
The Brazilian preprint adds a warning. On profit, its Black Friday shoppers edged those who stayed offline, though the gain was borderline, and their forecast incremental lifetime value came out about 60% below a forecast pooled with shoppers who moved online unprompted. Its own sum has a roughly 28% discount taking about 39 months to recover at that borderline profit gain, against about 13 months if you assume they behave like those organic adopters. That counts the whole discount as the cost, gives no credit for margin on the sale-day order, and assumes flat profit, no churn, and no discounting, well past the 11 months observed. It’s one retailer, not peer reviewed. [15]
What the whole sale does to your profit
Start with what you’d sell on the sale days with no discount, then be honest about the lift.
Model: orders × price × [(1 + lift) × (margin - discount) - margin - borrowed × lift × margin] - (ad cost - later profit) × orders × lift. At these defaults the sale doubles orders and loses 9,200. With nothing borrowed, it makes 4,000.
For scale, US online spending on Black Friday 2025 was $11.8 billion, about 2.8 times the average day of a season that totaled $257.8 billion, by our division. [7] That’s a whole market on one of its two biggest days, with Christmas demand inside it, so don’t read it as the lift your discount will cause.
Less than 25% off everything
A sitewide discount pays every buyer, including the ones who needed no push. The options below shrink that group or the depth.
| Move | Why it might work | Evidence and its limits |
|---|---|---|
| Cut depth first | Each point of discount comes off every order | At CVS, deeper discounts raised net units but cut net profit [20]. In the Dutch events, uniform-price formats correlated with higher profit, possibly because their deals were less deep and encouraged less stocking up. Depth itself wasn’t significant, though it barely varied within formats, and wider scope helped, so this isn’t a case for a narrow range [4] |
| Send coded offers instead of a sitewide price | Only people who redeem get the discount | 37% more spending during the offer in email experiments on a ticketing platform, with 90% of the gain not coming through redemption. That’s against no offer among past customers, not against a sitewide price, and the author warns the lift may leak to rival sites at general retailers [16] |
| Aim the deepest offer at new or lapsed customers | First-time buyers won at depth bought more later, and recent, frequent buyers showed no gain | One catalog firm’s experiments [9], and regulars’ lack of net gain in the Dutch events [4] |
| Ask suppliers to fund it | Their money replaces yours | In the Dutch model, where retailers bore 10% to 30% of the discount, the average moved from -0.44% when that share rose in event weeks to +1.47% when it fell [4]. At CVS, even after manufacturer payments, promotional margin was often well below regular margin [20] |
| Exclude what fails the calculator | A category past the line can’t be rescued by volume | Our arithmetic. CVS’s 45% is switching within a category, so if you exclude items inside a category, count orders that switch from its full-price items as borrowed too. Across categories CVS found a small gain instead, so this says nothing against excluding a whole one [20] |
There are good reasons to run the sale anyway. You may be clearing stock you’d otherwise write off, in which case your real margin is much higher than the one on your books. If your no-sale baseline collapses that week, the discount has less to recover. And the Dutch authors saw these events as a defense, holding on to light customers who might otherwise defect to discounters. [4] Those are reasons you can state and check. “Everyone’s doing it” isn’t one of them.
Four weeks to November 27
That last step is the one almost nobody does, because by January the sale is a closed success in a slide deck. It’s also the only way to replace our default third with a figure that’s yours. Do it once and the next argument about 25% off everything gets much shorter.
Sources
- Eurostat · Turnover and volume of sales in wholesale and retail trade, monthly dataDataset sts_trtu_m, net turnover and volume, unadjusted, NACE G47 and G47.91, downloaded October 31, 2026. The seasonal indexes and medians are our calculation.
- US Census Bureau · Monthly Retail Trade: sales 1992 to presentNot adjusted, through August 2026. Online is NAICS 4541, electronic shopping and mail-order houses. All retail excludes motor vehicle and parts dealers, matching Eurostat’s scope. The indexes are our calculation.
- Ailawadi, Harlam, César, and Trounce · Practice Prize Report: Quantifying and improving promotion effectiveness at CVSMarketing Science, 2007. We could read only the abstract, and the field-test figures aren’t in the 2006 paper at [20].
- Guyt and Gijsbrechts · Evaluating the effectiveness of retailer-themed super saver eventsJournal of Marketing, 2020. Full text. Household panel across seven Dutch grocery chains. Effects are simulated from the authors’ model against a no-event baseline, and the profit range is a scenario.
- US Securities and Exchange Commission · EDGAR company facts (XBRL)Revenue and cost of sales from each company’s 10-K and third-quarter 10-Q. Holiday quarter = fiscal year minus nine months, our arithmetic. Gaps use the rounded margins shown.
- Williams-Sonoma, via The Motley Fool · Q2 2024 earnings call transcriptLaura Alber, August 22, 2024. Transcript by a third party.
- Adobe · Holiday shopping season drove a record $257.8 billion onlineJanuary 7, 2026. US online spending, November 1 to December 31, 2025. The 2.8 times figure is our division over 61 days.
- van Heerde, Leeflang, and Wittink · Decomposing the sales promotion bump with store dataMarketing Science, 2004. Abstract only, as we couldn’t get the full text. Its abstract also says the roles of cross-brand and cross-period effects tend to decrease at higher discounts.
- Anderson and Simester · Long-run effects of promotion depth on new versus established customersMarketing Science, 2004. Full text. Three field experiments at one direct mail catalog firm, about 30% against about 60% off, tracked 22 to 28 months. Units and revenue, no profit.
- Srinivasan, Pauwels, Hanssens, and Dekimpe · Do promotions benefit manufacturers, retailers, or both?Full working paper of the 2004 Management Science article. Dominick’s Finer Foods, 25 categories, 1989 to 1997.
- Gap Inc. · Form 10-K, fiscal year ended February 1, 2025Reports cost of goods sold and occupancy expenses as one line in its margin discussion.
- Macy’s, Inc. · Fourth quarter 2025 results, Form 8-K exhibitConfirms the 35.2% holiday-quarter gross margin rate and the tariff impact.
- Lebesgue · Are Meta ads worth it on Black Friday?Lone vendor, $22.6 million of Meta spend among its largest customers, September to November 2025. Median brand. Purchases and ROAS are Meta-reported.
- National Retail Federation · Consumers expected to return nearly $850 billion in merchandise in 2025Retailer estimates from a survey, not final results.
- Biswas, Yoganarasimhan, and Zhang · Channel adoption pathways and post-adoption behaviorPreprint, revised July 2026, not peer reviewed. Full text. One Brazilian pet supplies retailer, existing store customers moving online. Monetary values are masked. The authors note parallel pre-trends fail against offline-only customers.
- Stanford Graduate School of Business · Navdeep Sahni: email promotions to prospective customers really do workStanford’s summary of Sahni, Zou, and Chintagunta, published in Management Science in 2017. We couldn’t get the full paper. One online ticket platform.
- Court of Justice of the European Union · Press release on Case C-330/23, Aldi SüdSeptember 26, 2024. We also read the judgment itself.
- Federal Trade Commission · Guides Against Deceptive Pricing, 16 CFR Part 233Section 233.1 covers former price comparisons.
- Which? · Don’t believe the hype: most Black Friday deals the same price or cheaper at other times of the yearNovember 25, 2025. 175 products at eight UK retailers, tracked May 2024 to May 2025.
- Ailawadi, Harlam, César, and Trounce · Promotion profitability for a retailerJournal of Marketing Research, 2006. Author’s manuscript, full text, via the Internet Archive. 12.9 million observations from 1,816 CVS stores in 2003. The 18% is our ratio of the authors’ 10 and 45.
- European Union · Directive 98/6/EC, consolidated textArticle 6a, inserted by Directive 2019/2161, defines the prior price and allows the exceptions.
- Harald van Heerde · Decomposing the sales effect of promotionsPhD thesis chapter, University of Groningen, 1999. Five canned tuna brands. An earlier study by the lead author of [8], using a similar three-part method at one chain. We use it to show what the residual can include, not as a stand-in for [8].
Frequently asked questions
How much extra volume does a Black Friday discount need to break even?
Divide the discount by what’s left of your margin after it. A 20% discount at a 50% contribution margin needs 20 / 30, or 67% more orders during the sale, compared with what you’d have sold on those days with no discount. That sum assumes every extra order is new. If a third of them were borrowed from the weeks around the sale, the same discount needs 150%.
Can a discount lose money however many orders it brings?
Yes, under simple assumptions. Each extra order earns your margin minus the discount now, and costs your margin times the borrowed share later. When the borrowed share reaches 1 minus discount divided by margin, an extra order earns nothing, and the discount you gave on your ordinary orders is never paid back. Past that, each extra order adds to the loss. At a 50% margin and 25% off, that line is a borrowed share of 50%.
Does Black Friday create new sales or just move them?
National data can’t settle it, but across all retail it shows no sign of a net gain in the peak. Comparing 2010-13 with 2022-25 across 26 European countries, the median country’s November index rose 3.0 points in all retail and December’s fell 4.0, so the two together barely moved. Online-focused retailers’ November rose 10.9 points, netting +3.5 across October to January. An index shows share of the year, not sales, and across all retail most of December’s fall came after November’s rise, so this isn’t a clean transfer. A single store can still win share from rivals.
How much of a promotion’s sales lift is really incremental?
It varies, and most published evidence comes from grocery and drugstores. At CVS in 2003, leaving out holiday weeks, about 45% of gross lift was incremental, about 45% was switching inside the store, and about 10% came from later purchases. More than half of promotions earned less than a no-promotion baseline. We know of no published equivalent for Black Friday online.
Are shoppers pulled in by Black Friday worth less?
The evidence is thin and mixed. In one catalog firm’s experiments, about 60% off won more first-time buyers than about 30% off, and those buyers later bought more units. Established buyers later bought about 10% less, but that gap nearly vanished once the authors controlled for purchase history. A preprint on one Brazilian pet supplies retailer followed store shoppers whose first online purchase came in a Black Friday flash sale. On profit they edged shoppers who stayed offline, though the gain was borderline, and they trailed those who moved online on their own. It isn’t peer reviewed.
Do higher ad costs make Black Friday unprofitable?
Not by themselves. In one vendor’s sample, the median brand’s Meta CPM rose 59% in Black Friday week against September, but its conversion rate rose 48% and its cost per purchase fell 20%. Those are platform-reported purchases, so they don’t tell you how many of those orders were new.
What are the rules on advertising a Black Friday discount in the EU?
The EU’s Court of Justice held in 2024 that an announced price reduction must be calculated from the prior price, which Directive 98/6/EC defines as the lowest price the trader applied in at least the 30 days before it. Raising a price ahead of the sale doesn’t help, and an earlier sale in that window can lower the reference, though member states may allow an exception when a reduction is progressively increased. In the US, the FTC’s pricing guides treat an inflated former price as deceptive. Take legal advice on your own offer.
How do I measure how much of my Black Friday was borrowed?
Forecast what you’d have sold from early November to the end of January with no sale. Use a year before you ran a sale if you have one, scaled by this year’s September and October growth, because last year’s pattern already contains last year’s borrowing. Add up the shortfall in the weeks around the sale, then divide it by the extra orders during the sale. A random holdout from your email and SMS offers gives you a cleaner read on those channels. If customers buy rarely, treat the result as a floor, since the shortfall can run past January.
Mikołaj Salecki
Writes about media, tech, and AI business for people who actually run digital. Former agency lead. Skeptic of frameworks that read better than they perform.
