106 deals · 5 brands · 6 marketplaces · Oct 6 to 7, 2026 · against same-brand products without a deal · data through Oct 31, 2026
01On the deal days, a product sold 3.5 times what it would have sold anyway
For each deal we took the product's usual daily sales, from 35 to 8 days before the deal, and asked what it would have sold on the deal days without a deal. The event lifts some products without a deal, so the yardstick is the same brand's products in the same marketplace that had no deal. If they sold a fifth more on the event days, we expect the deal product to sell a fifth more too, and count everything above that as the deal's work.
The typical deal sold 3.5 times that (90% interval 3.1 to 4.2, 106 deals). Over the two days that is 6.4 days of normal sales. Bigger products gained a little less for their size: weighted by units, the gain is 5.5 days. Most of it was traffic. Sessions were 2.8 times the expected number, and the share of sessions that ended in an order rose 26%.
Usual level = each product's average from 35 to 8 days before the deal. Data through Oct 31, 2026. Hover for the values per day.
Show the numbers
| Day | With a deal | Without a deal | Deals |
|---|---|---|---|
| Sep 15 | 107 | 129 | 106 |
| Sep 16 | 100 | 113 | 106 |
| Sep 17 | 96 | 100 | 106 |
| Sep 18 | 87 | 84 | 106 |
| Sep 19 | 87 | 90 | 106 |
| Sep 20 | 104 | 98 | 106 |
| Sep 21 | 99 | 104 | 106 |
| Sep 22 | 99 | 92 | 106 |
| Sep 23 | 93 | 106 | 106 |
| Sep 24 | 96 | 85 | 106 |
| Sep 25 | 85 | 96 | 106 |
| Sep 26 | 78 | 68 | 106 |
| Sep 27 | 106 | 93 | 106 |
| Sep 28 | 103 | 108 | 106 |
| Sep 29 | 95 | 85 | 106 |
| Sep 30 | 104 | 98 | 106 |
| Oct 1 | 98 | 100 | 106 |
| Oct 2 | 88 | 90 | 106 |
| Oct 3 | 87 | 85 | 106 |
| Oct 4 | 104 | 88 | 106 |
| Oct 5 | 105 | 110 | 106 |
| Oct 6 | 444 | 129 | 106 |
| Oct 7 | 467 | 123 | 106 |
| Oct 8 | 91 | 84 | 106 |
| Oct 9 | 78 | 78 | 106 |
| Oct 10 | 82 | 79 | 106 |
| Oct 11 | 100 | 99 | 106 |
| Oct 12 | 87 | 98 | 106 |
| Oct 13 | 92 | 85 | 106 |
| Oct 14 | 91 | 87 | 106 |
| Oct 15 | 87 | 84 | 106 |
| Oct 16 | 79 | 68 | 106 |
| Oct 17 | 73 | 61 | 106 |
| Oct 18 | 83 | 92 | 106 |
| Oct 19 | 89 | 85 | 106 |
| Oct 20 | 80 | 79 | 106 |
| Oct 21 | 88 | 88 | 106 |
| Oct 22 | 82 | 58 | 106 |
| Oct 23 | 68 | 70 | 106 |
| Oct 24 | 65 | 66 | 106 |
| Oct 25 | 102 | 93 | 106 |
| Oct 26 | 90 | 58 | 106 |
| Oct 27 | 76 | 79 | 106 |
| Oct 28 | 86 | 71 | 106 |
Every number on this page needs one caveat. 76 of the 106 deals (72%) are one skincare brand, in Germany, the UK, Italy, Spain and France, and its deals lifted sales 4.1 times the expected units. Without it the typical lift is 2.0 (30 deals). A home and DIY tools brand's deals did less (1.73 times), and so did one US brand's: its 8 deals, all one product family, a median 19% off like the skincare brand's 19%, lifted sales 1.7 times. That is one product, so read it as a hint that brand matters too. The 106 deals are 49 different products, because the same product in several countries counts several times. Counted by product, the interval for the typical lift is 2.9 to 4.3; resampling the five brands gives 1.7 to 4.2. The table has each brand on its own.
| Brand | Event | Deals | Median lift | Extra days, median |
|---|---|---|---|---|
| Skincare brand | Oct 2026 | 76 | 4.1× | +8.7 |
| Home and DIY tools brand | Oct 2026 | 12 | 1.7× | +2.1 |
| Other brands | Oct 2026 | 18 | 2.4× | +4.0 |
| Skincare brand | Prime Day 2026 | 57 | 3.3× | +12.2 |
| Home and DIY tools brand | Prime Day 2026 | 16 | 2.1× | +2.9 |
| Other brands | Prime Day 2026 | 18 | 1.5× | +2.3 |
02We found no give-back in the weeks around the deal
A deal could just move sales around. Shoppers might hold off in the week before, then stock up and buy less for weeks after. If that happened, the deal products would dip before the event and sag after it, against products without a deal.
We saw no sign of it. Over the week before and the three weeks after, the deal products came out +4.5 days of sales ahead of the yardstick (90% interval +2.0 to +6.6). We checked that against fake dates: the same products six weeks earlier, when nothing happened. 65 of the deals are clean on both dates. For those, the deal products came out +3.7 days ahead of their siblings around the fake date and +3.2 around the real one, so the gap between the two is −0.4 days (−4.8 to +3.6). The other 41 deals have no fake-date twin and came out +6.4 days around the real date. Across all 91 deals that are clean on the fake dates, the weeks around the fake date average −0.1 days. So the evidence is mixed: the gap between real and fake dates is small, but it rests on the 65 deals that can be compared, and the 41 that can't came out further ahead.
Our best estimate is that close to none of the deal-day gain was borrowed from the weeks around, but that rests on 65 deals, and the data can't rule out that up to about 61% of it was.
Whisker: the deal-day estimate plus the 90% interval of the weeks around the deal.
Show the numbers
| Compared with | Deal days | Week before | 3 weeks after | All 5 weeks | Deals |
|---|---|---|---|---|---|
| Products without a deal | +8.4 | +1.2 | +3.3 | +12.9 | 106 |
| Same products, fake dates | +7.8 | −0.3 | −0.1 | +7.3 | 65 |
For Prime Day we can't say. The answer depends on which weeks count as the usual level. With the four weeks before, the weeks around Prime Day come out at +5.8 days. With eight weeks, they come out at −4.7.
03Products without a deal sold 28% more on the event days, and it differs by brand
The same brands' products without a deal sold 28% more per day than usual on October 6 and 7 (median of 124 products). The 89 of them that are also clean on the fake dates changed by +29% on the real dates and by +1% on the fake ones. It differs by brand: +35% for the skincare brand's 51 products, +28% for the other brands' 62 and −3% for the home and DIY tools brand's 11. Weighted by units sold the rise is 6%. Some of what a deal product sells during the event would have come anyway, and that part is what the yardstick takes out.
We left out comparison products that cut their price without a listed deal, because a hidden discount lifts their sales and makes the deal look weaker than it was. Of the 180 products without a deal, in the brands and countries where a deal ran, that passed our other checks in October, 52 (29%) sold at a price more than 5% below their usual one on the event days. On the fake dates it was 5%. For Prime Day it was 23%, against 4%. Some hidden discounting may remain below the 5% line.
On Prime Day the typical product without a deal sold 9% more per day (median of 197 products; +14% for the skincare brand's, +8% for the other brands', +4% for the home and DIY tools brand's). Weighted by units, it was 0% more.
04Deeper discounts went with bigger lifts
The discount here is the price shoppers paid on the deal days against what they paid in the weeks before. By that measure the typical October deal was 19% off.
The lift rose with the discount in both events (rank correlation 0.50 in October, 0.63 on Prime Day). Brands give deeper discounts on different products, so we also looked inside one brand. The skincare brand's own deals climb the same way, from 1.7× at under 10% off (4 deals) to 7.2× at 25%+ off (7 deals).
Discount = price paid on the deal days (sales divided by units, VAT included) against the same in the weeks before.
Show the numbers
| Discount | Oct median | Oct deals | Prime Day median | Prime Day deals |
|---|---|---|---|---|
| Under 5% off | 1.6× | 5 | 1.4× | 14 |
| 5 to 15% off | 2.8× | 33 | 1.9× | 26 |
| 15 to 25% off | 4.1× | 60 | 3.3× | 40 |
| 25%+ off | 7.0× | 8 | 4.1× | 11 |
| Discount, one skincare brand, Oct | Median lift | Deals |
|---|---|---|
| Under 10% | 1.7× | 4 |
| 10 to 15% | 3.2× | 21 |
| 15 to 20% | 3.8× | 17 |
| 20 to 25% | 4.6× | 27 |
| 25%+ | 7.2× | 7 |
Shallow deals lifted less. The home and DIY tools brand's measurable deals were a median 7% below the price its shoppers usually paid, and they lifted sales 1.73× in October (12 deals) and 2.14× on Prime Day (16 deals), against 4.1× and 3.3× for the skincare brand's deeper cuts. The two brands also sell different things, so treat this as a pattern.
05Ad orders on deal products got cheaper, but for the typical product the return didn't improve
We looked at Sponsored Products for the same products. On the October deal days, deal products paid 20% more per click, but their ads converted 31% better, so each ad order cost 14% less (medians over 76 products with enough clicks). Spend per day was 3.3 times the usual. On Prime Day it looked the same: clicks 26% dearer, conversion 51% better, each ad order 12% cheaper (66 products). The return on ad spend barely moved, because the price was lower too: the median rose 1% in October and 3% on Prime Day. Pooled over all products, October's return was up 27%, but one product was 51% of that spend.
Products without a deal had a harder Prime Day. Their cost per click rose 14%, their conversion fell 21%, so each ad order cost 42% more and the return on ad spend fell 44% (65 products). In October the picture is mixed: the typical product without a deal paid 1% more per ad order, and all of them together paid 19% more (53 products).
Amazon credits an ad with orders up to 7 days after the click, including other products. On deal days shoppers buy several deal items, so ad conversion here runs high.
Show the numbers
| Event and group | CPC | Conversion | Cost per order | ROAS | Products |
|---|---|---|---|---|---|
| Oct 2026, with a deal | +20% | +31% | −14% | +1% | 76 |
| Oct 2026, without a deal | +6% | −6% | +1% | −11% | 53 |
| Prime Day, with a deal | +26% | +51% | −12% | +3% | 66 |
| Prime Day, without a deal | +14% | −21% | +42% | −44% | 65 |
06Whether a deal pays depends on your margin
We can't publish our clients' product costs, so we ran the sums for three margins. Margin here means what's left of the net price (without VAT) after product cost and Amazon's fees, before ads.
A deal costs you margin on every unit you would have sold anyway and earns margin on the extra units. Amazon's referral fee and VAT shrink with the price, so a 20% discount takes a bit less than 20% off your margin. At the typical October discount, a deal needs 2.1 times the expected units to break even at a 30% margin, 1.7 times at 40% and 1.5 times at 50%.
Then we counted, deal by deal, which came out ahead on the deal days once the extra ad spend was paid. In October that was 56% of deals at a 30% margin, 67% at 40% and 80% at 50% (106 deals). On Prime Day it was 65%, 80% and 82% (82 deals with ad data). The two events have different brands and different amounts of ad data, so don't read the gap as a difference between the events. These are the deals that passed our checks, at a margin we assumed. Not in the sums: Amazon's deal fee, and whatever a new customer buys later.
Our free Deal & Coupon Profit Calculator runs the same sum with your own price, costs and discount.
If the referral fee stayed fixed per unit, 62% of October deals would come out ahead at a 40% margin instead of 67%.
07Prime Day 2026 showed a similar spike over four days
Prime Day ran from June 23 to 26, 2026. On the 91 deals we can measure, the deal days sold 2.7 times the expected units (90% interval 2.3 to 3.4; 2.4 to 3.0 across our checks), spread over four days. That is lower than October's 3.5, though the intervals overlap. The typical deal gained 7.2 days of normal sales. The skincare brand is 63% of these deals too.
Show the numbers
| Day | With a deal | Without a deal | Deals |
|---|---|---|---|
| Jun 2 | 105 | 98 | 75 |
| Jun 3 | 100 | 110 | 91 |
| Jun 4 | 97 | 83 | 91 |
| Jun 5 | 97 | 112 | 91 |
| Jun 6 | 75 | 97 | 91 |
| Jun 7 | 101 | 110 | 91 |
| Jun 8 | 90 | 122 | 91 |
| Jun 9 | 104 | 111 | 91 |
| Jun 10 | 101 | 105 | 91 |
| Jun 11 | 102 | 94 | 91 |
| Jun 12 | 90 | 97 | 91 |
| Jun 13 | 90 | 86 | 91 |
| Jun 14 | 102 | 102 | 91 |
| Jun 15 | 111 | 98 | 91 |
| Jun 16 | 98 | 90 | 91 |
| Jun 17 | 100 | 96 | 91 |
| Jun 18 | 80 | 94 | 91 |
| Jun 19 | 75 | 76 | 91 |
| Jun 20 | 71 | 88 | 91 |
| Jun 21 | 80 | 98 | 91 |
| Jun 22 | 86 | 97 | 91 |
| Jun 23 | 392 | 112 | 91 |
| Jun 24 | 217 | 102 | 91 |
| Jun 25 | 233 | 88 | 91 |
| Jun 26 | 267 | 110 | 91 |
| Jun 27 | 80 | 80 | 91 |
| Jun 28 | 87 | 86 | 91 |
| Jun 29 | 91 | 98 | 91 |
| Jun 30 | 98 | 112 | 91 |
| Jul 1 | 90 | 86 | 91 |
| Jul 2 | 98 | 90 | 91 |
| Jul 3 | 90 | 83 | 91 |
| Jul 4 | 71 | 91 | 91 |
| Jul 5 | 97 | 81 | 91 |
| Jul 6 | 93 | 90 | 91 |
| Jul 7 | 98 | 98 | 91 |
| Jul 8 | 87 | 88 | 91 |
| Jul 9 | 88 | 77 | 91 |
| Jul 10 | 76 | 103 | 91 |
| Jul 11 | 71 | 101 | 91 |
| Jul 12 | 112 | 96 | 91 |
| Jul 13 | 102 | 90 | 91 |
| Jul 14 | 88 | 92 | 91 |
| Jul 15 | 64 | 79 | 91 |
| Jul 16 | 74 | 82 | 91 |
| Jul 17 | 65 | 72 | 91 |
08What we'd take into Black Friday
- 01
Deeper discounts went with bigger lifts. Deals under 5% below the price shoppers usually paid sold 1.6× times the expected units (5 deals), deals 15 to 25% off 4.1× times (60 deals). Brands differ too, so read it as a pattern. Measure the discount against what people actually paid in the weeks before. A list price can make a small cut look big.
- 02
Do the break-even sum before you book. At a 40% margin and the typical 19% off, a deal needs about 1.7 times its expected units on the deal days. 67% of our October deals came out ahead once the extra ads were paid, and 80% of the Prime Day deals.
- 03
Expect no hangover, but check. We found no sign that the weeks before and after the deal gave anything back in October, but the data can't rule out that up to about 61% of the gain was borrowed. Watch your own weeks after a deal.
- 04
Expect cheaper orders, not a better return. Deal products' ad orders got 14% cheaper in October, but the price was lower too, so the typical product's return on ad spend barely moved (up 1%). Pooled over all products it was up 27%, with one product at 51% of the spend. On Prime Day, ads for products without a deal got 42% more expensive per order.
We wanted to replay Black Friday 2025 as well. Our sales data for it has gaps, and only 1 deals pass every check, so these points come from October and June 2026.
09What we couldn't measure
The 2025 events. For Black Friday 2025 and Prime Big Deal Days 2025 our sales data for the main brands has gaps. 1 and 0 deals pass the checks.
Deal types. The deals we follow are Best Deals. 3 Lightning Deals started around the October event (October 4 to 9), too few to compare, and products with one are out of the sample.
Organic rank. Only 8 to 14 deal products per event have rank tracking, all from one brand, mostly with one or two keywords each. One product moves the medians by several positions, so we make no claim about rank.
New-to-brand customers. Only one brand and marketplace has the weekly Amazon Marketing Cloud data. There, new-to-brand orders rose 86% in the deal week, in step with all orders, so the share stayed about the same (77% before, 75% in the deal week, 17 products).
Running out of stock was rare. The check in our funnel is the Buy Box: of the deals that sold enough and had the Buy Box before, 7 of 203 lost it during or after the deal (and are out), against 6 of 345 products without a deal. Amazon's inventory report covers 98 of the 106 deals we kept: 5 of them had zero stock on a day within three weeks after the deal. For the 93 with snapshots on the deal days themselves, none had zero stock on a deal day.
Who this describes. Established products of brands with large catalogues, mostly one skincare brand in Europe. Of 687 October deals, 106 pass every check. They are 49 different products.
What the comparison assumes. The products without a deal belong to brands that also ran deals. If a brand's deals lifted its other products, our lifts are too low. If they pulled shoppers away from them, our lifts are too high. The screen sees price cuts of more than 5%, so smaller hidden discounts would make the lifts too low. And the screen matters: without it October's lift would be 3.9 and Prime Day's 2.2.
10How we measured it
- Data. Amazon's sales and traffic report per product and day (units, sessions, sales, buy box), the deal dates from Amazon's promotion report, the Sponsored Products advertised product report, and the live deal badge Amazon shows on a product page. Data through Oct 31, 2026.
- Windows. Deal days as Amazon scheduled them, in local time. Usual level: days 35 to 8 before the deal. Week before: days 7 to 1. Then the three weeks after the last deal day.
- Which deals count. Amazon didn't cancel the deal. The product sold at least 14 units in the four weeks of the usual level, had the Buy Box before the deal, had no other deal or coupon from five weeks before to three weeks after, didn't sell more than 5% below its usual price in the week before (so no price cut that the deal reports don't show), and had sessions on the deal days (4 deals had none and are out). Its brand had at least 3 products in the same marketplace without a deal, selling at least 3 units a day together, and those products didn't sell more than 5% below their usual price on the event days either. The deal product kept the Buy Box during the deal and after it (a week under half the Buy Box, or two unbuyable days in a row, counts as lost). A coupon that ran through the whole period counts as part of the normal price.
- Yardstick. Expected units = the deal product's usual level times how much the brand's products without a deal moved on the same days. Lift = actual units / expected. Days of normal sales = extra units / usual daily units.
- Checks. The same sums on fake dates six weeks (October) and five weeks (Prime Day) earlier, where they should find nothing. They find a median lift of0.8 and 0.8 (a shortfall of 18% and 17%), but about 0 extra days of sales on average (+0.1 in October's sample, −0.8 in Prime Day's). The median lift reads low when nothing happened, so our typical lifts are probably not too high. Then variations: a shorter and an eight-week usual level, leaving out comparison products that share a parent listing with a deal product, leaving out groups where one product is more than half of the yardstick, leaving out deals with under 3% realised discount and, in October, deals that ran out of stock. October's lift is 3.3 to 3.6 across them, Prime Day's 2.4 to 3.0. Prime Day's weeks around the deal flip sign, so we don't report those.
- Discount and profit. Discount = price paid on the deal days (sales / units, VAT included) against the same in the usual period. Profit uses margin on the net price, product cost and FBA fee fixed per unit, VAT and a 15% referral fee moving with the price. Extra ad spend = deal-day spend minus the usual spend, adjusted for how the brand's other products' spend moved.
- Intervals. 90% bootstrap intervals, resampling the deals and their comparison products within each brand and marketplace. Counting each product once (the same product in several countries, or variations of one parent, is one), there are 49 products in October and 43 on Prime Day, and the typical lift's interval is 2.9 to 4.3 in October and 2.1 to 3.2 on Prime Day. Resampling the five brands gives 1.7 to 4.2 and 1.5 to 3.2.
- Re-check. An independent re-check recomputed the numbers on this page from the raw data, in four rounds, without seeing our code. The counts of products with rank tracking were not re-checked in the last round. The first rounds found mistakes in dates, thin comparison groups and our profit sums. The later ones found the bigger problem: 29% of the products we first used as comparisons without a deal had cut their price anyway, in ways no deal report lists, and some deal products had cut theirs the week before. We now screen for that, which changed the typical lifts and what we say about the home and DIY tools brand's deals. They also found a canceled deal counted as a deal and a stock check that was really a Buy Box check, which we fixed, and asked us to count products rather than deals when we say how sure we are.
- Limits. 106 deals in one event, 72% of them one skincare brand. These are brands we manage, so our own ad decisions are part of what happened. Price cuts that only Prime members get are in no report we have. We see them only in the prices shoppers paid, so the 5% screen catches the ones that move the average price that much and misses smaller ones.