7 stops and cuts in 4 brands · 19 budget-out days in 5 accounts · 101 campaign pauses in 16 brands · March to October 2026
01How we found the moments the ads stopped
Turning the ads off on purpose, to see what happens, is an expensive test, and few brands run it. So we went looking for the times it happened anyway in the accounts we manage. There are three kinds. Sometimes a whole account's ads stop or get cut deeply for days. Sometimes an account's budget runs out mid-day and the ads stay dark until midnight. And sometimes a single campaign gets paused.
For each one we compare the drop in total sales with the sales Amazon had credited to the ads before. We call that the lost share. 1 means every ad-credited sale was really lost. 0 means all of them would have happened anyway. Above 1 means total sales fell by more than the ads were credited with.
Total sales come from Amazon's sales and traffic report, per product and day. Ad spend and ad-credited sales come from the advertising reports. For the budget-outs we used the hourly spend and sales snapshots our app takes for every account. Section 11 has the details.
02The blackout we went looking for was a daily spending cap
We started from one case we thought we knew: a pet food brand that "stopped its UK ads" in September 2026. The data told a different story. The brand had only started selling on Amazon in August, and its UK ads started on September 8. From September 10 to 19 its total ad spend landed on the same amount every day, within 98 to 102% of it. Every campaign went dark at the same time each day, and that time kept moving earlier: about 16:00 on the first day, about 9:00 by the fourth.
That's a ceiling above the campaigns, most likely the new account's credit line. It lifted on the evening of September 20. Over the next two weeks the brand spent about 4 times as much a day.
The brand sells only in the UK, had no Amazon sales a year earlier and launched in the same weeks, so there was nothing to compare a blackout with. Its capped days went into the budget-out analysis in section 06 instead.
Show the numbers
| Day | Ad spend (cap = 100) | Capped | Ads dark from about |
|---|---|---|---|
| Sep 7 | 1 | ||
| Sep 8 | 19 | ||
| Sep 9 | 72 | ||
| Sep 10 | 99 | yes | 16:00 |
| Sep 11 | 100 | yes | 13:00 |
| Sep 12 | 100 | yes | 12:00 |
| Sep 13 | 98 | yes | 9:00 |
| Sep 14 | 99 | yes | 9:00 |
| Sep 15 | 103 | yes | 8:00 |
| Sep 16 | 100 | yes | |
| Sep 17 | 98 | yes | 9:00 |
| Sep 18 | 100 | yes | 8:00 |
| Sep 19 | 101 | yes | 9:00 |
| Sep 20 | 132 | ||
| Sep 21 | 458 | ||
| Sep 22 | 532 | ||
| Sep 23 | 570 | ||
| Sep 24 | 316 | ||
| Sep 25 | 328 | ||
| Sep 26 | 321 | ||
| Sep 27 | 390 |
03A third of the ad stops came with something else breaking
We scanned every account for stretches where ad spend fell under 30% of the previous two weeks for at least three days. We found 15. In 3 of them, a home textiles brand's whole store was offline in five marketplaces at once, including one where it ran no ads at all. In another, the products became unbuyable two days after the ads stopped. In a fifth, the two best-selling listings went dark while the ads were still running. Ads get switched off for a reason, and the reason moves sales on its own.
Another 4 stops had no sales data, and 1 was a client leaving. That left 5. A second scan for deep cuts, spend under half of normal for a week or more, found 3 more. Of these, 1 was usable. Of the other two, one brand cut its ads in five marketplaces during a dip that hit all of them about equally, and one ran out of stock. One more cut, to 44% of spend, came up while we checked the data.
- The whole store was offline−3
- Out of stock or unbuyable−1
- Main listings went dark−1
- No sales data−4
- Client leaving, data ends−1
- Clean enough to measure5
- Brand-wide dip in every marketplace−1
- Products out of stock−1
- Clean enough to measure1
Plus one cut to 44% that sits just above the line, found while checking the data.
- No hourly sales for that day−25
- Spend resumed late in the day−3
- Deal-event day−1
- Too few hours without ads−1
- Too few comparable days−4
- Clean budget-out days19
A stop counts once per brand and marketplace. Budget-out days need hourly sales data for that day and at least two comparable days with the ads running all day.
Show the numbers
| Group | Step | Days or events |
|---|---|---|
| Ad spend fell to under 30% of normal for 3 or more days | account stops found | 15 |
| The whole store was offline | −3 | |
| Out of stock or unbuyable | −1 | |
| Main listings went dark | −1 | |
| No sales data | −4 | |
| Client leaving, data ends | −1 | |
| clean enough to measure | 5 | |
| Ad spend fell to under 50% of normal for 7 or more days | more cuts found | 3 |
| Brand-wide dip in every marketplace | −1 | |
| Products out of stock | −1 | |
| clean enough to measure | 1 | |
| Budget ran out mid-day | days in accounts with hourly sales | 53 |
| No hourly sales for that day | −25 | |
| Spend resumed late in the day | −3 | |
| Deal-event day | −1 | |
| Too few hours without ads | −1 | |
| Too few comparable days | −4 | |
| clean budget-out days | 19 |
04A week or two without most of the ads lost about as much as the ads were credited with
That leaves seven stops and deep cuts in four brands, from 5 to 17 days. For each, we estimated what total sales would have been with the ads still on, in up to five ways: against the two weeks before, against the last week before (in case sales were already moving), against the same brand's other marketplaces where the ads didn't change, against other brands in the same marketplace, and, for one account, across its own products. No single comparison is perfect, so the chart shows each one.
In six of the seven, the lost share sits between 0.81 to 1.74. In 3 of them every comparison comes out above 1: the personal care tools brand in the US (1.02 to 1.49), the intimates brand in the US (1.18 to 1.38) and the home textiles brand in Italy (1.30 to 1.74). Another 3 sit around 1, from 0.81 to 1.11. The seventh, the home textiles brand in France, is lower, 0.49 to 0.81. We don't know why. Its sales were already sliding before the cut, and allowing for that pushes its number lower still.
Hollow circles compare with the last week before the stop, which follows sales that were already moving. The home and DIY tools brand in Germany cut its ads between two deal waves. We compare its two quieter stretches between the waves, with the ads cut and with them back, across its 89 product families (filled and hollow diamonds: the two orders of the comparison).
Show the numbers
| Event | Days | Ads left | Ads' share of sales | Branded | Two weeks before | Last week before | Other marketplaces | Other brands | Products | Sales | Sessions |
|---|---|---|---|---|---|---|---|---|---|---|---|
| Personal care tools · US | 12 | 44% | 67% | 18% | 1.09 | 1.02 | – | 1.49 | – | -36% | -33% |
| Intimates · US | 14 | 0% | 63% | 0% | 1.38 | 1.18 | – | 1.30 | – | -87% | -87% |
| Home textiles · IT | 13 | 1% | 57% | 3% | 1.30 | 1.49 | – | 1.74 | – | -74% | -68% |
| Home and DIY tools · DE | 6 | 56% | 31% | 1% | – | – | – | – | 0.93 / 1.02 | -15% | -12% |
| Home and DIY tools · IT | 17 | 20% | 28% | 0% | 0.94 | 1.11 | 0.96 | – | – | -15% | 0% |
| Home textiles · DE (March) | 5 | 15% | 57% | 1% | 1.07 | 0.81 | – | – | – | -51% | -59% |
| Home textiles · FR | 7 | 9% | 44% | 0% | 0.62 | 0.52 | 0.49 | 0.81 | – | -25% | -51% |
| Budget ran out mid-day (5 accounts) | 19 | 0% for the evening | 16 to 64% | 0.33 (-0.02 to 0.71) |
A comparison with the two weeks before assumes sales would have stayed at that level. Where they were already falling, it overstates the loss. Sales fell between the two halves of those weeks in 4 events: from 112 to 88 for the intimates brand, 115 to 90 for the home textiles brand in Germany, 104 to 96 in France and 102 to 98 for the personal care tools brand (index, the two weeks = 100). Comparing with the last week before instead brings them to 1.18 (from 1.38), 0.81 (from 1.07), 0.52 (from 0.62) and 1.02 (from 1.09). So for these four, the comparison with the two weeks before is likely too high. A straight-line trend doesn't settle it: for the intimates brand, a line through the two weeks before gives 0.20, and one through the four weeks before gives 1.58.
Stock and deal events can move sales too. The table below shows what we checked for each event. Amazon's inventory report starts in late July, so it covers 5 of the seven. Only the home textiles brand in Italy had more of its top sellers out of stock during the stop: 11% of product days before, 19% during. 2 of them, 15% of its sales before, had no stock for most of the stop and sold nothing. Counting all of their drop as stock rather than ads brings its lost share down to 1.04 to 1.40, still above 1. For the two events before the report starts, sessions fell along with sales (87% and 59%), which points to lost traffic rather than empty shelves.
Three events sit next to one of Amazon's deal events. The intimates brand's stop is the two weeks before Prime Day, when some shoppers wait. The personal care tools brand's cut ends the day before the October deal event, and the home and DIY tools brand in Italy cut its bids the day after it. The comparisons with other marketplaces and other brands in the same weeks take some of that out. A brand's own deals on a few products also sit in some windows, at 3% to 13% of sales. For the home and DIY tools brand in Germany they made 13% of sales with the ads cut and 9% with them back, which if anything props up the stretch without ads.
| Event | Sales in the weeks before, first half to second | Top sellers out of stock (share of product days) | Deal events nearby |
|---|---|---|---|
| Personal care tools · US | 102 to 98 | None out of stock | Ends the day before the October deal event. Own deals: 4% of sales in the weeks before |
| Intimates · US | 112 to 88 | No inventory data (June) | The two weeks before Prime Day |
| Home textiles · IT | 90 to 110 | 11% before, 19% during; 2 top sellers (15% of sales) empty most of the stop | Italy's mid-August holiday in the weeks before |
| Home and DIY tools · DE | Compared across products | 0.3% with ads back, 1.0% with ads cut | Between two of its own deal waves. Own deals: 9% of sales with ads back, 13% with ads cut |
| Home and DIY tools · IT | 97 to 103 | 0.5% during, no data before | Starts the day after the October deal event. Own deals: 10% of sales before, 3% during |
| Home textiles · DE (March) | 115 to 90 | No inventory data (March) | None |
| Home textiles · FR | 104 to 96 | 7% before, 8% during | None |
- Sales in the weeks before, first half to second
- 102 to 98
- Top sellers out of stock (share of product days)
- None out of stock
- Deal events nearby
- Ends the day before the October deal event. Own deals: 4% of sales in the weeks before
- Sales in the weeks before, first half to second
- 112 to 88
- Top sellers out of stock (share of product days)
- No inventory data (June)
- Deal events nearby
- The two weeks before Prime Day
- Sales in the weeks before, first half to second
- 90 to 110
- Top sellers out of stock (share of product days)
- 11% before, 19% during; 2 top sellers (15% of sales) empty most of the stop
- Deal events nearby
- Italy's mid-August holiday in the weeks before
- Sales in the weeks before, first half to second
- Compared across products
- Top sellers out of stock (share of product days)
- 0.3% with ads back, 1.0% with ads cut
- Deal events nearby
- Between two of its own deal waves. Own deals: 9% of sales with ads back, 13% with ads cut
- Sales in the weeks before, first half to second
- 97 to 103
- Top sellers out of stock (share of product days)
- 0.5% during, no data before
- Deal events nearby
- Starts the day after the October deal event. Own deals: 10% of sales before, 3% during
- Sales in the weeks before, first half to second
- 115 to 90
- Top sellers out of stock (share of product days)
- No inventory data (March)
- Deal events nearby
- None
- Sales in the weeks before, first half to second
- 104 to 96
- Top sellers out of stock (share of product days)
- 7% before, 8% during
- Deal events nearby
- None
Stock comes from Amazon's inventory report: the share of product days with nothing in stock. The buy box share can stay high while a product is out of stock, so we don't use it for this.
All seven accounts bought mostly generic searches. Between 0 and 18% of their ad-credited sales came from people searching the brand's name. Ads were also a big part of their sales before the stop, 28 to 67%. Most are small accounts, from 5 to about 200 units a day, plus one at about 1,000.
The timelines below show what happened day by day, with every series as an index of the weeks before (100).
Hover a day for the numbers. The home and DIY tools brand in Italy cut its bids, not its campaigns; its deal weeks are left out of the comparison.
Show the numbers
| Event | Comparison window | Ads off or cut | Sales change | Sessions change | Units a day before |
|---|---|---|---|---|---|
| Personal care tools · US | Sep 10 to Sep 23 | Sep 24 to Oct 5 | -36% | -33% | 193.6 |
| Intimates · US | May 20 to Jun 2 | Jun 9 to Jun 22 | -87% | -87% | 8.4 |
| Home textiles · IT | Aug 6 to Aug 19 | Aug 20 to Sep 1 | -74% | -68% | 5.4 |
| Home and DIY tools · DE | Sep 15 to Sep 20 | Aug 25 to Aug 30 | -15% | -12% | 997 |
| Home and DIY tools · IT | Sep 8 to Sep 21 | Oct 8 to Oct 24 | -15% | 0% | 72.6 |
| Home textiles · DE (March) | Mar 4 to Mar 16 | Mar 17 to Mar 21 | -51% | -59% | 19.7 |
| Home textiles · FR | Aug 17 to Aug 30 | Aug 31 to Sep 6 | -25% | -51% | 35.9 |
05The sales came back when the ads came back
Two accounts let us watch the other direction. The home textiles brand started ads in Italy in August, after weeks without any. Its sales went from 20 to 100 (the weeks with ads), fell back to 26 when the ads went off, and climbed to 57 when the ads came back at about half the spend (46%). The personal care tools brand in the US cut its Sponsored Products spend to 17% for 12 days and kept its Sponsored Brands and Display ads running, 44% of its ad spend in all. Its sales fell to 64, and returned to 97 when the spend did.
The personal care tools brand's October event days (October 6 and 7) are left out of both phases.
Show the numbers
| Account | Phase | Total sales (ads on = 100) | Ad spend (ads on = 100) |
|---|---|---|---|
| Home textiles · IT | No ads | 20.1 | 0 |
| Ads on | 100.0 | 100 | |
| Ads off | 25.6 | 1 | |
| Ads back | 57.2 | 46 | |
| Personal care tools · US | Ads on | 100.0 | 100 |
| Ads cut | 63.5 | 45 | |
| Ads back | 97.1 | 94 |
06When the budget ran out mid-day, a third to a half was lost that evening
Our app takes a snapshot of every account's ad spend each hour. On some days the spend goes flat in the afternoon and stays flat until midnight: the account's budget has run out and every ad is dark. We found 19 such days, clean enough to use, in 5 accounts of 4 brands. Of these, 9 are the pet food brand's capped days from section 02.
For each one we compared the sales in the dark hours with the same hours on days the ads ran all day, same account, same weekday, within four weeks. We scaled by the sales in the hours before the cut. Pooled over the 19 days, the evening lost 0.33 of what Amazon would have credited to the ads in those hours. The 95% interval runs from −0.02 to 0.71.
The placebo takes a normal day and treats it as if the budget had run out at the same hour. It lands near 0, as it should.
Show the numbers
| Row | Result | 95% interval | Days | Accounts |
|---|---|---|---|---|
| Pet food · UK (46% of the weight; without it 0.43) | 0.22 | 9 | 1 | |
| Skincare · ES (21% of the weight; without it 0.33) | 0.34 | 4 | 1 | |
| Skincare · FR (21% of the weight; without it 0.40) | 0.07 | 2 | 1 | |
| Women's health gummies · US (11% of the weight; without it 0.21) | 1.31 | 1 | 1 | |
| Intimates · US (1% of the weight; without it 0.33) | 0.54 | 3 | 1 | |
| Main: 19 budget-out days | 0.33 | -0.02 to 0.71 | 19 | 5 |
| Comparison days spending 0.5 to 3× the event day | 0.43 | -0.07 to 0.96 | 10 | 4 |
| Comparison days spending 0.75 to 2× the event day | 0.45 | -0.07 to 1.00 | 7 | 3 |
| Hours with ads held at least a quarter of a normal day | 0.44 | -0.03 to 0.95 | 12 | 5 |
| Report delay of 1 hour instead of 2 | 0.35 | 0.03 to 0.72 | 19 | 5 |
| Report delay of 3 hours | 0.26 | -0.26 to 0.76 | 19 | 5 |
| Report delay estimated per account | 0.36 | 0.03 to 0.73 | 19 | 5 |
| Including the last, partial hour of the day | 0.36 | -0.01 to 0.74 | 19 | 5 |
| Missed hours valued at the budget-out day's own return on ad spend | 0.48 | -0.03 to 0.91 | 19 | 5 |
| Placebo: a normal day treated as a budget-out | -0.16 | -0.70 to 0.30 | 14 | 5 |
The answer moves with the choices behind it, from 0.26 to 0.48 across the variants, and from 0.21 to 0.43 when we leave one account out. The pet food brand's comparison days spent 3 to 5 times as much as its capped days. Using only comparison days with a similar spend leaves it out and gives 0.43 on 10 days. Our spend snapshots run one to three hours behind real time, depending on the account, so the exact hour the ads went dark is an estimate. And valuing the missed hours at the budget-out day's own return on ad spend, instead of the comparison days', gives 0.48.
Two brands carry most of the result: the pet food brand's 9 capped days make up 46% of the weight, and the skincare brand's 6 days in Spain and France 43%. Resampling whole brands instead of days, the 95% interval runs from 0.21 to 1.25. So the most we can say is a third to a half, with a wide range.
These accounts had more of their ad sales on brand-name searches than the stops in section 04: 16% for the pet food brand on its capped days, 28% for women's health gummies and 61 to 64% for the skincare brand in Spain and France. And this number only counts the same evening. Purchases that the missing clicks would have brought on the following days aren't in it, so the full loss is larger.
07Organic search didn't make up for it
If organic search took over when the ads stopped, product page sessions would hold up. They didn't. During the stops and cuts, against the weeks before, sessions fell 87% for the intimates brand, 68% and 59% for the home textiles brand in Italy and Germany, 51% in France and 33% for the personal care tools brand. They fell 12% or less for the home and DIY tools brand, whose sales also fell the least, 15% in both its marketplaces.
For the intimates brand we also had daily organic rank on 116 product and keyword pairs. The rank slipped a median 22 positions while the ads were off and hadn't recovered two weeks later. Its ads only came back at a sixth of the old spend. For the home textiles brand the ranks didn't move, but most of them were deep, past page 3, where a move makes little difference.
08Pausing a campaign rarely switches a product's ads off
Campaign pauses are the most common ad change of all: 8,752 in our data since June. We hoped they would let us compare brand-name and generic keywords. They mostly can't, because a pause rarely stops the ads for a product. We took the 348 Sponsored Products pauses with real spend before (at least 60 euros over two weeks) that stayed paused for three days or more. They fall into 111 groups of same-day pauses, and 101 had product-level ad data.
The paused campaigns carried a median 12% of their products' ad-credited sales. The products kept the rest through their other campaigns. The other campaigns didn't pick up what the paused ones had carried: they changed by a median −41% of it, with a very wide spread. When the paused campaigns carried half or more (15 pauses), the products' ad-credited sales fell to a median 0%.
Dots on the dashed line kept exactly what the remaining campaigns had carried before. Only Sponsored Products has product-level ad data.
Show the numbers
| Group | Pauses | Median ad-credited sales kept |
|---|---|---|
| Paused campaigns carried less than half | 86 | 88% |
| Paused campaigns carried half or more | 15 | 0% |
| All pauses | 101 |
After we left out pauses near deal events, stock-outs, price changes and missing data, only 7 groups in 2 brands were left to measure total sales against. That's too few to split by keyword type, so we don't report it.
09What we can't tell from this data
- Ads on the brand's own name. Brand-name ads are the ones most often called a waste: the shopper was looking for the brand anyway, so the organic listing would have caught the sale. Our data can't confirm or rule that out. None of the seven stops was in an account that mostly bought brand-name searches, and the pauses were too few. The closest case was a skincare brand in Spain, where 57% of the ad-credited sales came from brand-name searches in the weeks before. In July it cut its ad spend by 71%. Of the ad-credited sales it lost, 66% were generic, and the cut came during a dip that hit all five of its marketplaces about equally, whatever their own cut. So it can't answer the question either.
- More than two weeks without ads. Our stops lasted 5 to 17 days. Whether sales find their way back after a month or two without ads, or keep sliding as organic rank slips, this data can't say.
- The days after a budget-out. The budget-out result counts the same evening only. Clicks the evening missed may have turned into purchases later, so the full loss is larger by an amount we can't measure.
- Bigger accounts. Six of the seven events were in accounts selling 5 to about 200 units a day, the seventh about 1,000. They come from 4 brands, one category each. A large brand with strong demand for its own name may lose less when its ads stop.
- Why three events landed above 1. In 3 events total sales fell by more than the ads were credited with, in every comparison. That can mean the ads helped sales Amazon didn't credit to them, through organic rank or shoppers who came back later. It can also be the noise of small accounts over two weeks. Seven events can't tell those apart.
- One result we withdrew. For a while it looked as if a mostly brand-name account had lost nothing in six days without ads. That turned out to be two brands' sales data swapped at the source. We found it in our independent re-check, and it's not in any number on this page.
10What to do with this
- 01
Don't switch off generic ads expecting the sales to stay. In six of the seven stops and cuts we could measure, total sales fell by about as much as Amazon credited to the ads, and in 5 of them sessions fell by a third or more.
- 02
Check the store before you blame the ads. A third of the account stops in our data came with a store outage, a stock-out or dark listings. Look at stock in the inventory report, sessions and deals for the same days. The buy box share can stay high while a product is out of stock.
- 03
Let budgets run out late, if at all. An evening without ads lost a third to a half of what the ads would have been credited with, before counting the next day. A budget that runs out after a morning rush also misses the evening.
- 04
Before pausing a campaign, look at its share of each product's ads. Most paused campaigns carried a small part. When one carried half or more of a product's ad-credited sales, those sales went to zero and nothing else picked them up.
- 05
Test brand-name ads on their own. Our data can't say how much of them is incremental. Cut them in one marketplace or for some product families, keep the rest running, and compare total sales for two weeks.
- 06
Expect the sales to come back with the ads. In both accounts where the ads returned, total sales came back roughly in step with the spend.
11How we measured it
- Data. Amazon's sales and traffic report per product and day (total sales, units, sessions, buy box), the advertising reports per campaign, product and search term, our app's hourly snapshots of each account's ad spend and total sales, Amazon's inventory report, deal and promotion records, and Helium 10 organic ranks. March to October 2026, accounts we manage in the US, UK and EU.
- Lost share. The drop in total sales against what they would have been with the ads on, divided by the ad-credited sales lost. For a full stop that's the ad-credited sales before; for a cut, the drop in them. All per day. Money is never shown; where we pool accounts, euros at fixed rates (a dollar at 0.86, a pound at 1.15).
- Stops and cuts. Ad spend under 30% of the previous 14 days' median for 3 or more days, or under 50% for 7 or more. We dropped stops with a store outage, dark listings or missing data in their windows, and left Amazon's deal-event days out of every window. A brand's own deals on a few products stay in some windows (3% to 13% of sales). Stock is the share of days the products making 80% of sales had nothing in stock, from Amazon's inventory report, which starts in late July.
- Comparisons. The two weeks before; the last week before, which follows sales that were already moving; the same brand's other marketplaces, used only where their own ad spend moved by less than 20%; the median of all other brands we manage in the same marketplace; and for one account, a comparison across its 89 product families that lets the whole account rise or fall together. The same design on the personal care tools brand gives 1.57 on 17 families, too uncertain to use (−1.41 to 4.56). None of the year-earlier windows had data, and straight-line trends changed too much with the number of weeks they were fitted to.
- Budget-outs. A day where the hourly spend stops rising by 18:00 and stays flat, and the day's final spend matches. The snapshots run one to three hours behind real time, depending on the account. The main estimate assumes two hours, the variants one, three and a delay per account, and we leave out the hours around the cut. The lost ad-credited sales are the spend the dark hours would have had on normal days, at those days' return on ad spend. Intervals resample days. Resampling whole brands gives 0.21 to 1.25.
- Pauses. Every campaign that went from enabled to paused for at least three days. A product's ad-credited sales add up all its Sponsored Products campaigns. Product families are parent products.
- Brand-name share. The share of ad-credited sales in the search term report where the search contains the brand's name, a misspelling or a translation of it.
- Re-check. A second analysis, written separately on the raw tables, checked the numbers in four rounds. It caught two snapshot glitches counted as budget-outs, comparison marketplaces that had cut their own ads, a currency mix-up, offboarding gaps read as sales drops, a missed stop and two brands with swapped sales data. After we published, a review from another of our studies found three more problems: we had used the buy box share as a stock check, we had described two windows as away from deal events when they sit next to one, and sales were already falling before four of the events. The fourth round checked those fixes. Everything on this page includes them.