American ecommerce sales dropped 2.2 percent in July against June, and almost every reading of that number you encounter this week will be wrong. July online sales were 136.9 billion dollars, and they were up 7.7 percent against July last year. The fall is month against month, and it has a single, boring, entirely knowable cause: Amazon moved its summer Prime Day out of July and into June.
Prime Day ran from 23 to 26 June and did 26.4 billion dollars. June ecommerce grew 14.2 percent year on year. July, deprived of the event that normally sits inside it, grew 7.7 percent instead. Nothing about consumer demand changed between those two months. The calendar did.
This matters to an operator for a reason that has nothing to do with the American market. It is the cleanest available example of a distortion that is about to run through every year-on-year comparison you make for the next twelve months, including your own.
What the figures actually say
| Figure | What it measures |
|---|---|
| 136.9 billion dollars | US retail ecommerce sales, July 2026 |
| Up 7.7 percent | July 2026 against July 2025, year on year |
| Down 2.2 percent | July 2026 against June 2026, month on month |
| Up 14.2 percent | June 2026 year on year, the month that took the event |
| 26.4 billion dollars | Prime Day, 23 to 26 June 2026 |
| 763.6 billion dollars | Total US retail sales, July 2026, against 768.07 billion in June |
Two of those six numbers are growth and four are context, and the two that are growth both point up. July online sales were the lowest since March in absolute terms, and they were still 7.7 percent bigger than the same month a year earlier. Both statements are true at once, and which one you lead with decides whether your reader panics.
Why the timing shift is not a footnote
Amazon has held its summer Prime Day in July almost every year. The exceptions before this one were 2020 and 2021, and both had obvious pandemic explanations. So 2026 is the first ordinary year in which the event has moved, and that makes the resulting distortion unfamiliar rather than routine.
The consequence is arithmetic rather than commercial. An event worth 26.4 billion dollars sitting in June instead of July inflates June’s comparison and deflates July’s. Anyone reading July in isolation sees a market losing momentum. Anyone reading June and July together sees a market growing at roughly the rate it was already growing at, with a large lump of demand having shifted four weeks earlier.
The trap is that the distortion does not end here. It reverses next year. In July 2027, if the event returns to its usual slot, July will be compared against a July that had no Prime Day in it, and the growth rate will look extraordinary. Neither number will describe anything real about demand.
If a moved event can distort a national figure this much, it is doing the same thing to yours. Sellerboard breaks profit down by period and by SKU from your settlement data, so you can line up the event weeks themselves instead of comparing a month that had a peak against one that did not.
Affiliate link. E-CommSphere may earn a commission at no extra cost to you.
The version of this that lands on your own dashboard
Every seller who took part in Prime Day now has a June that looks like a triumph and a July that looks like a collapse, and neither is a fact about their business. The same distortion propagates into the decisions that get made off those numbers.
Inventory planning is the first casualty. A forecast built on last year’s monthly shape will order for a July peak that is not coming, or will read this July’s softness as a demand signal and under-order for a season that has not actually weakened.
Advertising budgets are the second. Month-on-month efficiency comparisons across a moved event are meaningless, because the June figures include peak-event competition and peak-event conversion rates and the July figures do not. A cost per acquisition that rose in July is not evidence that anything got worse.
And the third is the conversation with whoever you report to. A 2.2 percent monthly drop in a market that grew 7.7 percent year on year is a sentence that can be honestly written two completely different ways, and the person hearing it usually cannot tell which one they are getting.
What we would check before drawing any conclusion
- Compare the event weeks, not the months. Line up 23 to 26 June 2026 against last year’s Prime Day dates directly. That is the only comparison in this period that holds anything constant.
- Recalculate June and July as a single block. Two months containing exactly one Prime Day, against two months containing exactly one Prime Day. The distortion cancels out.
- Annotate your own dashboards now, not in six months. Whatever tool you use, the event dates need to be visible on the chart. The alternative is explaining the same anomaly repeatedly for a year.
- Expect the mirror image in 2027. If the event returns to July, next July’s year-on-year growth will look spectacular for the same non-reason. Write that down somewhere you will find it.
A demand dip and a calendar artefact look identical on a monthly chart, and only one of them should change what you order. Helium 10 gives you search volume history per keyword, so you can check whether interest in your category actually softened or simply arrived a month early. Code ECOMMGM10.
Affiliate link. E-CommSphere may earn a commission at no extra cost to you.
The European note, and it is an absence
These are United States figures throughout, drawn from US retail data. No European equivalent is published in this reporting, and the European Prime Day calendar is Amazon’s to set market by market.
The mechanism transfers even though the numbers do not. Any market where Amazon moved its event will show the same artefact in its national ecommerce statistics and in the accounts of every seller in it. If you sell in Germany, Spain, Italy, France or the UK, the question worth answering is a local one: when did the event run in your marketplace this year, when did it run last year, and have you compared like with like since. Most sellers have not, because monthly reporting makes it easy not to.
The one line to take
US online sales did not fall in July. They grew 7.7 percent against last year while a 26.4 billion dollar event moved out of the month. A monthly comparison spanning a shifted peak is not a measurement, it is an artefact, and the operators who lose money on this will be the ones who treated it as a signal and reordered accordingly.
Sources
- Digital Commerce 360, Abbas Haleem, Retail ecommerce sales growth in the U.S. freezes in July 2026, 14 August 2026