Prime Day 2026 set a record and quietly broke most brands’ forecast models at the same time. US shoppers spent $26.4 billion across the event window, up 9.3% year over year. The average household spent 8.3% less than it did in 2025. Both numbers are true, and the gap between them is the most important thing that happened in June.

Record totals, smaller carts

The $26.4 billion figure, from Adobe Analytics, measures total US online retail spend during the event window — all retailers, not Amazon’s sales alone. It describes the gravitational pull of the event, not Amazon’s own results. Treat it as a measure of how much of the calendar Amazon owns, not as a revenue number.

Underneath that record, household behavior moved the other way. Numerator’s verified purchase panel puts average household spend at $143.45, down from $156.37 the year before. Households placed more orders, and each one was smaller.

So where did the growth come from? More buyers. Household penetration has climbed from 35.3% of US households in 2023 to roughly 43% in 2026. Breadth is the growth engine now, not basket size.

The four-year cart shrink

This is not a one-year blip. Both order size and spend per unit have declined every year since 2024, and the trend is steep.

The average Prime Day order is now $11.92 smaller than it was three years ago, and the average unit inside it is $8.06 cheaper. If your planning model still carries 2023-era AOV, you are over-forecasting revenue and under-forecasting order volume at the same time — which shows up as overstated topline targets, understaffed fulfillment, and ad budgets pegged to the wrong denominator.

The barbell basket

Sixty-nine percent of items sold for under $20. Three percent sold for more than $100. Average spend per item slipped from $24.59 to $23.23. Two in three items cost less than lunch.

The distribution has two masses and a hole in the middle. Below $20 is the stock-up zone — 69% of all items, concentrated hardest between $5 and $15. At the far end, premium held up: premium electronics took a 51% larger share of spend than their year-to-date average, and buy-now-pay-later financed $2.1 billion in purchases, up 9.5%.

What collapsed is everything between. A $45 product is too expensive to toss into a stock-up cart on impulse and too cheap to feel like a considered splurge worth financing. It gets deferred, and on a four-day event, deferred means lost.

That has a direct pricing implication. If your hero SKU sits in the $30–$60 band, Prime Day is the wrong event to defend that price point. Either build a genuine multipack or trial size that lands under $20, or lean into the top end with bundling and financing eligibility. Straddling the middle is how listings quietly lose the event.

Pantry Day: what actually sold

The top five sellers by unit volume were not doorbusters. Not one was electronics.

Premier Protein was also number one by units in 2025. That is acceleration, not novelty. The biggest event in e-commerce is now won by consumables — multipack-sized, repeat-purchase, low-consideration goods that people were going to buy anyway and simply timed to the sale.

Category participation reinforces it. Health and beauty is the growth engine: six of the ten fastest-growing categories in 2025 were health and beauty, and the GLP-1 economy keeps compounding protein and supplement demand.

The intent data says the same thing from the shopper’s side. Asked what they did on Prime Day, 45% waited for a specific item to go on sale, 36% bought the same items they always buy, 32% picked up general Prime Day deals, and 26% stocked up on sale items. Planned purchases beat impulse deals — which means the demand was already there and the event only decided the timing.

For brands, that reframes the job. You are not persuading someone to want your product during a four-day window. You are making sure that when a shopper who already wants it goes looking, you are the listing they find.

Deal fatigue and the Walmart problem

High deal satisfaction fell nine points in a single year: 68% of shoppers said deals were great in 2025, against 59% in 2026. The obvious explanation would be shallower discounts, and the obvious explanation is wrong. Discount depth in electronics was 24%, versus 23% the year before — flat. The discounts held. The delight came off.

Part of that is competition for the same week. Nearly half of Prime Day shoppers also shopped a rival sale.

Walmart ran seven days, June 22–28, wrapping around Amazon’s four. A shopper could price-check for a day before Prime Day opened and for two days after it closed. When a 20% discount is available at three retailers in the same week, the discount stops being a reason to choose you.

Then there is the loyalty paradox: 89% of participants had shopped Prime Day before and 84% are Prime members, but only 37% said Prime Day was their main reason to shop. The audience is deeply habitual, and the event itself persuades barely a third of them. These are your customers arriving on schedule, not new demand the promotion created.

What it cost to show up

For advertisers, the default playbook got structurally more expensive. Across Sponsored Products:

• Cost per click rose 25%, with peaks of 40–80% in contested categories

• Conversion rate fell 22.5%

• ACoS rose 21.6%

 Read those together and the picture is blunt: you paid a quarter more for a click that converted a fifth less often. That is not a bidding error you can fix mid-event. It is what happens when every advertiser crowds the same placements while the average basket is shrinking underneath them.

The formats around Sponsored Products told a different story.

Sponsored Brands carried the lowest cost inflation of any format while spend against it more than tripled. Video lifted ad-attributed sales 24%. Brands running all three formats together beat their category baseline by 139%.

The winners did not outspend the squeeze. They changed the mix before the event started. That is the part that cannot be retrofitted on day one — Sponsored Brands and video creative need approval time, and a display audience needs history to target against. By the time your ACoS spikes on the morning of day one, the window to restructure has closed. This is exactly what regular advertising audits are for: the format mix decision is made in April, not June.

What to do before Prime Day 2027

1. Model breadth, not basket size

Forecast more orders at a lower average value. Check that your fulfillment, packaging, and customer service assumptions survive a higher order count at a lower value per order, because that is the shape of the event now.

2. Pick an end of the barbell

Under $20 or genuinely premium. If your catalog only has middle, build a multipack or a trial size before next June rather than discounting your way down into the stock-up zone and training customers to wait.

3. Treat consumables as the center of gravity

If you sell a replenishable product, Prime Day is now your event and your competition is other consumables, not your usual category rivals. If you do not, expect to fight for attention against protein shakes and trash bags.

4. Buy visibility, not discount depth

Discount depth was flat and satisfaction still fell, which means an extra five points off is unlikely to buy you anything except margin loss. Spend that margin on placement instead. With 45% of shoppers waiting for a specific item, the brand that owns the search result wins the sale — and once the click lands, the listing does the converting, not the ad.

5. Shift the mix before the event, not during

Sponsored Brands, video, and display need lead time for creative approval and audience building. Lock the mix a full quarter out. The velocity you buy during the event also feeds back into organic rank, so the return on getting the structure right extends well past the four days — which is the real argument for treating Prime Day as a ranking event with a sales spike attached, rather than the other way round.

Final word

Prime Day 2026 was a record event built on a weaker consumer. More households showed up, each one spent less, and almost half of them shopped a competitor the same week. The brands that did well were not the ones with the deepest discount — discounts were flat and shoppers noticed anyway. They were the ones who had already decided, months earlier, which end of the barbell they were selling into and which ad formats they would show up in.

The event rewards preparation now, not aggression. If your Prime Day plan starts in May, you are already buying at the worst possible price.

Want to know how your category performed against these benchmarks? A Prime Clicks audit reads your conversion paths, cost inflation, and format mix against the panel data above. Book a 30-minute read-out — no commitment.