Average ROAS on Amazon: What's Normal, and What to Do When Yours Isn't

What counts as an average ROAS on Amazon?

Ask ten sellers what's a good ROAS on Amazon and you'll get ten different numbers, because the honest answer depends on three things: which ad product you're running, what category you're in, and whether you're looking at one campaign or your whole account. Sponsored Products and Sponsored Brands campaigns commonly get quoted in the 2x to 4x range — that's the figure most benchmark articles settle on, and it's a reasonable anchor for a mature, exact-match campaign in a normal-margin category.

That range hides enormous variance. A low-margin consumables brand running broad match to build market share will look "bad" against that anchor and still be profitable. A premium accessory brand running tight exact-match retargeting can clear well above it and still be under-spending against real demand. The average is not the target — it's a starting point for asking why your number is different from it.

Amazon DSP muddies this further because it's a different buy entirely: programmatic display and video, often bought for upper-funnel goals like new-to-brand reach, not last-click conversion. Judging a DSP campaign against a sponsored-ads benchmark is comparing two different jobs.

How ROAS is calculated — and where the confusion with ACoS starts

ROAS is revenue divided by ad spend. ACoS is ad spend divided by revenue, expressed as a percentage. They measure the identical relationship; the only difference is which number sits on top. The table below shows the same set of results both ways.

Once you can flip between the two without a calculator, "what's a good ROAS" turns into "what ACoS can this product's margin absorb" — which is a business question, not a marketing one. If a product carries a 40% margin, a ROAS of 2.5x, an ACoS of 40%, is roughly your break-even point, not your target.

Why one campaign's ROAS and your account's ROAS tell different stories

A single high-performing exact-match campaign can run well above your account average while the account underneath it runs far lower, and both numbers are true at once. The mistake is reporting the first one as if it represented the whole account. We measure at the book level for this reason: across 30 of our advertisers in July 2026, the portfolio delivered 6.04x return on ad spend — measured across the whole book, not cherry-picked from the best line item.

That same book ran 78.4 million impressions at a $4.00 CPM and a blended $1.42 cost-per-click. The $0.41 CPC figure that circulates in this category is online-video only — quoting it as a blended number overstates efficiency by a wide margin. Blended cost per acquisition across that book was $5.49 on 57,137 attributed purchases, and 20.1% of those purchases came from a shopper new to the brand. None of those numbers are "the average ROAS on Amazon" either — they're one group of advertisers, in one month, on one book. Use them to sanity-check your own reporting, not to replace it.

If your ROAS number is bad news, check these before you panic

• Attribution window mismatch: a 7-day and a 14-day window on the same campaign will produce two different ROAS numbers from identical spend. Confirm which window your report is using before comparing it to last month.
• Double-counted conversions: if DSP and sponsored ads both claim credit for the same last-click sale, your combined ROAS looks worse, or artificially better, than either channel actually performed. That's a reconciliation problem, not a performance problem.
• New launch penalty: a product with no reviews and a thin catalog page will always show a worse ROAS than a mature listing, regardless of targeting quality. Fix the page before you fix the bids.
• Margin, not marketing: sometimes the ROAS is fine and the product simply doesn't have the margin to support any ad spend at all. That's a pricing conversation, not a campaign one.

If you've checked all four and the number is still bad, the campaign is probably the problem — pause it and rebuild rather than nudging bids for another month.

The mistake almost everyone makes with this metric — including us

The common mistake is treating ROAS as proof of causation. Last-click attribution tells you which ad was touched last before a sale, not whether that ad caused the sale or simply intercepted a shopper who was buying anyway. A branded search campaign will almost always show a flattering ROAS because it's catching demand another channel already created.

We've made a version of this mistake ourselves: reading an early DSP display campaign as underperforming because its last-click ROAS was low, when the actual effect showed up a week later as a lift in branded sponsored-ads conversion. Last-click attribution couldn't see that connection. It took reconciling DSP and sponsored ads in Amazon Marketing Cloud, and comparing exposed shoppers against a holdout group, to see the display spend had actually worked. That's the difference between a metric and a measurement.

Where reMKTR fits

reMKTR runs Amazon DSP as a managed service and reports at the book level for this reason — we hold real DSP seats, reconcile spend in Amazon Marketing Cloud so display and sponsored ads stop double-counting each other, and use holdouts rather than last-click to test whether a campaign actually added sales. That's the discipline behind the 6.04x figure above. Whether or not that's the right setup for your account, the questions in this piece — which attribution window, whose credit, what margin — are the ones worth asking before you accept any number as "average."