How to Calculate ROAS on Amazon (With a Worked Example)

The Amazon ROAS formula

ROAS = ad-attributed revenue ÷ ad spend. Spend $500 on Sponsored Products, generate $2,500 in attributed sales within that ad type's attribution window, and ROAS is 5 — written as 5x or "5:1."

Amazon calculates this for you inside Sponsored Products, Sponsored Brands and Sponsored Display reports as "Total ROAS," built from the revenue and spend columns already in your account. You don't need a separate calculator. You need to know what's inside those two numbers, because that's where the confusion actually starts.

The revenue side is ad-attributed sales only — purchases Amazon's system credits to a click or view on that specific ad, inside its attribution window. It is not your total Amazon revenue, and it is not profit. That one distinction matters more than the arithmetic.

A worked example, at scale

The formula is trivial for one campaign. It gets more useful once you're managing many campaigns and want an honest number for the whole account rather than the best line item held up as proof.

Here's what that looks like on an actual book of business. Across 30 Amazon DSP advertisers we manage, July 2026 delivered 6.04x ROAS measured across the entire portfolio — every campaign, every advertiser, not the top decile pulled out and presented alone. That book ran 78.4 million impressions at a $4.00 CPM, with a blended $1.42 CPC across all placements (the $0.41 CPC figure that circulates in this category is online-video only — quoting it as a whole-book number is its own kind of error). Blended cost per acquisition was $5.49 across 57,137 attributed purchases, and 20.1% of those purchases came from a shopper new to the brand.

None of that changes the formula. It changes what you should trust when someone hands you a ROAS figure: ask whether it's blended across the whole account or lifted from the one campaign that happened to work.

Minimum ROAS: the number that actually decides whether to keep spending

Total ROAS tells you campaign efficiency. It doesn't tell you whether the campaign made money. For that you need minimum ROAS — your break-even point.

Minimum ROAS = sale price ÷ gross profit before advertising. Take a $40 product with $28 in cost of goods and Amazon fees before a cent is spent on ads. Gross profit before ads is $12, so minimum ROAS is 40 ÷ 12 = 3.33. Below 3.33x, that campaign loses money on every sale even while the headline ROAS number looks fine. Above it, every additional point of ROAS is profit.

The table below shows why margin — not a category benchmark someone quoted you — should set the target.

Where the calculation quietly breaks

The formula never changes. The inputs do, and three mistakes show up constantly — including ones worth watching for even in accounts we manage.

• Double-counting DSP and Sponsored Ads. If a shopper sees a DSP display ad and later clicks a Sponsored Products ad before buying, last-click attribution can credit that sale to both channels' reports. Add the two ROAS figures together and you're not measuring incremental value — you're measuring the same purchase twice.
• Blended vs. campaign-level. A 6x account average can sit on top of a branded-search campaign running at 15x and a prospecting campaign running at 1x. Both numbers are true. Neither tells you what to do next without seeing the split.
• Mismatched attribution windows. Sponsored ads, DSP, and any third-party tool can each run a different window. Comparing a 7-day Sponsored Products figure against a 14-day DSP figure and calling it apples-to-apples is a common, avoidable error.

Last-click attribution, on its own, can't tell you whether the ad caused the sale or simply arrived last for a customer who was buying anyway. It only reports what happened last. Holdout tests and matched-control groups are what answer the causality question ROAS was never built to answer.

ROAS vs. ACOS vs. ROI vs. TACoS

These four get used interchangeably and shouldn't be.

• ACOS is the inverse of ROAS, expressed as a percentage: ad spend ÷ ad-attributed revenue. A 4x ROAS is a 25% ACOS — same math, different frame.
• ROI is broader. It nets ad spend against total business cost — inventory, overhead, agency fees — rather than media spend against media-attributed revenue.
• TACoS divides ad spend by total sales, not just ad-attributed sales, which makes it the closest of the four to answering whether advertising is growing the business rather than claiming credit for organic sales that would have happened anyway.

None of the four measures profit directly. All four can look healthy on a product that's losing money once storage fees, returns, and fulfillment costs are counted.

When ROAS is bad news: what to check before you touch a bid

Three situations, three different fixes.

The number itself might be wrong. Check the date range against the attribution window — a campaign launched nine days ago will show artificially low ROAS if you're comparing against a 14-day window that hasn't closed. Check whether you're looking at blended account ROAS when you meant to look at one ASIN.

The setting you're about to change might already be right. Raising a bid on a campaign already hitting top-of-search placement rarely improves ROAS — it just pays more for the same auction position. Check placement and impression share before assuming the fix is more budget.

The fix might genuinely not work, and that's useful information too. If you've tightened match types, added negatives and adjusted bids for two full attribution cycles and ROAS hasn't moved, the campaign may be capped by category competition or listing conversion rate — problems a bid change can't solve. At that point the honest move is a holdout: pause spend on a matched set of ASINs or audiences and compare against a control. If sales don't drop when the ads stop, the ROAS was never incremental in the first place.

Doing this yourself vs. having someone reconcile it

Everything above is the arithmetic. Most sellers can and should run it themselves — the formula fits in a spreadsheet and the break-even math takes ten minutes. It gets harder at the account level, once Sponsored Ads and DSP are both running and you need one number that isn't double-counting itself, or once you want to know if the display spend is causal rather than coincidental.

reMKTR runs Amazon DSP as a managed service — 109 live advertiser seats, reconciled through Amazon Marketing Cloud so DSP and sponsored ads stop claiming the same sale twice, tested with holdouts rather than last-click alone. That's not the calculation covered above; it's the layer above it, for accounts where the blended number stopped being enough. Full Circle, the group reMKTR is part of, has managed more than $500M in Amazon spend across 100+ brands, including HexClad, Ridge, BK Beauty, Beardbrand, Epic Gardening, The Woobles and Walkize.