A report that says your ads returned 5x sounds like a win. Whether it is one depends on a number that report doesn't contain: what you keep on each unit before the ad is bought.
ROAS is attributed ad sales divided by ad spend. ACoS is the same two numbers upside down. Neither one knows your landed cost, your Amazon fees or your returns, so neither one can tell you whether an ad made money. A ROAS only means something next to the SKU's break-even ROAS, and a food brand's break-even is different on every product it sells.
What's the difference between ACoS and ROAS?
ACoS and ROAS are the same measurement written two ways: ACoS is ad spend divided by attributed sales, and ROAS is attributed sales divided by ad spend. Amazon's own guide to ACoS calls ROAS its inverse.
ROAS stands for return on ad spend. Spend $1 on ads, get $5 in attributed sales, and ROAS is 5. ACoS, the advertising cost of sales, runs the other way: $1 of spend against $5 of sales is an ACoS of 20%. To convert, divide 100 by either number.
| ACoS | ROAS | Ad cost per $100 of attributed sales |
|---|---|---|
| 10% | 10.0 | $10 |
| 20% | 5.0 | $20 |
| 25% | 4.0 | $25 |
| 33% | 3.0 | $33 |
| 50% | 2.0 | $50 |
| 100% | 1.0 | $100 |
They are the same fact stated two ways, so switching between them changes nothing about what the ad did. A lower ACoS and a higher ROAS are the same improvement.
Both numbers stop at revenue. Attributed sales are what the customer paid, before the product cost, before the referral fee, before fulfillment, before storage and before the return that comes back three weeks later. That is why a campaign can sit at a ROAS that looks healthy while the channel loses money.
Should you use ACoS or ROAS on Amazon?
Use whichever your team reads fastest, with one practical lean: ACoS is easier to check against margin, because both are percentages of the sale. If a SKU keeps 35% before ads, a 28% ACoS leaves 7 points and you can see that without a calculator. Hold the same SKU against a 3.6 ROAS and you have to do a division first.
ROAS earns its place when you're comparing channels, because Google and Meta default to it while Amazon's console defaults to ACoS. If the same person reads all three, one number across the page is worth more than the small convenience of ACoS. What doesn't work is one person reporting ROAS and another setting ACoS targets with nobody converting between them.
What is break-even ROAS?
Break-even ROAS is 1 divided by your pre-ad gross margin on that SKU. If you keep 40% of the sale price after landed cost and Amazon's fees, break-even is 2.5. Every dollar of ad-attributed sales leaves 40 cents to pay for the ad, so the ad can cost 40 cents per dollar of sales before the unit stops contributing. Break-even ACoS is simpler still: it equals the margin. Amazon's guide makes the same point, that ACoS has to sit below your profit margin to stay profitable.
The same formula on a thinner product tells a different story:
| Pre-ad gross margin | Break-even ROAS | Break-even ACoS |
|---|---|---|
| 50% | 2.0 | 50% |
| 40% | 2.5 | 40% |
| 30% | 3.3 | 30% |
| 20% | 5.0 | 20% |
A brand running a 4x ROAS across a catalog is comfortably profitable on the 50% product, barely profitable on the 30% product and underwater on the 20% product. The dashboard reports one blended ROAS. The P&L experiences three different outcomes.
One unit shows it more plainly. These numbers are illustrative, so swap in your own landed cost and your own current Amazon fees. Take a $30 case of snack bars with $9 of landed cost and $10.50 of Amazon fees. It keeps $10.50, a 35% pre-ad gross margin, and breaks even at a ROAS of about 2.9.
| ROAS on the sale | ACoS | Ad cost per unit | Left after the ad |
|---|---|---|---|
| 5.0 | 20% | $6.00 | $4.50 |
| 4.0 | 25% | $7.50 | $3.00 |
| 2.9 | 35% | $10.50 | $0 |
| 2.0 | 50% | $15.00 | -$4.50 |
That break-even also moves whenever the price does, and in food the price moves a lot. Put a $3 coupon on the same case, or give that discount to a subscriber. The referral fee shrinks a little with the price, so call it about $8 kept on a $27 sale. Margin falls to roughly 30% and break-even climbs to about 3.4. A campaign that was clearing break-even on Friday can be under it during the weekend's promotion, with the ad report showing nothing unusual. Run your break-even at the price the ad-driven shopper pays, and that includes the coupon and the Subscribe & Save discount.
This is why we build margin per SKU before we touch a campaign. The method is in the 30% floor piece, and the reason ads need the margin first is that a ROAS target set without it is a guess. If you want the version of this argument that covers the total-spend view, what a good TACoS looks like picks it up from there.
What is a good ROAS on Amazon?
A good ROAS on Amazon is one above that SKU's break-even by the profit you want to keep, so the answer is different for every product. Jungle Scout puts the industry average around 4, and search the question and you'll usually be told 3 to 4. That's a fair description of the average account and no help with yours: at a 4x ROAS our illustrative snack case keeps $3 and a 20% margin product loses money.
Set the target with one subtraction. Target ACoS is your pre-ad gross margin minus the profit you want left on an ad-driven sale. A 35% margin with 10 points of profit wanted gives a 25% ACoS target, which is a 4.0 ROAS. Take off whatever else has to come out of that margin first, including what you pay someone to run the account.
Josh Grillo, who runs brand management here, usually aims for an ACoS around 30% on a consumables brand, can push it to 40% or 50% when the job calls for it, and doesn't want to see anything near 100%. Those are working ranges, and your margin still sets the ceiling. At the 30% consumables floor, a 30% ACoS is break-even with nothing left over, so a brand sitting at that floor has to run tighter than Josh's usual aim or be buying customers on purpose.
One target for the whole account won't work, because the campaigns are doing different jobs:
| Campaign job | What to expect from ROAS | What to judge it on |
|---|---|---|
| Branded defense | Far above anything else in the account | Share of your own searches you're winning |
| Proven non-branded exact match | At or above target ROAS | Profit after ads on that term |
| Discovery, such as auto campaigns | Around break-even, sometimes under | New converting terms found, on a capped budget |
| New-customer pushes | Under break-even by a planned amount | Payback from reorders, worked out below |
How is ROAS different from ROI?
ROAS compares revenue to ad spend, and ROI compares profit to ad spend, so ROI is the one that can go negative. To get from one to the other, multiply ROAS by pre-ad gross margin and subtract 1.
A 4x ROAS at a 35% margin returns $1.40 of margin on each ad dollar, for an ROI of 40%. A 5x ROAS at a 15% margin returns 75 cents on each ad dollar, for an ROI of -25%. The second campaign has the better ROAS and loses a quarter of every dollar it spends. That one line of arithmetic is what a CFO is asking for when they ask what the ads returned.
Three ways a good-looking ROAS misleads
Branded campaigns carry the average
Searches for your own brand name convert at very high rates, because the shopper already chose you. A defense campaign on your brand terms can post a ROAS several times higher than anything else in the account, and it should, because that is what defending your own name looks like.
The trouble starts when the account reports one blended figure. We often see branded campaigns running at exactly the high ROAS you'd expect while most of the non-branded spend loses money. The blend looks fine. Split apart, the picture is a defense campaign paying for a prospecting program that isn't working, and the biggest leak is usually broad match with no negative keywords behind it.
The first thing to ask any person or agency reporting ROAS to you: how much of it is branded? If the answer takes a week to produce, the number has not been looked at closely. Josh calls an agency reporting a strong ROAS that turns out to be almost all branded sales "the biggest manipulation in reporting I have seen in my career," and the PPC agency piece has the rest of that conversation.
Attribution counts sales that would have happened anyway
Amazon credits an ad with a sale when a shopper clicks it, or on some Sponsored Brands and Sponsored Display campaigns views it, and buys within the attribution window. It cannot show you the sale that would have come through organic placement at no cost. On a listing that already ranks well for a term, a share of the ad-attributed sales on that term is the same shopper arriving by a more expensive door.
This is the reason a low TACoS can be a warning and a high ROAS can be a harvest. Neither says anything about new demand. The cleaner test is whether total sales and new-to-brand customers rise when spend rises, which the TACoS piece walks through step by step.
The window itself also changes the number. On seller accounts, Sponsored Products reporting has generally defaulted to 7 days and Sponsored Brands and Sponsored Display to 14, according to agency and tool documentation, so check the label on your own report. A 14-day campaign set beside a 7-day campaign has a longer run at collecting credit, and the comparison flatters it before anyone has changed a bid.
A click can also credit a sale of something other than the product advertised. That matters in food, where a shopper clicks the ad for one flavor and buys the variety pack. The spend sits on one SKU and the sale on another with a different margin, so a per-SKU ROAS needs the advertised sales read apart from the rest. The purchased product report is the place to start.
Returns and fees arrive after the ad report closes
A ROAS figure is only as final as its attribution window, and returns can land later still. For food, returns run lower than for apparel or gadgets, but damaged-in-transit and expired-on-arrival claims still land weeks later, and the ad that drove the sale never carries that cost. A SKU with a high return rate has a break-even ROAS above what its margin implies, and nothing in the ad console will tell you so.
The adjustment is small and worth making. Say 3 in every 100 of those illustrative snack cases come back unsellable at about $15 each in lost product and fees. That's $45 across 100 units, or 45 cents a unit, so the case keeps about $10.05 and break-even moves from about 2.9 to 3.0. Spread the cost across every unit sold, and use your own rate.
"Don't judge ROAS at the campaign level" is half right
A common reply from agencies, when a client questions one weak campaign, is that ROAS can't be judged one campaign at a time and has to be read across the whole program. The part that is true: campaigns do different jobs. A prospecting campaign running at a lower ROAS can feed the one that harvests, and judging each in isolation leads to cutting the thing that fills the funnel. Campaign age matters too, since a new campaign hasn't earned the benchmark an older one has.
The part that is wrong is the conclusion. Read across the whole program, the program still has to clear break-even, and the way to know is the same: split brand from non-brand, look at new-to-brand customers, and compare the blend to the margin. "Look at it holistically" is a reason to put more numbers on the table. It is not a reason to stop looking at any of them.
When is it worth running below break-even ROAS?
It's worth running below break-even ROAS when the campaign is winning new customers and their reorders will pay back the loss, and you can show the arithmetic before you spend. This is the strongest reason a food brand has to break the rule, because a consumable gets bought again and a one-time purchase doesn't.
Back to the illustrative snack case. A new-customer campaign running at a 2.0 ROAS spends $15 to sell a unit that keeps $10.50, a $4.50 loss on the first order. A reorder with no ad behind it keeps the full $10.50. Each customer needs to average about 0.43 reorders for that campaign to wash. If those reorders come through Subscribe & Save at the $8 we worked out earlier, it's about 0.56.
The catch is that the math only applies to customers who are new. Say half of that campaign's sales went to people who already buy from you. Two sales now cost $30 in ads, kept $21 and produced one new customer. That customer starts $9 in the hole and has to average about 0.86 reorders. Same ROAS, roughly double the bar.
Before the campaign starts, write down the first-order loss you'll accept per new customer and the reorder rate, inside a set number of days, that pays it back. At the end of those days, check your new-to-brand share and your reorders against what you wrote. If you don't know your reorder rate, you can't make this case yet, and the campaign should run to break-even until you do.
How long before you can trust a campaign's ROAS?
Give it enough orders and a few days to settle, because a small sample moves too much to read and the newest days haven't finished collecting credit. Say a term has spent $40 and sold three of those $30 cases, an ACoS of 44%. One more order takes it to 33%, and that's the difference between cutting the term and keeping it. On a small sample, one order can flip the decision.
Leave the most recent days out of the read as well. A shopper who clicked yesterday can still buy inside the window, and last week's ROAS reads low today for that reason. We'd read a full period that ended at least a window ago, and compare it to the same length of time before it.
A term that never converts needs a different rule, because it has no ROAS to read. The margin gives you one. Our snack case can afford $10.50 of ad cost per sale, so a term that has spent two sales' worth of that, $21, with no order has already missed break-even twice. Pull the bid down there, and negate the term once it has had enough clicks for the silence to mean something, about 30 at a 10% conversion rate.
The scoreboard we manage to instead
We manage to profit after ads, per SKU, with four numbers beside each other:
- Break-even ROAS for the SKU, from true landed cost and current fees.
- Branded and non-branded ROAS reported separately, never blended.
- TACoS, which is ad spend over total sales, so growth in organic sales shows up as a lower number rather than hiding behind ad attribution.
- New-to-brand customers, because a program that only re-sells to people who already know you is not growing the brand.
One caution on that fourth number. As we read Amazon's help documentation, new-to-brand reporting covers Sponsored Brands and Sponsored Display and not Sponsored Products, so if most of your spend is Sponsored Products, the figure covers only part of the account. Read it as a trend.
When all four move together, the account is working. When ROAS is climbing and the other three are flat or falling, the ads are collecting credit for sales that were coming anyway.
For a manufacturer reporting to a parent company, there is a fifth line that matters more than any of these: contribution after ads. It is the only number the CFO is going to care about, and it is the one ROAS can't produce.
Do ACoS and ROAS work the same way on Instacart?
The formulas are the same on Instacart, but break-even usually sits higher, because, as we read the reports, attributed sales are counted at the shelf price and the brand only collects its wholesale price. ROAS is the number you'll see reported there, and Instacart says its attribution window is 14 days, with multi-touch as its default sales reporting. Don't set either beside a 7-day Amazon figure and call it a fair comparison.
Take an illustrative $6 item on a grocery shelf. If the brand receives $3.30 after the retailer and distributor take their share, and the product costs $1.80 to make and deliver, it keeps $1.50. Against the $3.30 that's a 45% margin, and against the $6 the ROAS is measured on it's 25%, so break-even ROAS is 4.0. Build the margin against shelf price, include your trade spend, and read Instacart spend as a cost of selling there, since its ads don't build organic rank the way Amazon's do.
What to do this month
- Pull your pre-ad gross margin for your top five SKUs and write the break-even ROAS next to each, at the price the shopper pays after coupons.
- Ask for ROAS split by branded and non-branded for the last 90 days.
- Find the campaigns running under break-even and decide, one at a time, whether they are filling the funnel or just spending.
- Write a payback rule for anything you decide to keep below break-even.
- Add negative keywords to any broad-match campaign that has none.
- Put the four numbers above on one page and look at it monthly.
The bottom line
ROAS tells you how much revenue an ad is credited with. Whether that revenue was worth buying depends on the margin underneath it, so the first job is to write that margin next to every SKU and read the ads against it. A brand that manages to true profit after ads usually finds one or two campaigns to cut and one to fund, and the account gets smaller on the dashboard and larger in the bank.
If you'd like a second read on yours, send us your ASINs and we'll split branded from non-branded, set each SKU's ROAS against its break-even, and tell you which campaigns we'd cut first.






