ROAS: The Formula, What a Good ROAS Is and Why Marginal ROAS Decides Budgets

ROAS (return on ad spend) is the revenue an advertising campaign brings in for every dollar spent on it, calculated as revenue divided by ad spend. A campaign that spends $20,000 and is credited with $86,000 of revenue has a ROAS of 4.3, which can also be written as 430% or 4.3:1.

A single channel carries three return figures at the same time. The ad platform reports one, a lift test or marketing mix model measures the incremental return, and the response curve gives the marginal return on the next dollar. Only the marginal figure, judged against your break-even point, should decide where the next dollar goes. Allocating on averages keeps sending money into channels that are already saturated.

What is ROAS?

ROAS stands for return on ad spend. A ROAS of 4 means the advertising was credited with $4 of revenue for every $1 it cost.

Return here means revenue, and ad spend means the money paid for media. Nothing else enters the formula: no cost of goods, no shipping, no salaries, no overheads. That makes the ratio fast to calculate and easy to compare between campaigns, and it is also why the ratio misleads once it reaches a finance meeting.

What ROAS does not tell you

It leaves three questions open.

• Whether the revenue would have happened anyway. A platform counts every conversion it can tie to an ad, including customers who were already going to buy. The complete guide to marketing mix modeling explains how a model separates these base sales from incremental ones.
• Whether the revenue was profitable. A ratio of 3 loses money for a business on a 30% gross margin.
• What the next dollar will return. The ratio is an average across all the spend so far, and says nothing about the next $10,000.

The ROAS formula, with a worked calculation

The formula is one division:

ROAS = attributed revenue / ad spend

Attributed revenue is whatever the platform or analytics tool credits to the ads, under its own rules.

How to calculate ROAS step by step

Take an illustrative paid social campaign.

1. Fix the scope and the period: one campaign, one calendar month.
2. Add up the ad spend for that period: $20,000.
3. Add up the revenue credited to the campaign over the same period: $86,000.
4. Divide revenue by spend: $86,000 / $20,000 = 4.3.

The campaign returns $4.30 of attributed revenue for every $1 of media.

What belongs in revenue and in spend

Two definitions decide the answer before any ad runs. Revenue can be gross order value, or net of returns, cancellations, discounts, tax and shipping charged. Spend can be media cost only, which is what platforms report, or media plus agency fees, technology fees and creative production.

Apply two plausible adjustments to the illustrative campaign. If 12% of the $86,000 comes back as returns, net revenue is $75,680 and the ratio falls to 3.78. Add a 15% agency fee to the $20,000 of media and spend becomes $23,000, which gives 3.74. Apply both and the campaign reads 3.29. Write both definitions down and keep them fixed between periods.

Campaign, channel and account level

The formula is the same at every level, and summing across platforms is where double counting enters (see MER below). Every aggregate is also a spend-weighted average, so a strong campaign can hide a weak one. Split the $20,000 campaign in two: $5,000 of retargeting credited with $40,000 (8.0), and $15,000 of prospecting credited with $46,000 (3.07). Together they still read 4.3, and nobody looking only at that 4.3 can see that three quarters of the money earns 3.07.

Is ROAS a ratio, a multiple or a percentage?

It can be written as any of the three, and they are the same number: 4.3:1, 4.3x and 430%. Multiply the ratio by 100 to get the percentage, and divide by 100 to go back.

Platforms differ on format. Google Ads expresses target ROAS as a percentage, so $5 of sales on $1 of ad spend is a 500% target. Most other dashboards, Meta Ads Manager included, show a plain multiple. The percentage form is the easier one to misread. "300%" sounds like a 300% return, when it means $3 of revenue per $1 of spend, which at many margins is a loss.

ACOS is the same ratio upside down

ACOS (advertising cost of sales), the efficiency metric used across Amazon advertising, divides spend by revenue:

ACOS = ad spend / attributed revenue = 1 / ROAS

A ratio of 4.3 is an ACOS of 23.3%, and an ACOS of 25% is a ratio of 4.0. Break-even ACOS equals your margin, so a product on a 40% margin breaks even at 40% ACOS.

ROAS vs ROI: margin makes the difference

ROAS asks how much revenue the ads were credited with per dollar. ROI asks whether the ads returned more profit than they cost.

Here, ROI means the gross profit on ad-driven revenue, minus the ad spend, divided by the ad spend. That is the structure of the ROMI definition in our guide to marketing effectiveness, which applies it to incremental profit. In terms of the revenue multiple it is one line:

ROI = (ROAS x gross margin) - 1

Worked example at a 40% margin

At a 40% gross margin, the illustrative campaign's $86,000 of attributed revenue carries $34,400 of gross profit. Subtract the $20,000 of spend and $14,400 is left, an ROI of 72%. The same arithmetic across a range of values gives this illustrative table.

ROAS	Gross profit per $1 of spend	ROI after ad spend
2.0	$0.80	-20%
2.5	$1.00	0%
3.0	$1.20	20%
4.3	$1.72	72%
6.0	$2.40	140%

Why a healthy-looking ratio can lose money

A ratio of 2.0 looks like doubling the money. At a 40% margin it returns 80 cents of gross profit per dollar spent, a loss of 20 cents before any fixed cost is counted. The lower the margin, the higher the ROAS a business needs to break even.

Check the vocabulary whenever ROI appears in a marketing mix modeling deck. Google's open-source MMM, Meridian, defines ROI as incremental outcome divided by channel cost: $25,000 of incremental sales on $10,000 of spend is an ROI of $2.50. That is revenue over spend, so an MMM "ROI" is usually what this guide calls incremental ROAS, and it says nothing about profit until a margin is applied.

Break-even ROAS: the line every campaign has to clear

Break-even is the ratio at which the gross profit from ad-driven revenue exactly covers the ad spend. Below it, the ads lose money once the media is paid for. Above it, they contribute to fixed costs and profit.

The formula: 1 / gross margin

At break-even, revenue x gross margin = ad spend, so revenue / ad spend = 1 / gross margin:

Break-even ROAS = 1 / gross margin

A 40% gross margin breaks even at 1 / 0.40 = 2.5. A 25% margin breaks even at 4.0, and a 70% margin at 1.43. Treating 1:1, or 100%, as break-even assumes revenue only has to equal ad cost, which holds only if the goods cost nothing to make, ship or take back.

Which costs belong in the margin

Gross margin in the accounts usually covers cost of goods only. For an advertising threshold, include every cost that scales with each order: cost of goods, payment processing, pick and pack, shipping you absorb and an allowance for returns. The result is a contribution margin.

Take an illustrative $100 order with $45 of cost of goods. Gross margin is 55%, so break-even is 1.82. Add $3 of payment fees, $7 of fulfilment and shipping and a $5 returns allowance, and the contribution margin falls to 40%, which moves break-even to 2.5. A team using 1.82 as its floor would keep scaling campaigns that lose money on every order anywhere between 1.82 and 2.5.

Break-even ROAS by industry

Margins vary widely by sector, and break-even moves with them. Aswath Damodaran's January 2026 dataset of US gross margins by sector, published at NYU Stern, shows the spread, and converting each margin with the formula gives these break-even levels.

Industry (US, January 2026)	Gross margin	Break-even ROAS
Food processing	23.23%	4.30
Retail (grocery and food)	26.31%	3.80
Retail (general)	33.18%	3.01
All US listed firms (Total Market)	37.76%	2.65
Shoe	43.88%	2.28
Household products	51.04%	1.96
Apparel	56.88%	1.76
Software (system and application)	71.72%	1.39

These are sector-wide gross margins across listed US companies, not your per-order contribution margin. For a business whose gross margin matches its sector, the figure is a floor, because contribution margin is lower than gross margin. Your own break-even could sit above or below it.

Set the target above break-even

At break-even the ads pay for themselves and nothing else. A target carries the return you need on top:

Target ROAS = (1 + target ROI) / gross margin

At a 40% margin, a 20% ROI needs 1.2 / 0.40 = 3.0, which matches the ROI table above. Brands with strong repeat purchase sometimes accept a first-order return below break-even. That is a lifetime-value bet, so write it down as one, with the payback period stated and checked against cohort data.

What is a good ROAS?

A good ROAS is one where the next dollar still clears break-even on your own contribution margin, measured on revenue the ads caused.

Why published benchmarks disagree

Wall Street Prep describes 4:1 as a widely referenced benchmark for an acceptable ratio. Vendor data points lower. Triple Whale reports a 2024 median of 2.04 for brands advertising on its platform, without saying whether that figure is platform-attributed or blended.

Set 4:1 against the break-even table. For a food processor at the sector's 23% gross margin, it sits below the 4.30 break-even, so a campaign that hits the "good" benchmark loses money. For a software company at the sector's 72% margin, it is nearly three times the 1.39 break-even, so a team holding to it will turn down campaigns that would have paid back.

Benchmarks also mix attribution windows, revenue definitions and retargeting shares, so two brands quoting 3.0 can be describing different quantities.

Set different targets inside one account

Retargeting reaches people who have already shown intent, so its platform figure runs high almost by construction, while prospecting reports lower. One shared target pushes budget towards the campaign best at collecting credit rather than the one creating demand.

One channel, three ROAS numbers

An illustrative direct-to-consumer brand runs on a 40% gross margin, so it breaks even at 2.5, and spends $150,000 a month on paid social. Ask what paid social returns and three answers come back.

• Platform-reported: 4.6. The ad platform credits $690,000 of revenue to the $150,000.
• Incremental (iROAS): 3.1. A marketing mix model, calibrated with a lift test, estimates that the spend caused $466,000 of revenue that would not otherwise have happened.
• Marginal (mROAS): 1.3. At current spend, the next $10,000 would bring in about $12,700 of incremental revenue. At a 40% margin that is roughly $5,100 of gross profit for $10,000 of spend, a loss of about $4,900.

All three are correct calculations. Only the marginal figure answers where the next dollar should go, and at 1.3 against a 2.5 break-even it says cut back.

Platform ROAS vs incremental ROAS (iROAS)

Incremental ROAS divides only the revenue the advertising caused by the spend that caused it:

iROAS = incremental revenue / ad spend

Why platform-reported figures run high

A platform counts conversions it can link to an ad inside its attribution window, and many of those customers would have bought anyway because they were already searching for the brand or had a full cart. Retargeting and brand search show the effect most, because their audiences are selected for existing intent.

Field experiments at eBay show the effect. Blake, Nosko and Tadelis found that paid search returns were a fraction of what non-experimental estimates had claimed, and that brand-keyword ads had "no measurable short-term benefits". On non-brand keywords, average returns were negative, because most of the spend reached frequent users who would have bought without the ad.

How incremental ROAS is measured

• Geo lift and holdout tests withhold the media from comparable markets or audiences and measure the difference.
• Platform conversion lift studies randomise the platform's own users into exposed and held-out groups.
• Marketing mix modeling estimates each channel's incremental contribution from aggregate weekly data, and is strongest when calibrated with tests.

Our guide to incrementality testing works through a full geo lift read-out that lands at an iROAS of 1.21, set against a hypothetical platform-reported 4.2. Platform-run lift studies are randomised experiments, but the platform designs the holdout and reports the result on its own inventory, so check them periodically against an experiment you design or a model you control.

How attribution windows change reported ROAS

An attribution window is the period after a click or view during which a conversion is credited to the ad, and the defaults differ by platform. For Search and Display, Google Ads applies a 30-day default click-through window, adjustable from 1 to 90 days, and view-through conversions, which come from Display impressions rather than Search, default to a 1-day window. Meta Ads Manager defaults to 7-day click, 1-day engage-through and 1-day view attribution, and since March 2026 only clicks on an ad link count as clicks.

Wider windows credit more conversions, so the same campaign reports a higher figure. This illustrative table reads one $20,000 campaign four ways.

Attribution window	Attributed revenue	Reported ROAS
1-day click	$52,000	2.6
7-day click	$74,000	3.7
7-day click and 1-day view	$86,000	4.3
30-day click	$92,000	4.6

At a 40% margin the 1-day click read barely clears the 2.5 break-even, while the 30-day read looks comfortable. None of the four is incremental, and none is more correct than the others. A Google figure and a Meta figure on their default settings answer two different questions. The guide to marketing attribution covers windows and model choice in more depth.

Why a marketing mix model estimates its window rather than choosing one

An MMM does not credit conversions to clicks, so there is no attribution window in the ad-platform sense. It estimates carryover, how long a channel's effect persists after the spend, from the weekly data, within limits the modeller sets, such as a maximum lag and the allowed range for the decay rate. So the lag is an estimate with its own uncertainty rather than a single window picked in an ads interface.

Average ROAS vs marginal ROAS

Average ROAS divides all the incremental revenue a channel produced by all of its spend. Marginal ROAS looks only at the last increment:

mROAS = change in incremental revenue / change in spend

Why the average stays high after the extra money stops working

Media saturates. Later dollars reach people who have already seen the ad or were never likely to buy, so each added dollar returns less than the one before. The average still contains the cheap early returns and falls slowly, while the marginal return drops fast.

A worked saturation example

Give the paid social channel an illustrative response curve, incremental revenue = $600,000 x (1 - e^(-spend / $100,000)), keep the 40% margin and the 2.5 break-even, and read the curve at several monthly spend levels.

Monthly spend	Incremental revenue	Average iROAS	Marginal ROAS	Gross profit after spend
$50,000	$236,000	4.72	3.64	$44,400
$87,500	$350,000	4.00	2.50	$52,500
$100,000	$379,000	3.79	2.21	$51,700
$150,000	$466,000	3.11	1.34	$36,400
$200,000	$519,000	2.59	0.81	$7,500
$211,000	$527,000	2.50	0.73	$0
$250,000	$551,000	2.20	0.49	-$29,700

Profit after media peaks at about $87,500 of spend, where the marginal return equals the 2.5 break-even and the average is 4.0. The average does not fall to 2.5 until spend reaches about $211,000. At $200,000 it is 2.59, still above break-even, while gross profit after spend has shrunk to $7,500 and each extra dollar returns 81 cents of revenue.

Between $87,500 and $211,000, a team budgeting on averages sees a channel above break-even and keeps scaling, while every added dollar loses money. In the paid social example, the brand at $150,000 sees 4.6 in the platform and 3.1 in its MMM, and both look safe. The last $62,500 of spend costs it about $16,000 a month in gross profit.

Meridian's documentation says an mROI far below ROI signals a saturating channel, with the example of a channel near saturation whose mROI has dropped to $0.80, its cue to invest elsewhere. Meta's Robyn documentation calls mROAS "the most important metric for budget allocation". Our guide to the law of diminishing returns covers the curve shapes behind this.

The allocation rule

1. Take money from the channel with the lowest mROAS and give it to the channel with the highest.
2. Repeat until mROAS is roughly equal across channels, at which point no move between them raises the total return.
3. Grow the total budget only while that common mROAS sits above break-even, and stop where the two meet.

Minimum spends, contracts, flighting and brand rules constrain real plans, but the rule still sets the direction of every move. A what-if analysis on the response curves shows what a proposed reallocation does before any money moves.

Marginal ROAS is not incremental ROAS

Incremental versus platform asks whether the sale would have happened without the ad. Marginal versus average asks what one more unit of spend returns from where the channel sits now. Any mROAS worth acting on is incremental, but most iROAS figures are averages. In the paid social example the iROAS is 3.1 and the mROAS is 1.3, and reading the first as the second hides $16,000 a month of lost profit.

Blended ROAS and MER

MER (marketing efficiency ratio), also called blended ROAS, divides all revenue by all marketing spend:

MER = total revenue / total marketing spend

A business with $2.4 million of monthly revenue and $600,000 of total marketing spend has an MER of 4.0 (illustrative figures).

What MER catches and what it hides

MER exposes the double counting in summed platform figures. Suppose the business spends $450,000 of that budget on paid media, and Google, Meta and Amazon report $1.3 million, $1.1 million and $0.7 million of attributed revenue. The platforms claim $3.1 million between them, a combined 6.9, while the business took only $2.4 million in total. The gap is double counting, with each platform crediting itself for conversions the others also claimed. MER also picks up spend that no platform reports, such as influencer fees, TV, agency retainers and creative production.

MER cannot show causation. It moves with base demand, pricing, seasonality and organic growth, so a rising MER can mean the brand is growing on its own while paid media gets worse, and it cannot say which channel to fund. Use it as a guardrail on the P&L and a check on platform claims, and leave allocation to marginal returns.

How marketing mix modeling estimates marginal ROAS by channel

A marketing mix model fits each channel's response curve from aggregate weekly data, with an adstock term for carryover and a saturation term for diminishing returns. The MMM guide covers both transformations. Two numbers come out of each curve:

• Incremental return, often labelled ROI: the channel's incremental outcome over the period divided by its cost, as in Meridian's $2.50 example above.
• Marginal return, often labelled mROI: the slope of the curve at current spend. Raise spend by a small amount, take the extra incremental outcome the curve predicts and divide it by the extra cost.

The allocation rule needs the second number, and an MMM is the most practical way to get it for every channel at once, offline media included.

Calibrating the curves with experiments

A curve estimated from observational data is best identified across the spend levels a channel has run at, and least reliable where a plan extrapolates beyond them. Lift tests anchor it. A geo lift result for paid social can enter a Bayesian MMM as a prior on that channel's effect, act as a bound on its implied return, or serve as an independent check. When the model and the experiment disagree, find out why before acting on either.

Reading the uncertainty around the point estimate

Across 25 large field experiments with US retailers and brokerages, covering $2.8 million of digital ad spend, Lewis and Rao found that the median confidence interval on return on investment was over 100 percentage points wide. A modelled marginal return carries uncertainty of the same kind, so report it as a range and act on the decisions that hold across it. If paid social's mROAS is 1.3 with an illustrative interval of 0.9 to 1.8, the whole range sits below the 2.5 break-even and the cut is safe. An interval of 1.8 to 3.2 straddles break-even and calls for a test before any cut.

Common ROAS mistakes

1. Allocating on the average. Average returns stay above break-even long after the marginal dollar stops paying. Compare mROAS with break-even instead.
2. Treating 1:1 or 100% as break-even. Break-even is 1 / margin, which is 2.5 at a 40% margin.
3. Holding every product line to 4:1. That target loses money at food-processing margins and turns away profitable spend at software margins.
4. Adding up platform-attributed revenue across Google, Meta and Amazon. Each platform counts conversions the others also claim. Check the sum against MER and your order system.
5. Comparing platforms on different attribution windows. A 30-day click figure and a 7-day click, 1-day engage-through, 1-day view figure measure different things.
6. Reading brand search returns as fully incremental. The eBay experiments found no measurable short-term benefit from brand-keyword ads for one very well-known brand. Later Bing experiments across thousands of brands found small positive effects, and much larger ones when competitors bid on your brand terms, so test before you cut.
7. Calling incremental ROAS "marginal ROAS". One is about causation and the other about the next dollar, and a channel can pass the first test while failing the second.
8. Judging long-consideration products on a short window. A 1-day click window undercounts anything people research for weeks. Match the window to the purchase cycle.
9. Treating a point estimate as precise. Confidence intervals on ROI from field experiments are often more than 100 percentage points wide. Report ranges.
10. Changing what counts as revenue or spend between reports. Returns and agency fees alone moved the $20,000 example campaign from 4.3 to 3.29. Fix the definitions and log every change.

A ROAS stack that holds up in a finance review

Metric	Question it answers	Where it comes from	How often	Decision it can drive
Platform ROAS	Are tracked conversions pacing against spend?	Ad platform reporting	Daily	Bid, creative and audience changes in flight
Break-even ROAS	What return does the business need to break even?	Finance, from contribution margin	Quarterly	The threshold every other number is judged against
Incremental ROAS	Does this channel deserve budget at all?	Lift tests and MMM	Quarterly	Keep, cut or restructure a channel
Marginal ROAS	Where should the next dollar go?	MMM response curves	Each planning cycle	Reallocation between channels and the total budget
MER	Is total marketing spend paying back on the P&L?	Finance	Monthly	Overall spend level and a check on platform claims

Frequently asked questions

What does ROAS stand for?

ROAS stands for return on ad spend. It divides the revenue credited to advertising by the amount spent on that advertising, so $86,000 of revenue on $20,000 of spend gives 4.3. It measures revenue efficiency only and ignores cost of goods, so it says nothing about profit until you compare it with your break-even point, which is 1 divided by your margin.

What is a good ROAS for Google Ads or Meta?

Published benchmarks rarely transfer cleanly to your account, and the two platforms count revenue differently by default. Google Ads uses a 30-day click window by default for Search and Display, while Meta defaults to 7-day click, 1-day engage-through and 1-day view, so their figures are not comparable. A good result on either platform is one that clears your break-even, 1 divided by your margin, measured on incremental revenue.

Is ROAS the same as ROI?

No. In any ROI vs ROAS comparison the difference is margin. ROAS divides revenue by ad spend. ROI subtracts the ad spend from the gross profit on that revenue and divides by spend, so ROI = (ROAS x gross margin) - 1. At a 40% margin a ROAS of 2.5 is an ROI of 0%, and a ROAS of 4.3 is an ROI of 72%.

Can ROAS be negative?

No. The ratio divides revenue by spend, and neither can be negative, so the floor is 0, which means the ads were credited with no revenue at all. ROI can be negative. At a 40% gross margin, a ratio of 2.0 is an ROI of -20%, because the gross profit on the revenue is smaller than the ad spend that produced it.

What is the difference between incremental ROAS and marginal ROAS?

Incremental ROAS divides the revenue advertising caused by the spend that caused it, averaged across all the spend in a period. Marginal ROAS is the return on the next dollar at the current spend level. The first separates real effect from conversions that would have happened anyway, and the second measures saturation. One channel can show an iROAS of 3.1 and an mROAS of 1.3 at the same time.

How do you calculate break-even ROAS?

Divide 1 by your margin. A business on a 40% margin breaks even at 1 / 0.40 = 2.5. Use a contribution margin that includes every cost that scales with each order, such as cost of goods, payment fees, fulfilment, absorbed shipping and returns, rather than the gross margin in the accounts. Anything below the result loses money once the media is paid for.

How we approach it at Aryma

Aryma Labs has built marketing mix models since 2019. Of the three ROAS numbers in this guide, the marginal one, read from a marketing mix model's response curves, decides where the next dollar goes.

MMM Singularity is the interpretation layer for that work. It connects attribution, incrementality, saturation and prediction in one explainable view, the same ground this guide covers. Aryma Nebula, our agentic spend optimiser, turns MMM outputs into weekly and monthly budget allocation decisions. When the numbers have to survive a review, Aryma Dimension, our AI meeting analyst, brings eMMMy into the call to quote ROAS, contribution and saturation from your own model rather than from a platform dashboard.

If your measurement problem needs a design rather than a product, see our custom solutions. For help with ROAS, marginal returns or marketing mix modeling, get in touch with us.