What is a good ROAS?

A good ROAS is any return above your break-even — which is 1 divided by your gross margin — with enough cushion to bank a profit. At a 30% margin you break even at 3.3x; at 50%, 2x already makes money. The 4:1 “benchmark” quoted everywhere has no study behind it, and the newest real cohort data puts the median blended ROAS near 2x (Triple Whale, ~35,000 brands, 2025). So the honest answer is a calculation, not a number — and it takes about a minute.

That minute is why we built a ROAS calculator with break-even built in. This page is the longer answer: where the famous benchmarks actually come from (it’s worse than you’d think), what current data really shows, and the two ways your platform’s ROAS number quietly lies to you.

The ROAS formula — and what it leaves out

ROAS is revenue divided by ad spend. $4,200 of tracked revenue on $1,000 of spend is a 4.2x ROAS — every ad dollar returned $4.20. Ad platforms report the same number as a percentage (420%), and Google’s bidding now uses it directly: in June 2026 Google shortened the strategy name “Maximize conversion value with a target ROAS” to simply “Target ROAS” (Google Ads Help), so this metric is literally the steering wheel of most accounts.

What the formula leaves out is everything your revenue has to pay for. Product costs, shipping, fees — none of them appear in ROAS, which is why a campaign can post a proud 4x and still lose money. ROAS tells you what came back; it doesn’t tell you what you kept. For that you need one more number.

Break-even ROAS: the only benchmark that’s actually yours

Break-even ROAS is 1 divided by your gross margin — the return where ads stop losing money and start making it. It’s the fastest piece of arithmetic in marketing and the most skipped:

Read the third column twice: the same 3x ROAS is a loss for a 25%-margin store and a solid win at 50%. That’s the whole reason “what is a good ROAS” can’t have a universal answer — the question is really “what’s your margin,” and no benchmark table knows it. Your target is break-even plus a cushion: a straight profit cushion if every order must pay for itself, or a lifetime-value cushion if customers reorder — a business willing to spend 30% of customer lifetime value on acquisition can target roughly 3.3x against LTV and deliberately run near break-even on first orders. Run your own numbers before you accept anyone’s target, including ours.

Where the famous benchmarks actually come from

We traced the numbers that dominate this topic to their sources (all checked July 11, 2026). The results are an argument for never trusting an uncited benchmark:

“4:1 is a good ROAS.” No study has ever established this — it appears uncited on every page that states it. Reconstructed, it’s just the break-even of a roughly 25%-margin business: if ads may take a quarter of revenue, you need 4:1. Arithmetic folklore, useful only if your margin happens to be ~25–35%.

“The average ROAS is 2.87:1, per a Nielsen study.” There is no such Nielsen finding. The only real Nielsen ROAS benchmark (Nielsen Catalina, 2016) measured CPG campaigns from 2004–2015 and contains no 2.87 anywhere. Meanwhile Triple Whale’s 2025 cohort happens to produce a 2.87x mean — so the same number now circulates with two contradictory origin stories. Textbook citogenesis.

“Google Ads returns $2 for every $1 spent” (or “$8 in profit”). That’s Google’s own economic-impact math, resting on its chief economist’s estimate published in 2009. A seventeen-year-old modeling assumption, still quoted as a current benchmark — by pages ranking today.

The Shopping-ROAS-by-industry tables (automotive parts 12x, toys 11.9x…) come from Sidecar’s benchmark reports, which stopped publishing in 2021 after the company was acquired. The dataset is dead; the tables live on in “2026” articles.

The pattern matches what we found researching ad costs: a benchmark without a dataset name and a date is decoration.

What current data actually shows

Two sources publish ROAS data with a real methodology today. Triple Whale’s platform cohort (~35,000 ecommerce brands, 2025): median blended ROAS 2.04x, mean 2.87x, with platform-attributed medians of 1.93x on Meta and 3.68x on Google. Varos runs a data cooperative of several thousand connected ad accounts whose median-by-industry snapshots are what most “benchmark” articles re-table, usually a year late.

Two honest caveats and one lesson. The caveats: both samples skew DTC-ecommerce, and both count platform-attributed revenue (more on why that flatters the number below). The lesson: the real median is roughly half the 4:1 folk benchmark — and the gap between the 2.04 median and 2.87 mean means high-margin outliers drag every “average” up. If your account beats 2x blended, you’re above the median of the best available data; whether you’re profitable at 2x is your margin’s decision, not the benchmark’s. For context on our side of the fence: across the client campaigns Herdr manages, the average is 3.5x — and the spread behind that average is exactly the margin story this page is telling.

ROAS vs ROI, MER, and POAS

Four metrics get tangled here, and each answers a different question. ROAS (revenue ÷ ad spend) asks: did the ads bring revenue? ROI (profit ÷ total cost) asks: after product costs, fees, and the ads themselves, did we make money? A campaign can carry a healthy ROAS and a negative ROI — that’s the margin gap again. MER, or blended ROAS (total revenue ÷ total ad spend across everything), asks: is the whole marketing engine efficient? It’s the sanity check against platform numbers, because platforms attribute generously — summed platform-reported revenue routinely exceeds what actually landed. POAS (profit ÷ ad spend) is the newest label, pushed mostly by tool vendors; it’s a fine metric, but break-even ROAS gets you the same correction with a division you can do on paper.

The part no ranking page tells you: platform ROAS flatters itself

Your ads manager’s ROAS number has two systematic distortions, and 2025–2026 measurement work put numbers on both. First, attribution: platforms claim conversions that would have happened anyway. Across 225 DTC geo-lift tests (Stella, August 2024–December 2025), the median incremental ROAS of branded search was 0.70x — below break-even — and retargeting ROAS was overstated 40–70%, with roughly 60% of retargeting conversions happening regardless of the ads. Second, signal loss cuts the other way: since Apple’s tracking changes, platforms also under-report some genuinely incremental conversions.

The practical rule that survives both distortions: use platform ROAS to compare campaigns against each other, use MER to judge the account, and be most suspicious of beautiful ROAS numbers on branded search and retargeting — that’s usually your own demand being sold back to you. Incrementality testing has gone mainstream (52% of US brand and agency marketers ran it as of a July 2025 TransUnion survey) precisely because scaling decisions made on platform ROAS alone kept burning money.

Setting a target the bidding can actually use

Once you have break-even plus cushion, that number goes into the platform, not on a poster. Google’s Target ROAS bidding wants at least 15 conversions in the last 30 days to work with, and it will chase exactly the number you set — which is why feeding it a benchmark-table target instead of a margin-derived one automates the wrong goal at scale. Two refinements worth stealing from how we run paid accounts: set targets per campaign, not account-wide (a shopping campaign and a branded search campaign deserve different math), and watch new-customer ROAS separately — an account can hit any blended target by quietly farming existing customers.

Common questions

Is a 3x ROAS good?

It depends entirely on your gross margin. At 30% margin, 3x is below your 3.3x break-even — you’re losing about ten cents per ad dollar. At 50% margin, 3x banks fifty cents per dollar and is comfortably good. Against the market, 3x beats the 2.04x median in the best current cohort data (Triple Whale, 2025) — but “better than median” and “profitable for you” are different questions.

What does a 4:1 ROAS mean?

Four dollars of tracked revenue for every dollar of ad spend — platforms display it as 400%. What it doesn’t mean is four dollars of profit: at a 25% gross margin, a 4:1 ROAS is exactly break-even, returning your ad dollar and nothing more. Always translate the ratio through your margin before celebrating it.

What is a good ROAS for Facebook ads?

The best current cohort data puts the Meta median at 1.93x platform-attributed (Triple Whale, ~35,000 brands, 2025) — lower than Google’s 3.68x, which is normal: Meta interrupts, Google catches intent. Whether 1.93x is good is still your margin’s call — it only clears break-even above roughly a 52% gross margin, and loses money below. Meta’s costs also moved this year; we’ve broken those down in how much Facebook ads cost.

Is a higher ROAS always better?

No — past a point, a high ROAS means you’re underspending. ROAS naturally falls as budgets scale into colder audiences, so maximizing the ratio pushes you toward tiny budgets, branded search, and retargeting: the traffic that converts anyway. The goal is the most profit, which usually lives at a lower ROAS and a higher spend than the vanity optimum. A 6x on $2,000 a month earns less than a 3x on $20,000 if your break-even is 2.5x.

Why does my ROAS drop when I increase the budget?

Because the auction serves your cheapest conversions first. More budget reaches colder audiences, and smart bidding relaxes toward your target rather than beating it — so some decay is expected, not broken. The question is where it lands relative to break-even: scaling from 5x to 3.5x at a 40% margin (break-even 2.5x) is a bigger business, not a worse one. If it lands below break-even, that’s your scale ceiling at current conversion rates — which is a conversion problem before it’s a bidding one.

Figures on this page were checked on July 11, 2026, against the named sources — including tracing each circulating benchmark to its origin. We re-check the data blocks quarterly and mark real updates with a visible date. For your own break-even, skip the benchmarks entirely: the calculator takes a minute.