ROAS is the number performance marketers live by. Revenue credited to your advertising divided by what that advertising cost. Spend €1,000, get credited with €4,000 in sales, and you have a ROAS of 4 — Google writes that as 400%, while Amazon flips it upside down into a 25% advertising cost of sales.
But ROAS is also an instruction. Since Google launched Target ROAS on October 3, 2013, advertisers have been able to hand a target to an algorithm that bids in every auction to hit it. Facebook followed with minimum ROAS bidding in 2018. Today the dial sits inside search, social, retail media and programmatic buying.
Three dials build the number
Conversion value comes from the revenue passed with each sale — through a site tag, Meta’s Conversions API or Google’s Floodlight tags. Whether that revenue includes tax, shipping or returns depends on your setup.
Attribution decides which sales the ads get credit for. Meta defaults to seven-day click or one-day view windows; IAB Europe’s 2024 retail media standards recommend a 30-day lookback and demand that click-based and view-based ROAS be reported separately.
Cost is normally just media spend. Agency fees, creative and data charges usually sit outside the calculation.
The bidding maths is simple once you see it. A click with a 2% chance of leading to a €250 order has an expected value of €5. At a 400% target the system can pay up to €1.25 for that click; at 500%, only €1. Raising the target makes bidding more selective and usually cuts volume.
The seller grades its own homework
Here’s the uncomfortable part. You set the target; the platform measures the outcome and usually owns the attribution too. Meta’s suite of truth research found rules-based attribution misallocated a median 31% of conversions. A former Meta product manager alleged in a London tribunal that Shops ads ROAS was inflated by 17% to 19% by counting shipping and taxes as revenue. And Zalando research showed a marketing mix model reporting paid search ROAS at 10.61 against a true value of 4.20.
None of this makes ROAS useless. It makes it a grading curve rather than a truth machine.
Margin decides what the number means
A retailer with a 40% gross margin breaks even at a ROAS of 2.5 before other costs. That means two advertisers reporting the same ROAS can sit on opposite sides of profit. Revenue-based targets also favour expensive products over profitable ones — which is exactly why Meta started testing profit-based ROAS optimisation using margin data sent through the Conversions API.
And averages hide saturation. Google’s Meridian documentation warns that a marginal return well below the average signals you are hitting diminishing returns. Cutting spend can lift ROAS while total profit falls.
What changed in 2026
- On August 17, Google’s budget-constrained Target ROAS and Target CPA campaigns began steering toward the stated target instead of beating it.
- In September, Google introduced Qualified Future Conversions, counting conversions up to 180 days after a click.
- Amazon tightened view attribution on January 1, splitting ROAS reporting into a standard figure and an “all views” family.
- Dreamdata’s B2B benchmarks put LinkedIn at 121%, Google Search at 67% and Meta at 51% across buyer journeys averaging 272 days.
Use ROAS as a dial, not a verdict
Work out your break-even ROAS before you set any target. Audit what enters your conversion value — tax, shipping and returns all move the number. Then run incrementality tests against an unexposed control group so you know what the ads actually caused, not just what they were credited with.
Les Binet told Google Marketing Live in Sydney in June 2026 that “budget is nine times more important than ROI” in driving profit. ROAS tells you how efficiently a slice of spend is working. It does not tell you whether that slice is big enough — or profitable enough — to grow on.
Source: PPC Land
