Here is a small, extremely common disaster. Meta reports a 4.1x return. Google Ads reports 3.8x. The founder does the arithmetic on ad spend against revenue in Stripe and gets something closer to 2.2x, and then spends a week trying to work out who is lying.
Nobody is lying. Each platform is answering a question it was asked, and the question is not the one the founder cares about.
Every platform reports its own best case
A platform's reported conversions are the conversions it can attribute to itself, using its own attribution window, its own identity graph and its own definition of a touch. Meta counts a purchase it can tie to an impression or click within its window. Google does the same. The customer who saw a Facebook ad on Tuesday, searched your brand on Thursday and bought on Friday is claimed, entirely honestly, by both.
Sum those dashboards and you are double-counting on purpose. The total is not a measurement of your business; it is the sum of two marketing departments' self-assessments.
Three numbers, and what each is for
You need all three, and you need to stop using them interchangeably.
Platform-reported ROAS is a steering metric for that channel. It is good at relative questions: is this ad set better than that one, is this creative fatiguing. It is bad at the absolute question of whether the business made money.
Blended ROAS is total revenue over total ad spend, across every channel, over the same window. It cannot tell you which channel earned the revenue. It can tell you whether advertising as a whole is working, and it is the number that agrees with your accountant.
Marketing efficiency ratio is the same idea widened to all marketing cost, not just media. It is the one to watch when your spend is not just ads.
The useful discipline: steer with platform numbers, judge with blended ones. When the two disagree sharply, that gap is information rather than an error — it usually means your channels are claiming the same customers.
Break-even is the line, not a benchmark
A 3x ROAS is either excellent or a slow bankruptcy depending on your margins. The only threshold that means anything is your own break-even ROAS: the return below which every incremental sale loses money, computed from average order value, cost of goods and the fixed costs the campaign has to cover.
Work it out once — our calculator takes a minute — and then use it everywhere. It changes what "good" means, and it usually changes it a lot.
What reading across channels actually requires
Once you accept that no single platform can answer the question, the work becomes plumbing:
- Get the channels into one view. Marketing OS reads Meta and Google Ads alongside GA4 and Search Console, so paid and organic sit in the same picture instead of two tabs and a memory.
- Tag everything, consistently. Analytics can only group traffic it recognises. Most attribution problems are really a casing-and-spelling problem in UTM parameters — the tag checker exists because this mistake is so reliably the cause.
- Check the measurement itself. Before you interpret a number, find out whether it is being collected properly: is the pixel firing on the pages that matter, is server-side tracking deduplicating against it, are conversions landing where you think. Our measurement health check grades that, and it is honest about its limits — Meta exposes no event match quality at any scope, so no tool can tell you that number, including ours.
- Judge on a window, not a day. Attribution windows lag. Comparing yesterday's spend to yesterday's revenue produces panic on a schedule.
What to change on Monday
Put your blended ROAS next to your break-even ROAS on one screen, weekly. Keep the platform dashboards for steering campaigns, where they are genuinely good. And stop adding the dashboards together — that number has never once been true.