Blended ROAS divides every pound of revenue the business took by every pound it spent on advertising. Platform ROAS divides the revenue one ad account claims by what that account spent. Those two questions have different answers, and the gap between them is where most reporting arguments live.
Between June and August 2026 I ran Meta sales campaigns for a US flag and banner ecommerce brand. Depending on where you draw the line, those three months produced 1.11, 1.25, 1.58 or 1.99. All four are true. Only one of them belongs in a decision.
This is for anyone who has watched a client’s finance team and their ad platform disagree.
What blended ROAS measures that a platform figure cannot
It measures the business rather than the ad account.
Platform ROAS answers a narrow question. Of the revenue this platform can see and can claim, how much came back per unit of spend? Every part of that sentence is doing work.
The blended version ignores claiming entirely. Total revenue over total advertising spend, taken from the shop’s own reporting, with no attribution model anywhere in the calculation.
So it is duller and far harder to argue with. Nobody debates whether the bank received the money.
The trade is that it tells you nothing about which channel did the work. You get an honest total and no diagnosis at all. So the two figures belong next to each other rather than in competition.
Four numbers from one account in one quarter
Here are the four figures, in the order they tend to get quoted.
The account returned 1.11 when the reporting mixed attribution windows, which the platform does by default. Across all campaigns including awareness it returned 1.25. The six sales ad sets together returned 1.58. Remove one failed location test from those ad sets and the figure becomes 1.99.
Same account, same three months, same shop revenue underneath. The lowest and the highest are almost a factor of two apart, and that spread is entirely a definition problem.
Behind the 1.58 sit 64 purchases at $67.24 each. Behind the 1.99 sit 61 purchases at $53.72, on a $12.78 CPM and a $0.63 CPC.
The attribution window did the first piece of damage
The 1.11 is the figure I trust least, and it is the one most likely to end up in a screenshot.
Meta blends attribution windows when you read an account level total across campaign types. So the number at the top of the screen is an average of measurements taken with different rulers.
Change the attribution window and the attributed revenue changes with it. The money in the bank does not move. That is the tell. Any metric that responds to a reporting setting is describing the report rather than the business.
Blended ROAS is immune to this by construction, because it never asks the platform what it thinks it earned.
Awareness campaigns change the denominator, not the sales
Including awareness campaigns took the account from 1.58 down to 1.25.
Nothing about the sales campaigns changed. Awareness spend entered the denominator and brought almost no directly attributed revenue with it. That is precisely what awareness campaigns are built to do.
So both figures are correct, and they answer different questions. The 1.58 describes how the sales campaigns performed. Total Meta investment over the same weeks is what the 1.25 describes.
Comparing one quarter’s 1.58 against the next quarter’s 1.25 is the classic mistake. It happens when nobody notices that the second figure includes a campaign type the first did not. That comparison has caught out better analysts than me.
The ad set I excluded, and whether excluding it is honest
One ad set tested a neighbouring state and lost money at a rate I have rarely seen.
It returned 0.25, with three purchases at roughly $341 each on about $1,027 of spend, before I switched it off. Here is the part worth keeping. That ad set had the highest reach in the account at 17,086, and the most link clicks at 542.
Every traffic metric said it was the best thing running. Revenue said it was the only real loser.
Reporting 1.99 while hiding the test would be dishonest. Reporting 1.58 without saying that a stopped experiment sits inside it is misleading in the other direction. So I report both figures with the reason attached, which takes one extra sentence.
Blended ROAS when one of the three channels is email
This is not really a Meta account. It is a business with three channels, and I run all three.
Meta buys the impressions. Google Ads catches demand that already exists. Klaviyo sends email, which costs almost nothing at the margin and converts people the ads brought in months earlier.
So a Meta figure of 1.99 can look thin in isolation while the blended ROAS across all three sits somewhere comfortable. Total revenue over total spend is the marketing efficiency ratio in everything but name. It answers the question the owner is actually asking.
Meta also accounts for 68.3% of the ad spend Triple Whale tracks across its brands. For most ecommerce businesses the Meta figure gets treated as the whole picture. It is a large share of the spend and it is still not all of it.
Why I will not call this blended ROAS healthy or unhealthy
The client has never disclosed gross margin to me, so no verdict is available.
Break even sits at one divided by gross margin. Across my ecommerce clients that margin runs between 20% and 50%. That puts the floor anywhere from 5.0x down to 2.0x, so a 1.99 could be comfortably profitable or badly underwater. I genuinely do not know which.
That is an unsatisfying thing to publish. It is also more honest than the alternative, which is quoting a return figure as though it were a profit statement.
Plenty of case studies do exactly that. A ratio becomes a verdict without anybody mentioning the margin that would turn one into the other. So treat the blended ROAS here as a measurement, and leave the verdict to whoever knows the cost of goods.
How to build a blended ROAS number you can defend
Four rules, and the fourth is the one people skip.
Take revenue from the shop platform rather than any ad account, net of refunds where the platform will give you that.
Then take every pound of advertising spend, including the lines nobody counts. Agency fees, email tooling, affiliate commission, and any retainer that exists to produce sales.
Third, use a period long enough to cover the purchase cycle. Weekly is usually the shortest sensible window, and monthly is safer for considered purchases.
Fourth, write the definition down and never quietly change it. A blended ROAS that means something different each quarter is worse than having no number. It will be trusted, and it will be wrong.
What the platform figure is still good for
Platform reporting is a good ranking tool and a poor totalling tool.
Inside the flag account the best ad set returned 4.84 at $29.37 per purchase. The worst active one returned 0.98 at $93.96. Both were measured the same way, in one account, over the same weeks, so comparing them is legitimate.
That is the correct use. Rank ad sets against each other and move money towards the top of the list. Then resist adding the column up and presenting the total as truth.
Haus, which sells geo experimentation and so has an interest in the answer, published research on 13 August 2025 covering 640 Meta experiments. Those brands spent an average of $14 million a year on Meta, and every $100 of platform attributed revenue came with $115 of incremental revenue. Their argument runs the opposite way from mine, and both can hold at once. Platform totals can be wrong in either direction.
What to change this week
Three steps.
Pull total revenue from the shop platform for the last complete month. Then pull total advertising spend from everywhere it hides, and divide one by the other. That single figure is usually a surprise to somebody.
Put it beside the platform number in the same report, labelled clearly. Nobody should have to ask which one they are reading.
Then write your definition at the bottom of that report. What counts as revenue, what counts as spend, and over what period. Next quarter’s comparison depends entirely on that paragraph existing.
Why the platform’s reported cost never matches the client’s covers the same disagreement on the lead generation side, the purchase costs I can show with volumes attached gives dated numbers worth comparing against, the floor to calculate first is the arithmetic that turns any ratio into a verdict, and why industry tables cannot be trusted explains why borrowing somebody else’s figure fails. On Meta this is a measurement job before it is a media one, which my Meta Ads work treats as the starting point. To have both numbers built properly for your account, book a teardown.