ROAS vs ROI is not a matter of taste, because the two numbers answer different questions and only one of them mentions profit. Return on ad spend is revenue divided by ad spend. Return on investment is profit divided by everything the investment cost. I once reported 4.7x across a full year of ecommerce work. I still could not tell the owner his return on investment. Half the inputs sat on the client’s side of the wall.
This is for you if somebody has asked you to prove the advertising made money.
What ROAS vs ROI actually compare
Two formulas, and the distance between them is the entire argument.
Return on ad spend is revenue divided by ad spend. It is a ratio, it stops at the media, and anybody with a login can produce it in about four seconds.
Return on investment is profit divided by total cost, usually written as a percentage. It needs cost of goods sold, shipping, payment processing, refunds, discounts, and whatever you pay the person running the campaigns.
So ROAS vs ROI is partly a question about access. One figure lives inside the ad platform. The other lives in the accounts, and most media buyers never get to see it.
That is not a small technicality. It explains why one of these appears in every weekly report and the other appears almost never.
ROAS is a media metric, ROI is a business metric
Both are correct. They answer different questions, and those questions have different owners.
Return on ad spend answers an allocation question: which campaign, channel or creative should take the next $1,000? For that job it is exactly the right tool, because the comparison is like for like.
Return on investment answers something else entirely: should this business be spending on advertising at all, at this scale, this quarter?
A 3.0 return on ad spend is a fact about the media buy. Whether it counts as good news depends on gross margin, and margin is the one thing an ad account has never contained.
Across my ecommerce clients gross margin runs from 20% to 50%. Break-even therefore sits somewhere between 2.0x and 5.0x. That same 3.0 is comfortable at one end of the range and quietly loss-making at the other.
The 4.7x ROAS I could not turn into an ROI
An EU fashion ecommerce client, full year 2022, Meta and Google combined, with the majority of sales coming from Google Search and Shopping.
The account returned 4.7x across 1,715 orders at a cost per order of €14.18. Those are figures I can evidence, and I date them every single time I use them.
What I never had was the other half of the sum. No cost of goods, no return rate, no shipping subsidy, no payment processing percentage.
So the report said 4.7x and listed, in the same paragraph, everything that figure excluded. Nobody asked me to close the gap across twelve months of monthly reporting, which is common enough to be worth saying plainly.
A media buyer can hand you a defensible ROAS, but only the business itself can turn it into an ROI.
What sits between the two numbers
Six line items, and any one of them can be worth several points of margin.
Cost of goods sold is the largest and the most stable. Shipping and fulfilment come next, moving with basket size and destination. Payment processing looks trivial at two or three percent, right up until you set it against a twenty percent margin.
Then come the ones that swing week to week: discount codes, refunds, and the fees paid for media management, creative and tooling.
A store can lose the whole difference between a 4x and a 3x inside that list. Nothing has to change in the ad account for it to happen. Which is why a dull monthly reconciliation beats a clever dashboard.
Returns are the gap nobody reconciles
Refunds hide better than any other cost, mostly because of who handles them.
On every ecommerce account I have run, returns are managed by the client’s own in-house team. They never reach the reporting a media buyer sees. So the buyer optimises to a revenue figure that finance will later reduce. Nobody joins the two up.
Salesforce, on 8 January 2026, put holiday returns at $181B, or 14% of purchases, up 10% year on year. That data covered 1.5 billion shoppers across 89 countries. Their sample skews towards enterprise retail, so read it as a scale marker rather than your own rate.
No credible primary source exists for return rate by category, apparel included. That absence is why most profit estimates in ecommerce are guesses carrying a decimal point.
When a broken pixel makes both numbers worthless
One US automotive equipment ecommerce account taught me this the expensive way.
Its traffic data looked clean and unremarkable: an $8.28 CPM, a $0.28 CPC, and a 2.92% click-through rate across 43,415 link clicks. The conversion side was not clean at all. Reported conversions sat so far outside anything the business recognised that they were evidence of a tracking fault rather than evidence of performance.
Here is the trap. Take a corrupted revenue figure, apply a genuine margin to it, and you get a profit number that looks rigorous and is completely wrong. Precision is not accuracy. Both ROAS and ROI inherit whatever the tracking got wrong upstream.
Before anyone converts a media ratio into a profit statement, verify platform conversions against the store’s own order count for the same window. If those two disagree, that is the job for this week.
Choosing between ROAS and ROI for a bid target
Bid to the first and review against the second. That split settles most of this argument.
Smart bidding cannot read your margin, your refunds or your retainer. A revenue target is the only instruction the auction understands, so a target return on ad spend is what goes into the campaign.
Thresholds matter more than people expect. Google’s advertiser documentation sets them out plainly. Search and Shopping campaigns need 15 conversions in the past 30 days before a target return strategy becomes available. Demand Gen needs 50 in 35 days. Google also advises waiting four weeks, or three conversion cycles, after any reporting change.
Then the review happens outside the platform, monthly, with margin in front of you. When margin moves, the bid target moves with it. Most accounts never wire that loop up at all.
Reporting ROAS vs ROI without misleading anybody
Four things belong beside every figure: the window, the channels, the date range and the exclusions.
Triple Whale reports a platform median return on ad spend of 2.04 for 2024. Meta makes up 68.3% of the ad spend it tracks. That 2.04 arrives with no stated sample size or date range on the page where it sits. Avoid that habit in your own reporting.
My rule came out of being asked, twice in one quarter, whether a figure included Google. Now the channel mix goes in the sentence rather than the footnote.
Where ROAS vs ROI causes real trouble is the quiet upgrade: a media ratio presented in a board pack that is actually discussing profit. Label each number for what it measures, and let the finance team run the conversion.
What to change this week
Three things, and none of them takes long.
Ask for gross margin after discounts and returns, in writing, and record the date you were given it. If it never arrives, put that sentence in the report itself.
Then reconcile last month’s platform-reported orders against the store’s own order count for the same window. A gap beyond a few percent means the ratio you were about to convert is not solid enough to convert.
Last, add one line to your reporting template listing what the figure excludes. It costs a minute and it prevents the conversation where somebody has spent six months reading a media ratio as a profit number.
Cost per lead against cost per acquisition runs the same distinction in lead generation. The purchase costs I can date and evidence give the volume behind the 2022 figures above. Reconciling blended against platform reporting covers channel overlap. Total revenue over total spend is the account level version of the same idea. Organic revenue changes both sides of the sum, which is why my SEO work sits beside the paid accounts. To have your reporting checked against your actual margin, book a teardown.