Every ad platform you use is designed to make its own performance look good. That is not a conspiracy, it is simply how the business works. The platform owns one half of the attribution window and almost none of the margin calculation. The result is that reported ROAS can look healthy at the same time as the business is losing money on paid acquisition.
Understanding whether your advertising is actually profitable is one of the most consequential questions in ecommerce. Here is how to approach it in a way that holds up to scrutiny.
What reported ROAS leaves out
The ROAS you see in Meta Ads Manager, Google Ads, TikTok Ads Manager, or Amazon Advertising is Revenue Divided By Ad Spend. The revenue is almost always the attributed order value, counted at the point of order placement, including every cost that is going to come out of it later.
The honest version of ROAS, the one that answers "did this spend earn money," has to deduct every cost that scales with each order:
- Returns and refunds attributable to the same orders.
- Platform fees on each order (payment processing, referral, FBA, chargeback).
- Shipping and fulfilment cost per order, net of what the customer paid.
- Cost of goods for the SKUs sold.
- Promo and discount applied against the order.
What is left is the contribution margin of the acquired revenue. True ROAS is contribution margin divided by ad spend, and it answers the real question: for every dollar of ad spend, how much contribution margin came back.
A simple true ROAS formula
True ROAS equals ((Attributed revenue minus returns minus platform fees minus shipping and fulfilment minus COGS minus promo discounts) divided by Ad spend).
For many brands, a reported ROAS of 3 reconciles to a true ROAS of less than 1 once every cost is accounted for. That does not mean the campaign is a failure. It means the question "is this campaign profitable" has a quantitative answer that is different from the one the ad platform is showing you.
Last click versus multi touch
The second honesty check is attribution. Every ad platform credits itself when it sits inside the attribution window. If a shopper sees a Meta ad, then clicks a Google search, then opens a TikTok reel, then buys, three platforms will claim the sale.
There is no single correct attribution model. There are useful ones. A pragmatic combination for most ecommerce teams:
- Measure last click or last platform touch for operational decisions, because that is what each platform optimises against.
- Measure first touch separately, so you can see where new customers are actually discovering you.
- Measure a weighted multi touch model monthly for budget decisions, because that is the model closest to the full acquisition journey.
- Reconcile the three views against your real new customer rate every month. If the models disagree with the real customer growth, something in the attribution model is wrong.
An illustrative example
A brand running Meta and Google ads sees a 3.2 reported ROAS on Meta and a 4.1 reported ROAS on Google. True ROAS, after returns and fees, is 0.9 on Meta and 1.4 on Google. One platform is losing money per acquired customer. The other is paying for itself but barely. A reasonable response is to cut Meta spend by half, hold Google steady, and invest the saving in a retention program that lifts repeat purchase rate on the customers both channels have already paid to acquire.
None of this is possible with the reported ROAS alone.
What to measure every month
A simple monthly report that will keep most brands out of trouble:
- True ROAS per platform.
- True ROAS per campaign, for campaigns above a spend threshold.
- Blended true ROAS across all paid acquisition.
- New customer rate from paid versus organic versus returning.
- Payback period: how many months of margin from a new customer are needed to pay back their acquisition cost.
If one monthly report is enough to defend every budget decision for the next quarter, you have built the right one.