Why platform ROAS overstates profit
Platform ROAS divides the revenue an ad platform attributes to itself by media cost. It leaves out cost of goods, shipping, payment fees and refunds, it counts sales inside the platform’s own attribution window, and it does not say whether the sale would have happened without the ad. In the worked model below, an account reporting ROAS 4.0 makes money if every attributed sale is new and loses money if 60% of them are. Budget decisions should rest on contribution after media, measured against a holdout.
Executive findings
- 1
Google’s Target ROAS strategy bids towards the conversion value you report to Google Ads. Margin enters the bidding only if the values you send are margin-based.
Documented: official source. Scope: Google Ads Search and Shopping bidding documentation, as of . [1]
- 2
Google Ads now uses data-driven attribution for conversion actions that were on the retired rule-based models; last click remains available. The attribution model decides how much credit each ad gets, so the same sales can produce different ROAS under different settings.
Documented: official source. Scope: Google Ads attribution documentation, as of . [2]
- 3
Meta’s standard attribution counts website conversions within 1 or 7 days of a click and within 1 day of an impression, and Meta also offers an incremental attribution model. A view-through purchase counted by Meta may also be counted by Google.
Documented: official source. Scope: Meta Ads Manager attribution settings, as of . [3]
- 4
In our illustrative model, an account with platform ROAS 4.0 earns a positive contribution after media only if more than about 68% of attributed revenue is incremental.
Illustrative model, not a benchmark. Scope: FOCUSADS worked model with labelled assumptions, not a benchmark, as of
Method and limitations
- Question
- Can a business use platform-reported ROAS to decide whether paid media is profitable and how much more to spend?
- Written for
- CFOs, ecommerce directors and performance leads who approve or defend monthly media budgets.
- Decision at stake
- Whether to raise, hold or cut media spend, and which product groups to stop advertising.
- Sources
- Official Google Ads and Meta help-centre pages on bidding, attribution and lift measurement, and IAB guidance on incremental measurement in commerce media. All read on the access date shown.
- Period
- Documentation as published on 17 October 2026. Platform settings change; recheck before applying.
- Markets
- Applies wherever Google Ads and Meta ads are sold. The model is currency-neutral.
- Definitions
- Platform ROAS: conversion value attributed by the ad platform ÷ media cost in that platform.
- Contribution margin: revenue kept after returns, minus cost of goods, shipping, payment fees and return handling.
- Incremental share: the part of attributed revenue that would not have happened without the ads, measured by a holdout or lift test.
- Limitations
- The worked model uses round illustrative numbers. It shows the mechanism; it says nothing about typical returns or margins in any category.
- We have no access to non-public account data for this article and do not quote client results.
- Lift tests need enough conversions to read a difference. Small accounts may need longer tests or geographic holdouts.
Four things the platform figure leaves out
Documented: official source
First, cost. The platform sees the order value you send it. It has no line for the goods, the courier, the payment provider or the parcel that comes back three weeks later.
Second, the attribution window. Meta counts a purchase within 7 days of a click or 1 day of an impression under its default standard setting. Google assigns credit with data-driven attribution unless you choose last click. Each platform reports from its own vantage point, so one order can appear in both reports.
Third, refunds after the fact. Unless you adjust or retract conversion values, a refunded order stays in the platform’s revenue.
Fourth, incrementality. Attribution records that an ad was present before a sale. Google’s Conversion Lift and Meta’s incremental attribution exist because presence and cause are different questions.
Worked model: when ROAS 4.0 loses money
Illustrative model, not a benchmark
The table follows one month of a single ecommerce account. Every input is an assumption chosen to make the arithmetic easy to check.
On attributed numbers the account clears its break-even of 2.74 comfortably. Once only 60% of the revenue is treated as new, the break-even rises to about 4.57 and the same month shows a loss.
| Line | Amount | How it is calculated |
|---|---|---|
| Media cost | 20,000 | Assumption |
| Revenue attributed by the platform | 80,000 | Platform ROAS 4.0 × media |
| Refunds on returned orders | -12,000 | 15% of attributed revenue |
| Cost of goods on kept orders | -27,200 | 40% of 68,000 kept revenue |
| Shipping and payment fees | -9,600 | 12% of attributed revenue, paid on returned orders too |
| Return handling | -2,000 | Assumption |
| Contribution before media | 29,200 | 36.5% of attributed revenue |
| Contribution after media, if all sales are incremental | 9,200 | 29,200 - 20,000 |
| Contribution after media, if 60% are incremental | -2,480 | 0.6 × 29,200 - 20,000 |
| Break-even ROAS on attributed revenue | 2.74 | 1 ÷ 0.365 |
| Break-even ROAS at 60% incrementality | 4.57 | 2.74 ÷ 0.6 |
Assumptions
- Return rate 15% of attributed revenue; cost of goods 40% of kept revenue; shipping and fees 12% of attributed revenue.
- Incremental share 60% is a placeholder for a measured holdout result, not an estimate for any category.
- Fixed costs (staff, software, warehouse) are excluded; contribution must also cover them.
Break-even per product group
Illustrative model, not a benchmark
One account-wide target hides the products that lose money. A homeware group with a 55% contribution margin breaks even at about 1.8. An apparel group with 30% after returns needs about 3.3 before incrementality is considered.
Put a margin band and a return-risk label in the product feed, then split campaigns or Performance Max asset groups so a low-margin line is not bid with the target of a high-margin one. If you send margin-based conversion values instead of order values, Target ROAS starts optimising towards contribution directly.
| Line | Homeware | Apparel |
|---|---|---|
| Order value | 100 | 100 |
| Cost of goods | -30 | -40 |
| Shipping and payment fees | -10 | -12 |
| Expected returns cost | -5 | -18 |
| Contribution before media | 55 | 30 |
| Break-even ROAS | 1.82 | 3.33 |
Assumptions
- Both groups use an order value of 100 so the margins compare directly.
- Cost, fee and return figures are examples chosen to show the spread, not category averages.
- Break-even here assumes every attributed sale is incremental; divide by the measured incremental share as in the table above.
Where FOCUSADS draws the line
FOCUSADS view
We let platforms bid on the values they can see, because that is what their systems are built for. We set budgets on contribution after media from the shop or ERP, and we recommend a holdout before any large budget increase. When the owner cannot share margin data, the plan says so and reports revenue after returns as the closest available figure.
What we would test, in order
| Step | Measured by | Counts as success when |
|---|---|---|
| Reconcile 30 days of platform-attributed orders with shop or ERP orders by order ID. | Share of attributed revenue matched to a real, unrefunded order. | The unexplained gap sits within a tolerance agreed with finance before the test starts. |
| Send margin-based conversion values for one product group. | Contribution after media for that group against the previous period and a comparable group. | Contribution rises without a fall in new-customer orders. |
| Run a geographic or audience holdout for one full purchase cycle. | Difference in orders and contribution between exposed and held-out groups. | The measured incremental share is used to restate break-even ROAS for the next budget decision. |
Which number answers which question
Use the figure that matches the decision. Mixing them is how a profitable report and an unprofitable month coexist.
| Decision | Number to use | Source of truth |
|---|---|---|
| Daily bids inside a platform | Platform conversion value or ROAS | Platform, with margin-based values where possible |
| Monthly budget level | Contribution after media, adjusted by incremental share | Shop or ERP orders and returns, plus a holdout result |
| Which products to advertise | Break-even ROAS per product group | Feed labels built from cost and return data |
| Board or CFO reporting | Contribution after media and new-customer count | Finance-reconciled order data |
| Comparing two platforms | Incremental contribution per unit of spend | Lift tests on each platform, not their own ROAS |
How this connects to our method: Define economic reality before media spend
References
- About Target ROAS bidding, Google Ads Help, accessed
- About attribution models, Google Ads Help, accessed
- About attribution models and attribution settings, Meta Business Help Center, accessed
- About Conversion Lift, Google Ads Help, accessed
- Guidelines for Incremental Measurement in Commerce Media, IAB, accessed
- About the search terms report in Performance Max, Google Ads Help, accessed