A 4.2 ROAS looks like a good month. It is also entirely compatible with a business that lost money, and the gap between those two readings is where a surprising number of eCommerce brands live for a year or two before anyone notices.
This is not a criticism of agencies specifically. ROAS is the number the ad platforms report, so it is the number that gets reported onward. But it is a marketing efficiency ratio being used as a business result, and those are different jobs.
ROAS answers: did the advertising work? Contribution margin answers: did the business make money? A report that only carries the first cannot support a spending decision.
Every one of these is a real cash cost of the order the advertising generated. None of them appears in the ratio.
The largest single omission, and it varies by SKU mix
Ad-driven traffic often returns at a higher rate than organic
Free delivery thresholds change this per basket
Return carriage and handling
Frequently larger than the outbound cost
Payment and platform fees
A few per cent, applied to every order
Depends entirely on whether revenue is gross or net
The first question to ask about any ROAS figure is whether its revenue is gross or net of returns and discounts. Reports frequently do not say.
Contribution margin, defined once
Contribution margin is what is left from an order after every cost caused by that order. Not after overheads — after the costs the order itself created. The stack is short and every line is knowable.
Start with net revenue: gross revenue minus discounts and minus returns.
Deduct cost of goods for what was actually kept, not what was shipped.
Deduct outbound carriage, return carriage and the handling cost of processing a return.
Deduct payment processing and, on marketplaces, referral and fulfilment fees.
What remains is contribution. Advertising is spent from this, not from revenue.
Net
After discounts and after returns, or the ratio is fiction
Per order
Basket mix moves margin more than channel does
Definitions we standardise on before reporting on any account. Deliberately conservative rather than flattering.
The report to build instead
You do not need to abandon the channel view — it is genuinely useful. You need to add four columns to it, and then read the last one.
Store, after returns and discounts
Net revenue minus order costs
What was available to pay for advertising
Contribution minus ad spend
Whether this channel funded itself
The fourth column is the one that changes decisions. It regularly reorders a channel league table that ROAS had looked settled on.
Blended and channel views do different jobs
Teams argue about this as though one has to win. They answer different questions and you need both, provided each is used for its own question.
- Blended contribution after ads tells you whether the business is working. It cannot be gamed by attribution settings, which is exactly why it is the number to govern by.
- Channel-level contribution tells you where to move budget next. It is attribution-dependent, so treat it as directional rather than exact.
- Incrementality tests settle arguments the other two cannot. Expensive, slow, and the only honest answer to “would this have happened anyway?”.
The useful reporting argument is not blended versus channel. It is which question each view is allowed to answer.
Break-even ROAS is a moving number
If you must keep a ROAS target, at least derive it rather than inherit it. Break-even ROAS is one divided by your contribution margin percentage — and that percentage moves constantly.
It moves with product mix. A promotion that shifts volume toward low-margin lines lowers the whole target without anyone changing a setting.
It moves with return rate. Apparel in January is a different business from apparel in June, and one target across both is wrong twice.
It moves with carriage and fee changes. A courier increase or a marketplace fee change quietly moves your break-even and nothing in the ad account notices.
“The most uncomfortable meeting we have with a new client is the first one where their best-performing channel turns out to be their least profitable. It is also the most useful.”
Dazzle Commerce · eCommerce Management team
Rebuilding reporting around contribution usually needs finance and marketing data in the same place. It is one of the first things we set up on a management engagement.
Questions to ask about any report
Five questions, none of them hostile, that establish quickly whether a report can support a decision:
Is the revenue in this report gross or net of returns and discounts?
Which costs are deducted before the number I am looking at?
What attribution window and model produced the channel split, and what happens to it at a different setting?
What is our current break-even ROAS, and when did it last change?
If we switched this channel off for two weeks, what does this report predict would happen — and have we ever checked?
WHAT A GOOD ANSWER SOUNDS LIKE
A good agency will answer all five without defensiveness, and several answers will be “we do not have that yet”. That is a fine answer. “ROAS is the standard metric” is not.