The ROAS Trap: Why Your Best-Performing Campaigns Are Quietly Destroying the Business

The campaign your team is proudest of is, statistically, the one most likely to be losing you money. Not because the targeting is wrong or the creative is weak — but because the number everyone is celebrating was never designed to tell you whether you made any money. Return on ad spend measures revenue per advertising dollar. It says nothing about what is left after the product, the box, the card fee, the return, and the discount you used to win the order in the first place.

This is not a small accounting footnote. It is the single most common reason we see otherwise healthy direct-to-consumer brands scale revenue while quietly running out of cash — and it almost always traces back to one decision: optimising the entire growth engine against a top-line efficiency ratio that has no relationship to profit.

We want to be precise about this, because the fix is not “ROAS is bad.” ROAS is a perfectly good diagnostic in the right place. The problem is using it as a north star — the number you scale toward, set targets against, and reward your media team for hitting. Do that, and you will reliably scale the campaigns that are destroying contribution margin fastest, because those are precisely the campaigns that look best on a metric that ignores margin.

01 / THE MEASUREMENT PROBLEMWhat ROAS actually measures — and what it hides

ROAS is revenue divided by ad spend. A 4.0× ROAS means four dollars of revenue for every dollar of media. The instinct is to read that as “four dollars back for every dollar in,” but that four dollars is gross revenue at the top of the P&L — it is not margin, and it is certainly not profit. Between that revenue line and the money that actually clears into the business sits a stack of costs that ROAS is structurally blind to.

The discipline that fixes this is contribution-margin accounting — building the unit economics in layers so you can see what each order is worth at each stage. We work in three layers:

The contribution stack

Layer	Definition	What it subtracts
CM1	Revenue − cost of goods	Product cost, inbound freight, duties
CM2	CM1 − variable post-purchase costs	Pick-pack, outbound shipping, payment processing, returns
CM3	CM2 − variable marketing	Customer acquisition cost (the ad spend itself)

CM2 is the number that matters most for media decisions, because it is the contribution available to pay for acquisition and then leave a profit. CM3 is what is actually left after you have bought the customer. A campaign can post a 4.0× ROAS and a positive CM3, or a 4.0× ROAS and a deeply negative CM3 — the ROAS is identical in both cases. That is the whole problem in one sentence: ROAS is a constant across wildly different profit outcomes.

02 / THE UNIT ECONOMICSFollowing a single order down the waterfall

Let us make this concrete with an illustrative brand — representative of the skincare and supplement businesses we work with most. Average order value is $82. Cost of goods runs 28% of revenue. Variable post-purchase costs — fulfilment, shipping, processing, and a realistic return rate — come to another 18%. That leaves a CM2 of $44.28, or 54% of revenue, as the pool available to acquire the customer and still make money.

At a 4.0× ROAS, the acquisition cost per order is $82 ÷ 4 = $20.50. Subtract that from CM2 and the order delivers $23.78 of CM3 — 29% of revenue dropping to contribution. That is a genuinely healthy order. The waterfall below traces exactly where the money goes.

From order value to contribution — the unit-economics waterfall

Illustrative DTC order · $82 AOV · CAC at a 4.0× ROAS

Revenue Cost deducted Contribution pool Profit after acquisition

03 / THE MARGINAL TRAPWhy your best campaign breaks first

Here is where the trap closes. The campaign posting your highest ROAS is, almost by definition, the one reaching your warmest, highest-intent audience — retargeting, branded search, your most efficient prospecting pocket. It looks like the obvious place to add budget. So you do.

But that audience is finite. As you push more spend into it, you exhaust the cheap, high-intent impressions and start paying to reach colder, less-likely buyers. The marginal cost of each additional customer rises, and the marginal ROAS on each new dollar falls — fast. Meanwhile the number on your dashboard is blended ROAS: the average across every dollar you have spent, including all the cheap early ones. The average is heavy with history. It moves slowly. So it keeps reading “healthy” long after the marginal dollar has gone underwater.

You do not feel the trap, because you are watching the average, not the margin. The chart below shows the divergence on our illustrative brand, whose breakeven sits at a 1.85× ROAS — the point where CAC exactly equals CM2 and CM3 falls to zero.

Blended ROAS hides the marginal dollar going negative

ROAS vs. audience saturation · breakeven = 1.85× for this brand

Blended ROAS (the dashboard number) Marginal ROAS (the real decision) Value-destruction zone

04 / THE REPLACEMENT METRICPOAS: the number that doesn’t lie

The fix is to optimise against the metric that ROAS only pretends to be. Profit on ad spend — POAS — divides the contribution generated (CM2 attributable to the campaign) by the ad spend that generated it. A POAS of 1.0 is breakeven on contribution; above 1.0 the campaign is funding profit, below 1.0 it is consuming it. Unlike ROAS, POAS moves when your margin moves, which means it actually reflects the business.

The most useful translation we give clients is the breakeven ROAS implied by their margin structure — the ROAS below which a campaign is guaranteed to be CM3-negative. It is simply one divided by your CM2 margin. Once a team internalises this single number, the entire ROAS conversation changes, because a target ROAS without a margin model behind it is just a number someone picked.

Breakeven ROAS by contribution margin

CM2 margin (contribution before marketing)	Breakeven ROAS	Implied max CAC on $82 AOV
30%	3.33×	$24.60
40%	2.50×	$32.80
50%	2.00×	$41.00
54% — our example brand	1.85×	$44.28
60%	1.67×	$49.20
70%	1.43×	$57.40

Read this table the way a CFO would. A brand with 30% contribution and a brand with 60% contribution are playing entirely different games: the first cannot survive below a 3.33× ROAS, the second prints contribution all the way down to 1.67×. A blanket “keep ROAS above 3” rule simultaneously starves the second brand of profitable volume and bankrupts the first — which is exactly why blanket ROAS rules fail.

05 / THE NUANCEBut profit isn’t always the goal

Here is where most “ROAS is dead” takes go wrong: they assume every brand should optimise for profit on every dollar, always. That is not how disciplined growth works. The right objective — and therefore the right metric — depends on the company’s stage and its access to capital. Optimising for profit when the business is funded to grab a category is as much a mistake as optimising for ROAS when the business needs to make money.

The decision is genuinely strategic. What is your north star right now? A well-funded brand in a land-grab phase may rationally accept thin or even negative first-order CM3 to win market share — provided it has a modelled, believable LTV and a defined payback window that turns those customers profitable on a horizon the balance sheet can fund. A brand optimising for profit should be managing to CM3 and POAS, ruthlessly. The metric follows the goal; it does not lead it.

Matching the metric to the stage

Stage & capital position	Primary objective	What you optimise toward
Early / well-funded land-grab	Revenue & market share	Payback window & LTV:CAC, with a CM3 floor
Scaling / disciplined growth	Profitable growth	POAS ≥ target, blended & marginal CM3
Mature / profit-focused	Operating profit	CM3 maximisation, marginal POAS, mix efficiency

The connecting thread is the one principle we refuse to bend on: LTV over CAC. The only thing that justifies a thin or negative first-order CM3 is a second, third, and fourth order you can actually model. Negative CM3 with no credible repeat behaviour behind it is not a growth strategy — it is a subsidy you are paying to your own customers, and it ends the moment the funding does.

06 / THE OPERATING SYSTEMTest small, scale winners, kill losers

None of this works without a testing discipline that evaluates on the right metric. The instinct in most accounts is to test against ROAS and scale whatever wins. But a campaign that “wins” on blended ROAS while losing on marginal CM3 is exactly the loser you should be killing — and the ROAS framing tells you to do the opposite.

The doctrine we run is simple to state and hard to hold: test concepts small, fast, and at the lowest spend that still produces a statistically meaningful read; evaluate every test on contribution, not on revenue efficiency; scale only what is POAS-positive (or payback-positive, if you are in a funded growth phase with a modelled LTV); and kill the rest without sentiment. The discipline is the point. You never scale a losing machine — and ROAS is very good at disguising losing machines as winners.