ROAS is a bad target, and most accounts are optimised for it anyway

Every paid media dashboard leads with ROAS. It's the number clients ask about first and the one agencies report most proudly. It is also, on its own, close to meaningless.

Here's the problem in one sentence: ROAS measures revenue, and you cannot pay salaries with revenue.

The arithmetic nobody runs

Two campaigns, same account, same month.

Campaign A has the better ROAS and loses money. Campaign B looks worse on every dashboard in the industry and is the only reason the business is solvent.

This is not a contrived example. It's what happens in any account selling a mixed catalogue, which is most ecommerce businesses. The high-ROAS campaigns are very often the discount-driven, low-margin ones, because discounting is precisely what makes revenue rise faster than profit.

What to optimise for instead

Contribution margin after ad spend

The only number that answers "did this campaign make us money?"

Contribution margin = Revenue − COGS − Shipping − Payment fees − Ad spend

Feed it back into the platform as a conversion value and you change what the bidding algorithm is actually chasing. In Google Ads this means passing margin rather than order value in your conversion tracking. In Meta it means a value-optimisation campaign fed with margin as the value parameter.

Most accounts pass order value because that's the default. The default is optimising your budget toward whatever you sell most cheaply.

Break-even ROAS, calculated per product tier

Once you know margin, break-even ROAS falls out of it:

Break-even ROAS = 1 / contribution margin %

At 22% margin you need 4.5x just to break even. At 61% you need 1.64x. A single account-wide ROAS target applied across both tiers guarantees you're over-investing in one and starving the other.

CAC payback period

For anything with repeat purchase or subscription, ROAS on a first order is the wrong window entirely. What matters is how many months it takes for a cohort's cumulative margin to exceed what you paid to acquire it. A 1.2x first-order ROAS is excellent if payback lands at month three and the retention curve holds.

When ROAS is still useful

It is a perfectly good diagnostic. A sudden ROAS movement tells you something changed, whether that is creative fatigue, a competitor's bid, or a broken tracking tag. It is a smoke alarm, and smoke alarms are worth having.

The failure is treating a diagnostic as a target. Anything you set as a target gets optimised, including by algorithms that don't know or care about your margin structure.

The uncomfortable part

Moving to margin-based targets usually makes your reported numbers look worse for a quarter. Blended ROAS drops. Revenue growth slows. Someone in a board meeting asks why.

The answer is that you stopped buying revenue at a loss. It's an easy thing to say and a genuinely hard thing to sit through, which is why most accounts never make the switch.

If you're going to do it, agree the new target and the expected optics with whoever reads the reports before you change anything. The strategy survives the quarter or it doesn't, and that's decided by the conversation, not the arithmetic.