Why your ROAS looks good
but you're not making money.
A 6x ROAS and a flat P&L are not a contradiction. They're the default outcome of optimising to a revenue metric in a business that runs on margin.
If your dashboard says the ads are working and your accountant says otherwise, believe the accountant - then find the leak. There are four usual places.
1. Margin mix: the algorithm is selling your worst products hardest
ROAS treats a £100 order at 15% margin and a £100 order at 60% margin as identical. Smart Bidding, fed revenue values, will systematically push budget toward whatever converts easiest - usually high-demand, thin-margin bestsellers, discounted lines and clearance. Revenue climbs, blended ROAS holds, contribution margin quietly falls.
The account isn't broken. It's obediently optimising the wrong number.
The check
Pull last month's paid orders, apply SKU-level margin, and rank campaigns by profit rather than ROAS. On most mixed-margin catalogues, the "best" campaign by ROAS drops out of the top three by profit.
2. Brand cannibalisation: buying traffic you already owned
Performance Max and broad match harvest branded demand - searches that would have converted through the free organic listing anyway. Platform ROAS books that revenue as ad-driven; the P&L knows it was coming regardless. The bigger your brand, the better your ROAS looks and the less it means.
The check
Brand exclusions on, or a proper incrementality read on brand terms. Expect reported ROAS to fall. That fall was never profit.
"The bigger your brand, the better your ROAS looks and the less it means."
3. Variable costs the platform never sees
Shipping subsidies, payment fees, returns, packaging. A category with a 25% return rate and free returns can post a healthy ROAS on every order and lose money on half of them. Fashion and footwear brands: this one is usually yours.
The check
POAS by category - gross profit after variable costs, per pound of ad spend. Anything under 1.0 is a paid acquisition programme for products that cost you money.
4. The incentive problem: who benefits from more spend?
Percentage-of-spend agency fees reward scale, not profit. Nobody managing your account on that model has a financial reason to tell you that the last £20k of monthly spend returned less than it cost. That's not an accusation of bad faith - it's an observation about incentives, and it's why we charge fixed fees.
The check
Ask whoever runs your account to show you the marginal return on the last tranche of spend, not the blended average. Blended averages are where marginal losses go to hide.
What to do about it
Fixing this isn't a bid tweak - it's re-plumbing what the account optimises for: margin-adjusted conversion values in, SKU-level commercial roles set, brand traffic separated, and a profit floor agreed instead of a ROAS target. That's the substance of managing to POAS. The full method is on our POAS agency page, and if the symptom is specifically Performance Max, start with why PMax spends but profit stays flat.
Or shortcut it
Our audit prices the gap between your reported ROAS and your actual paid-media profit, SKU by SKU.
FAQs
Can ROAS and profit really move in opposite directions?
Yes. Whenever spend shifts toward lower-margin, higher-return-rate or already-owned (brand) revenue, ROAS can rise while contribution margin falls. Margin mix does the damage silently.
What ROAS should I target instead?
Wrong question - there is no universal ROAS target, because break-even ROAS is a function of margin, and margin varies by SKU. A 3x target on 60% margin prints money; the same target on 20% margin loses it. Set a profit floor (POAS) and let acceptable ROAS fall out of it per product.
Is MER the answer?
MER (blended revenue over total ad spend) fixes the attribution flattery but is still revenue-based. It tells you the machine is efficient, not that it is profitable. Use it as a sanity check, not a target.
How fast can this be diagnosed?
The margin-mix and brand checks take a competent operator a day with feed data and order-level margins. If an audit takes three weeks to tell you whether your ads make money, it is a sales process, not an audit.