Client identifiers removed. Metrics directionally preserved.
This is a real BSD audit with the brand and exact figures anonymised. The pattern, the diagnostic logic, and the directional numbers are reported as they were found.
The account looked spectacular. Meta was reporting a 12.5 ROAS. By any in-platform read, this was a brand to leave alone. The true acquisition MER was 1.68, the first-order economics were below breakeven, and the business was paying more to acquire a customer than that customer returned in first-purchase profit. The reported number was not a sign of health. It was the thing hiding the problem.
Omnichannel women's fashion. Meaningful retail footprint alongside a substantial online business.
Blended across Meta, Google, and two tertiary social channels. Heavily Meta-weighted.
Meta platform-reported ROAS against true blended acquisition MER. Same spend, two very different stories.
First-order gross profit divided by new customer acquisition cost. Under one means losing money on the first purchase.
The pattern.
A spectacular platform ROAS sitting on top of acquisition economics that did not work. Meta reported 12.5. Strip it back to 7-day click only and it fell to 8.5. Apply incremental attribution and it landed at 6.44. Look at new audiences only on an incremental basis and it was 4.75. Reconcile the whole account against the P&L and the real acquisition MER was 1.68. Every layer you removed took a chunk of the number with it, which is the tell. A healthy account does not lose most of its reported return the moment you stop crediting it for things it did not cause.
The same shape repeated on Google. The top spending campaign reported around a 24x ROAS. When we pulled its search terms, it was overwhelmingly brand. The budget was defending demand that already existed, not buying anyone new, and the platform was crediting itself for purchases it would have got for free.
Why it was happening.
Three things were inflating the read at once. The first was attribution. Every campaign ran 7-day click plus 1-day view, which lets Meta claim a sale where a single frame of an ad appeared on screen and the person bought within 24 hours, with no interaction at all. In an omnichannel business this is particularly corrosive. Someone visits a store, decides to buy, later scrolls past an ad they never touch, and Meta books the credit.
The second was existing customers. Roughly $52k a month, about half the Meta budget, was being spent on people who already buy, at a frequency of around 30 ad impressions per person per month. That spend was claiming credit for repeat purchases that would have happened anyway, and dragging the blended number up with it.
The third was the metric itself. The account was being judged on blended MER, which folds returning customer revenue into the numerator. This brand was generating around half a million a month in returning customer revenue, and lift data says paid media causes a fraction of that. Most of it is email, organic, and direct. Counting it as paid performance is how a barely-breakeven account reads like a winner.
If your acquisition layer is underperforming and your platform ROAS is spectacular, the ROAS is not telling you the account is healthy. It is telling you what it is allowed to take credit for.
What it was costing.
The honest number was LTGP:NCAC below 1.0, and deteriorating. New customer acquisition cost was running ahead of the gross profit on a first purchase. Even on generous margin assumptions the position stayed under breakeven. For context, we hold fashion brands to a 2.0 to 3.0 range on this metric. This account was buying new customers at a loss and the reported ROAS gave nobody a reason to look closer.
It was worse than it appeared, because retail and online customers were never de-duplicated. Existing in-store customers buying online for the first time were being counted as new. That inflated the cohort and made the acquisition cost look better than reality. The true new customer economics were below even the below-1.0 figure on the page.
When platform ROAS sits far above your P&L, the platform is taking credit, not creating value.
A 12.5 reported against a 1.68 real acquisition MER is not a measurement quirk. It is the distance between what paid media is credited for and what it actually caused. Close that gap before you scale anything, or you scale the credit, not the customers.
The fix.
- Move to 7-day click only. Drop 1-day view across every campaign and use incremental attribution as a secondary read for budget and creative decisions. This alone removes the largest single source of inflation.
- Cut existing customer spend hard. Take it from around $52k to roughly $17k a month, pulling frequency back to the 5 to 7 range, and redirect the freed budget straight into cold acquisition. This was the single biggest immediate win in the account.
- Judge the account on acquisition MER, not blended MER. New customer revenue over total spend, reconciled against first-order contribution margin. Keep blended MER in the finance layer where it belongs, not running the media decisions.
Alongside that, de-duplicate retail and online customers before calculating acquisition cost, and isolate cold Google campaigns from brand terms so the cold read stops borrowing brand-search credit. None of this is exotic. It is the difference between optimising to a number that flatters you and optimising to one you can hand a CFO.
What changed.
The diagnosis reframed the entire account. The brand stopped reading a 12.5 ROAS as permission to keep spending the way it was, and started measuring the only thing that decides whether paid media is working: whether it is buying new customers at a profit. With the existing customer spend cut, attribution tightened, and the budget redirected to cold, the blended acquisition MER had room to move toward the 2.0 to 3.0 we hold fashion to, and every future scaling decision was being made against a number that reconciles with the P&L instead of one that argues with it. The reported ROAS never told them anything was wrong. The real metric did.