Platform dashboards said everything was fine while profit shrank. Rebuilding measurement and budgets around contribution margin changed which campaigns deserved to exist.

Client identity withheld. Figures compare the two quarters after restructure to the two before.

The situation

A DTC brand was scaling spend across Google and Meta with dashboards that looked excellent and a bank account that disagreed. Platform-reported ROAS held steady while contribution shrank quarter over quarter. Finance suspected the marketing numbers; marketing suspected finance’s allocations. Both were half right: the accounts were optimizing beautifully toward metrics that no longer meant profit.

What we found

The audit followed the money instead of the dashboards. Added together, the two platforms claimed credit for more revenue than the store had earned, the signature of attribution double-counting. Brand search consumed a large share of Google budget converting people who had already decided. Meta’s spend skewed toward existing-customer retargeting reported as acquisition. Bidding treated all revenue as equal, so budgets chased high-return low-margin bundles. And no one had ever run an incrementality test, so every belief about what worked was a belief about what claimed credit.

The strategy

Rebuild the measurement floor first: one blended scoreboard tracking spend, new customers, contribution per cohort and payback, immune to attribution politics. Then test incrementality where the money concentrated. Then restructure budgets and bidding around what the honest numbers said, and scale only where payback held at the margin.

The work

Phase 1: an honest scoreboard (weeks 1-3)

Conversion tracking was rebuilt with margin-weighted values, new-versus-returning separation and offline reconciliation against the order system. The weekly blended sheet became the meeting: spend, new customers, blended CAC, contribution, payback. Arguments got shorter immediately.

Phase 2: interrogate the beliefs (weeks 3-8)

A brand-search holdout ran in matched regions: a majority of that spend was buying clicks the brand would have received free. A Meta prospecting-versus-retargeting geo split showed retargeting claiming conversions prospecting had created. The two tests reallocated a third of the combined budget with evidence nobody could argue with.

Phase 3: rebuild for contribution (weeks 6-14)

Google restructured around margin-weighted values with brand isolated and capped. Meta consolidated into fewer campaigns with real creative testing volume, pointed at genuine new-customer acquisition. Creative angles multiplied; the winning ones were iterated on a calendar instead of by mood.

The results

Over the following two quarters, blended CAC fell 27% while new-customer revenue rose 52%, on roughly flat total spend. Payback tightened enough that the brand began scaling spend again in the third quarter, this time with tests that proved each increment earned its keep. Platform ROAS, amusingly, barely moved; it had never been measuring the thing that mattered.

What made the difference

The holdout tests. Dashboards argue; geography does not. Once incrementality was measured instead of assumed, every budget decision got easier, and the account stopped paying for applause it would have received anyway.

Explore the services behind this engagement: Google Ads, Meta Ads and Funnel Optimization.