Case study
Subscription training platform
A training app run by a creator, with a subscription model and sales driven largely by organic reach. The account was taken over after a previous agency: sales campaigns switched off, acquisition through comments and direct messages, making it unmeasurable by definition.
99 purchases according to Meta about33 confirmed by payment
The ratio from a client's account. Meta counts a purchase for anyone who clicked an ad within seven days before buying - including someone who would have bought anyway. Hard attribution requires a trace on the payment side. Both numbers are true, they simply answer different questions.
I do not count sales toward my own fee if they would have come without ads.
- 56 campaigns, 274 ad sets and 244 creatives on the account
- one creative was bidding against itself across 26 ad sets at once
- pixel audiences empty, 4 of 5 lookalike audiences broken
- in the last full month before I took the account over, spend rose by 39 per cent while purchases fell by 29 per cent
- server-side measurement distinguishing a first purchase from a renewal
- rebuilding the account from scratch instead of tidying the old one
- a dashboard connecting the ad account to the payment system
- automated recovery of failed payments
- Cost to acquire a new customer down 45% (July vs August, counted from the first real payment).
- New customers from ads x3.3 with an 81% higher budget. Results grew faster than spend.
- Meta reported roughly three times more purchases than could be confirmed by payment.
- August: the best acquisition month in the account's history.
How I counted it
Both numbers come from the payment system, not Ads Manager. I match each new customer to their first payment and attribute that customer to an ad only when the payment can be linked to an ad click within 7 days.
I compare July and August on the same account. In August the budget was 81% higher and the number of new customers from ads grew 3.3 times, so the cost to acquire one customer fell by 45%. At the same time, Meta reported roughly three times more purchases than the payment system confirmed.
This is not a controlled experiment. Creatives and the promotion calendar changed between the months. I claim only that results grew faster than spend and that budget decisions were based on payments, not the dashboard.
How it went
- 01
Diagnosis and shutdown
Reading the state of the account and stopping the spend. Rather than optimising an account where creatives were bidding against each other, I stopped it entirely and started with measurement.
- 02
Measurement and dashboard
Server-side measurement, a first purchase separated from a renewal, and a dashboard connecting ad cost to revenue in the payment system. Only then did campaigns start.
- 03
Rebuilding the account
A new campaign structure built from scratch, with assets filmed by the founder based on my briefs. The first promotional window already ran on the new measurement.
Sound familiar?
We start with an audit: five working days and a specific answer to how much of your reported revenue is visible in your payment system.