246
Campaigns run
Three years of the Italian Meta account, reconstructed from account structure, naming, bids and timing. €1.48M of spend and 280.8k signups of evidence, read back into the experiments they ran and the budget bets they made.
Early ad sets were hand-segmented on gender, age and lookalike tier, with manual A/B tests on landing pages and creative. Over three years almost all of it collapsed toward one setup: broad, all ages, all genders. The account encodes each test in the ad set name, so the whole experiment log is readable.
83% of spend was controlled ad-set-side (ABO) at €4.75, well below the 17% run on campaign-budget optimisation (CBO) at €11.22. What is live today is the whole strategy in one screen: pour into the proven sitter engine, keep stress-testing the expensive parent side.
The two biggest ad sets fail in opposite directions. On the sitter side, the €225k set is starved: manual, capped, and boxed in by Employment rules that block the usual fixes. On the parent side, a €50k/day Advantage+ set is over-fed: handed to Meta on a loose leash with a cost cap as its only guardrail. Both were opened in the live account.
The opposite failure. This parent set runs Advantage+ with a €50,000/day campaign budget and, unlike the sitter side, it is not restricted: custom audiences and detailed targeting are available. Yet the only thing steering it is a cost cap. No customer lifecycle strategy, no custom audiences included, no existing-customer control. Give Advantage+ that much budget and that little guidance and it will spend to the edge of the cap on whoever is cheapest to convert.
Of €1.48M, only €7.9k ever ran as remarketing (0.5%), and the last real attempt was in 2023. Worse, the cold prospecting is optimised straight to bottom-funnel events: Parent Registration, CompleteRegistration, InviteSent. In pirate-funnel terms (AAARRR) the account only occupies A1 awareness and A2 acquisition, optimises straight to A3 activation, and leaves R1 retention, R2 referral and R3 revenue with no warm layer at all. Cold traffic is asked to run the whole funnel in one hop.
The mechanism: a broad prospecting campaign optimises for conversion probability, so it harvests the warmest, highest-intent users first, exactly the people a remarketing layer should catch cheaply. With no warm layer and no exclusions, prospecting keeps re-converting users who were already going to register. Headline CPAs look fine, but the Advantage+ engine is starved of net-new conversion density, the momentum it needs to get cheaper. Cannibalisation dressed up as efficiency.
Each date-stamped campaign is a fresh attempt. On the sitter side, relaunches stayed cheap as they scaled. On the parent side, 21 generations later the cost per signup is stuck around €9.5, and it has drifted up, not down.