Meta app install, AEO and VO with SKAN value schema design

Meta documents the threshold and then almost nobody builds to it: an ad set needs roughly fifty optimisation events in seven days to leave the learning phase, and value optimisation needs those fifty to be purchases. An app producing three hundred registrations and thirty deposits a week can support one app-event-optimised ad set, not the twelve-cell test matrix that arrives in most media plans. The first deliverable is therefore subtraction, collapsing the account until events concentrate where the model can use them. The second is the conversion value schema, which is the part most shops will not touch because it needs an SDK release and a conversation with your engineers. Which event you optimise on: Registration is usually the answer, with deposit or subscription used as the gate rather than the target. The ratio between them is reviewed monthly, and I will tell you when volume has earned the right to optimise deeper in the funnel. That decision belongs to the event count, not to anybody's preference. Value optimisation goes on only once the purchase value the SDK sends reconciles with your own ledger to within a point or two, because Meta will bid with total confidence on numbers that are wrong. iOS and the conversion value schema: Postbacks arrive across three windows and, under Apple's privacy threshold, as a coarse low, medium or high. I set bucket boundaries from your own first-week revenue distribution, at percentiles, so the three coarse values separate cohorts that genuinely behave differently instead of round numbers that all mean the same thing. The lock window is a deliberate choice: lock early and you get a signal you can still act on, lock late and you get a truer value that arrives after the budget decision. That trade-off is commercial, not technical, and it goes to you in writing rather than being defaulted quietly. Android: Deterministic install data still exists here, so Android carries the honest read and I use it to sanity-check what iOS is claiming. When the two disagree by more than a modest margin, the schema is wrong before the market is. Restricted categories: Real money gaming needs Meta's prior written permission per country plus a licence Meta recognises, and app install placement rules differ market by market. Eligibility gets checked before scoping. Being told in week one that the answer is no in six of your eight markets is worth considerably more than a plan that assumed otherwise. I will not: Build ad sets I already know cannot reach fifty events, so the plan looks thorough. Run Advantage+ app campaigns against a manual structure on a budget that cannot feed both, then declare a winner out of noise. Report Meta-attributed revenue as revenue. Write a creative refresh cadence into the contract, which is a subscription to production rather than to judgement. Or take an engagement where I cannot get twenty minutes a fortnight with whoever owns the SDK. Not a fit: Apps under roughly two hundred qualifying events a month. Paid social at that scale is buying data rather than users, and it should be budgeted and defended as research or not run at all. Teams that cannot ship an SDK change inside a release cycle, since the schema work then has nowhere to land. And anyone who wants this account run to a cost per install: Meta will deliver that number precisely and you will not enjoy the cohort it brings. The schema, the event choice and the reasoning behind both are handed over as documentation your team keeps, whether or not the engagement continues.