SKAdNetwork and Aggregated Event Measurement Setup

Before anything else: if your traffic is Mexico-only, this is probably not worth buying, from us or from anyone. iOS sits at roughly a fifth of the smartphone base here, and ATT opt-in across LatAm runs in the mid-twenties, so the population this engagement recovers is a fraction of a fifth of your users. The work starts paying when iOS media crosses about 15,000 USD a month, or when a real share of your spend runs in the United States, Spain or Canada, where the device mix inverts. We would rather lose the sale than take it on a Mexico-only account. Assuming the volume is there: On the app side: We design the SKAdNetwork 4 conversion value schema against a business event, not against install. The first postback window carries a fine value between 0 and 63; the second and third carry only low, medium or high; and all of them are subject to crowd anonymity tiers that will silence your source-app breakdown when a campaign is small. That constraint drives the design. We spend the fine-value range where your decisions actually differ — for a game, on early payer signal and session depth inside the first 48 hours; for a booking app, on search-to-checkout progression, because the booking itself frequently lands after the second window has already closed. Then we configure it in AppsFlyer, Adjust or Singular, verify postbacks against raw network data, and write down which decisions the schema can support and which it cannot. On the web side: Aggregated Event Measurement allows eight ranked events per verified domain, counts only the highest-priority event a person triggers, and reports with a delay of up to 72 hours. Most accounts we open have that ranking sitting at its default, which means a newsletter signup outranks a purchase and the campaign optimizes toward the cheaper thing. We verify the domain by DNS record, rank events against margin rather than volume, and align the priority list with the event each campaign optimizes toward. A mismatch there quietly starves the campaign that matters most, and it is invisible in every report you currently look at. What we refuse: We do not implement fingerprinting, probabilistic device matching, or any vendor marketing the recovery of ATT-denied users. It breaks Apple's terms, it puts your app at risk at review, and the numbers it produces cannot be falsified, which is the same as saying they cannot be trusted. We will not build a predicted lifetime value mapping for an app with fewer than about 200 paying users a week; below that the fine value is carrying noise and a three-bucket schema is the more honest instrument. We do not sell this as a subscription. The schema is rebuilt when your product or pricing changes, not every month. Who this is not for: Advertisers who want the pre-2021 numbers back. They are gone. What this produces is a smaller, slower, coarser signal that is actually true, together with a written list of the questions your measurement can no longer answer. If the brief is to restore user-level iOS attribution, decline us and be suspicious of whoever accepts. What you receive: The conversion value schema as a table with the reasoning behind each bucket, the MMP configuration made in your own account, a verified AEM priority list, a postback validation report against raw network data, and a note naming every decision this setup cannot support.