AppLovin MAX and AdMob Mediation Built to Be Audited

Bidding demand does not switch on for a market with a few thousand daily users in it. Under roughly 10,000 daily actives in a single country most bidders will not price your inventory at all, and the stack that fits is three networks, not eleven. The first deliverable in this engagement is usually a shorter network list than the one you arrived with. What the build covers: SDK and adapter integration on iOS and Android with versions pinned in your dependency file rather than floating, so no adapter can change behaviour inside a hotfix. An ad unit taxonomy that survives contact with a data team: one naming scheme across MAX, AdMob and your analytics, with placement identity carried all the way to the revenue event. SKAdNetworkItems written into Info.plist from the networks you actually run, not the union list copied off a blog post. Amazon Publisher Services only where US or UK traffic justifies the extra request. Consent and ATT: A CMP configured against TCF v2.2, Didomi or Google UMP depending on your release process, with the consent string reaching every adapter and a check that vendors you never authorised are not being called anyway. The ATT prompt is placed after the first moment the app has done something for the user. The pre-prompt is written by us in Italian and English. We will not write a pre-prompt that implies the app stops working without permission, and that has ended one engagement at proposal stage. Revenue data, non-negotiable: Impression-level revenue leaves the SDK and lands in your warehouse: the MAX user-level revenue API or AdMob's impression-level ad revenue, forwarded to Firebase, AppsFlyer or Amplitude with placement, country, network and precision flag intact. We do not take an engagement where a client will not turn this on. Without it, every later decision about floors, bidding or ad load is made from a dashboard belonging to a party with an interest in the answer. Device reality: Ad rendering is tested on the hardware your users hold. For the industrial apps we work on that means rugged Android handhelds on Android 11 with 3GB of RAM and the camera held open by a scanner: banner refresh on the main thread becomes an ANR there long before it does on a Pixel. We measure cold-start impact, memory after an hour of continuous use, and frame drops on the screen the ad sits in. How it runs: One week reading the current implementation and the traffic. A build phase with your engineers rather than instead of them, where we write the integration and your team reviews and merges it. Then two weeks of watching after release, with the rollback path written before the release rather than during it. What is not included: Creative production, user acquisition, store optimisation, and network business terms. We will tell you what a revenue share ought to look like; you negotiate it. We do not operate the account after handover unless you take a retainer, and most publishers should not. Who this is not for: Apps aimed at children. Apps with no revenue other than advertising. Teams that want eleven networks live because a deck said fill rate. If your app runs on a company licence and the person seeing the ad is not the person paying, advertising is the wrong instrument and you will hear that on the call rather than in the build. Handover: A written runbook covering how to add a network, how to read the mediation debugger, what each ad unit exists for, and the three graphs to open when revenue moves.