Rewarded Placement Buying in Mobile Games, Publisher-Level

A rewarded video that completes at 94 percent tells you a player wanted the in-game currency. It does not tell you your brand registered, and the two are routinely reported as the same number. We buy rewarded inventory at bundle-ID and placement level, we require Open Measurement SDK on every unit we bid on, and we leave 15 percent of the target geography out of the buy so the readout is a difference between exposed and unexposed regions rather than a completion rate. How the buy is assembled: Publisher-direct where the app is worth naming, programmatic where it is not. On the mediation side that means AppLovin MAX, Unity LevelPlay and AdMob bidding; on the demand side Moloco, Liftoff and Smadex, with pre-bid invalid traffic filtering from HUMAN or Pixalate and a sellers.json path check on every seat before a bid is placed. Inventory with no valid app-ads.txt entry for the seat we are buying through does not enter the plan. Placement rules we do not bend: The reward is granted after the ad, never before it. No offerwall, Tapjoy and AdJoe included, because the reward economy there attaches your spend to a player's intent to earn currency and produces a cost per action that looks excellent and predicts nothing. No rewarded interstitials in hyper-casual titles with sessions under two minutes, where the ad lands mid-frustration and the brand carries that. Creative: Vertical, 15 or 20 seconds, claim inside the first three seconds, because rewarded skip rules make the closing ten optional in practice. Playable end cards only where there is a real product surface to simulate: a class booking flow, a store locator, a symptom triage form. We do not commission a generic playable from a network's creative studio and we do not run a television cutdown. If neither exists in your library we scope production separately or tell you to run static. Measurement: SKAdNetwork 4.0 and AdAttributionKit conversion value schemas designed by us rather than taken from the MMP's default template, wired through Adjust or AppsFlyer, with Privacy Sandbox attribution reporting on Android. The lift readout runs on the geo holdout using GeoLift, and the number we report is incremental, not attributed. Where the business is offline, a gym chain or a clinic network, the dependent variable is bookings or footfall in the exposed regions, not installs. Not included: App store optimisation. App development. User acquisition for game publishers, because that auction is contested by teams who do nothing else and we would be the worse bidder. Incentivised installs of any kind. Streamer and influencer integrations. Who this is not for: Advertisers whose only conversion is a phone call to a sales office, and advertisers who need a result inside four weeks. A rewarded buy with a holdout needs roughly six weeks of delivery before the difference between cells is worth reading, and we would rather say that now than write it into a wrap deck later. Fees and media: Package prices are our fee. Media is bought on your seat where you have one, or ours at cost with the platform invoice passed through. Below roughly 25,000 euro of media per market per quarter the holdout cannot carry a readable result, and we will propose a single-cell buy with honest delivery reporting instead of a lift claim we cannot defend.