Zone-Level Whitelisting for Push and Pop Campaigns

On a push or popunder account of any size, most of the conversions that survive a 30-day refund window come from a small minority of zone IDs, and roughly half the spend lands on zones that never produce one. This engagement finds that minority and writes it down: the zone IDs, the network each sits on, the bid it holds at, and the date the entry was last re-tested. The list lives in your tracker, under your licence. You keep it whether or not you keep us. How a zone earns its place: Every zone starts on a metered test with a fixed click budget, and gets no verdict until it clears the sample the offer's conversion rate requires, usually between 300 and 800 clicks, higher when the payout is small and variance is wide. We judge on the lower bound of the interval rather than the point estimate, because a zone showing three conversions on 200 clicks is noise, and buying more of it is exactly how an account drifts. Each verdict is stored with the click count it was made on, so six weeks later you can see which decisions were thin. Push and pop do not behave alike: On push we cohort by subscriber age. A base opted in nine days ago and a base opted in eleven months ago are different inventory sold under one zone ID; the fresh cohort converts and exhausts, the aged cohort is cheap and mostly fatigued. Sources that will not expose subscriber age get bid as though the whole zone is aged. In-page push is kept on its own list entirely, because it reaches Safari and iOS where service-worker push does not, and its click rate is inflated by placement rather than intent. On pop we cap one impression per user per 24 hours, hold dayparting to the buyer's local time, and separate straight popunders from tab-under and back-button implementations, which are different products even when a network prices them identically. The list rots, and the budget says so: Traffic gets resold. A zone that was a Mexican file-sharing audience in March can be an aged Indian push base in June under the same ID. Entries older than 21 days without spend get re-tested or dropped, and roughly 8 percent of monthly budget stays reserved for exploration: new zones, re-tests of blacklisted ones, and the occasional re-buy of a zone that was killed by weak creative rather than weak traffic. A frozen whitelist is a decaying asset and we will not hand you one and call the work finished. Invalid traffic comes out first: Datacentre ASNs, headless user agents, impossible click-to-conversion intervals and repeated device fingerprints are filtered before any of it reaches the whitelist logic. Skip that order and you build a list of the zones best at faking. We run one third-party fraud signal alongside our own filters, and we name it, with its cost, in the proposal rather than folding it into a fee. Not included: Creative production, landing pages, offer selection. We will say early when no zone list can rescue a lander, and that is an uncomfortable conversation we would rather have in week one. We do not buy through resellers who refuse zone-level reporting, we do not accept network-side optimisation that hides the placement, and we do not price this as a percentage of media spend, because that arrangement pays us to keep bad zones alive. Who this is not for: Buyers spending under about eight thousand dollars a month across these formats. The list will not see enough traffic to stay current and you will be paying for maintenance of a stale file. If what you actually want is the whitelist to hand to a cheaper buyer, that is fine, it is yours, but say so and take the audit rather than the retainer.