Popunder Buying at Zone Level, Blocklist First

On the popunder buys we have run, under a tenth of the zone IDs that received spend in week one were still receiving it in week three. That is the work. A popunder buy is a subtraction exercise carried out under time pressure, and the money is made in the first fortnight by whoever is willing to sit with the zone-level report every day instead of letting the network's optimiser decide. How we buy: Zone level from the first dollar. Every campaign runs with source-level postback so a converting click ID is attributable to a specific zone, and a zone with no conversions and no dwell gets blocked rather than bid down. Bidding down keeps the zone in the auction, and a zone in the auction keeps spending. We make one prelander decision rather than five: the prelander either survives its own test against direct linking or it goes. Frequency is capped at one impression per user per 24 hours on every buy. Networks will happily let you set three, and three is how a source that was working stops working inside a week. Fraud is a spend problem before it is a quality problem: We check datacentre ASNs, headless signatures, and browser-and-device combinations that do not exist, such as an eight-version-old Chrome on a handset released last year. Session depth and time on page are measured client-side and compared against the network's own click log. A zone whose clicks arrive with no measurable session is not underperforming, it is not real. Where volume justifies the cost we add a third-party filter, but ASN and session checks catch most of it and cost nothing. Japan: Japanese-language popunder inventory is thin, and a large share of what exists sits on manga and streaming piracy properties. We do not buy it for brand advertisers, and we say so on the first call rather than after the invoice. For Japanese-market clients the honest version of this channel is buying international inventory with Japanese-language targeting, accepting a small addressable pool, or not running the channel at all. What we refuse: No pharmaceutical, medical device or supplement offers on popunder inventory in any market. No legal-services acquisition. No campaign that requires a brand-safety guarantee, because adjacency here cannot be pre-approved and any agency promising otherwise is selling a document rather than a control. No buy we cannot measure server-side; if your stack cannot fire a postback, this is not your channel. Who this is not for: Advertisers whose lifetime-value model only pays back past ninety days. The feedback loop here runs in days, and slow payback turns every blocking decision into a guess. Anyone who needs the placement report shown to a brand team. Anyone expecting spend to scale linearly after the cull, because it does not; a buy that has been properly cut usually has a ceiling, and finding that ceiling early is worth more than a month of scaling attempts. What arrives: The zone allowlist with evidence attached to each entry, and the blocklist with the reason each zone was cut. Tracker configuration with the postback path documented end to end. Creative and prelander set with the test that decided between them. A daily spend-and-block log during qualification, then a weekly read. And a written ceiling estimate: what this channel can carry for you before quality falls, which is the number most buyers are never given.