How a single user gesture decides whether popunder ads earn anything or simply burn a deposit nobody refunds

Last updated August 27, 2026 · Smith Jones, Performance Media Buyer

Pricing starts near twenty cents per thousand in cheap regions and reaches several dollars in the largest English markets. The difference between a source that pays and one that quietly drains a deposit comes down to landing page speed rather than to the bid, because a visitor arrives at the hidden window within seconds and abandons it when nothing has rendered. Offers asking for one small action survive on popunder ads, while anything needing comparison or a long form will not recover the cost of a single window. Deposits are not refundable.

How popunder ads open

Browsers stopped allowing windows to open without a user gesture years ago, so every unit sold in this format depends on a real click somewhere on the publisher page. The script listens for that click, hands the new window to the background and leaves the original page as it was. Anyone counting impressions the way display buyers do will misread every report on popunder ads.

Blockers remove another share. Extension level filtering catches part of the scripts on desktop, shifting delivery toward mobile browsers where extension support is thinner, and that is why one source looks different across two device breakdowns of the same campaign. Buyers who compare a desktop line against a mobile line without accounting for filtering conclude that mobile audiences convert better, when what actually happened is that the desktop half was never delivered at all. Check delivery against anti-adblock traffic rates before concluding anything about audience quality. Mobile.

Tab under

Tab under is the sibling format. The script moves the current page into a new tab and loads the advertiser destination in the original one, which several networks sell under the same line item without saying so anywhere in the interface, the rate card or the monthly invoice.

Reporting rarely separates the two, so a buyer comparing sources may be comparing formats and calling the result a geography effect, while one gesture produces exactly one unit in both cases. Networks appearing to deliver several windows per visitor are chaining scripts across one click, and the second and third windows in that chain convert at a fraction of the first, which is where a suspiciously cheap thousand of them usually comes from, invoiced at the first rate.

Capping rules that decide how much attention is left in a session by the time popunder ads arrive

Frequency capping is set by the publisher, not the buyer, which makes it the most important variable a buyer cannot touch, and a visitor who already collected four windows in one session treats the fifth as an obstacle, closes it unseen, while the impression is recorded exactly like the first. Nothing distinguishes them, and the same decay drives push ads on any base. That gap is regularly mistaken for a quality difference between publishers when the publishers may be identical and only the cap changed, and the distance between a source capped once per day and a wholly uncapped source explains most of the price spread in popunder ads.

Uncapped inventory is cheap for that reason. It sells at a fraction of capped pricing because the marginal window arrives carrying almost no attention, and a buyer paying by volume rather than by attention keeps buying it happily for months without noticing. Cheap volume.

Frequency settings the publisher controls and the buyer pays for

Ask for the cap in writing before the first deposit, because a source capped once per twenty four hours per unique visitor behaves like a different product from one capped per session, since a heavy user starts six sessions in an evening and receives six windows under a rule that sounds strict. The seller knows. Ask in the same message as the price. Position inside the session matters as much as the cap, because the first window a visitor gets after arriving carries more attention than one triggered after twenty minutes. Sequence.

Sellers rarely report position, so it has to be inferred, and conversion rate by hour of session separates early windows from late ones well enough to justify a bid difference. The inference is rough and better than assuming equivalence because a rate card says so. Infer it, price it, and let adult traffic tracking confirm the split before a second order.

Page weight as the practical ceiling for popunder ads

The window opens behind everything else, and the visitor reaches it whenever they finish with the page they were reading, a gap measured in seconds rather than minutes, so a landing page that has not rendered by the time attention arrives has already lost the click it was paid for. A heavy page does not convert badly, it never gets seen, and both failures look identical in a report counting conversions against spend, so page weight gets blamed last. Page weight is the practical ceiling on what popunder ads can return in a month.

A working target is a first meaningful render under one second on a mid range connection, meaning one stylesheet and no tag manager loading vendors ahead of the headline. Anything heavier is a straight donation to the publisher rather than a media buy. Nothing else fits.

Redirect chains, and what every additional hop actually costs in milliseconds on a mid range handset over mobile data

Each hop between the network and the final page costs time nobody budgets for, and a tracker redirect, an affiliate redirect and a geographic filter add most of a second on mobile, which here separates a rendered page from an abandoned window. Two hundred milliseconds per hop is a fair estimate, and measuring the chain once per campaign gives a number worse than anyone guessed, because nobody ever counts the hops that were added for reporting purposes rather than for delivery. Compression and caching handle the rest, and serving the landing page near the target region removes latency that no creative work can offset. Milliseconds.

Source tiers behind popunder ads

Source type predicts quality more reliably than country does, since streaming portals and download aggregators generate huge volume from visitors who came for something else, while forums produce smaller volume with visibly higher engagement. A tier one label attached to file hosting traffic is accurate and misleading at once. Both get sold as popunder ads at prices rarely reflecting that difference, exactly as tube inventory and native ads come off one card.

Published tier definitions are worth reading before agreeing to any rate, and I went through the way sources are grouped on popunder-ads.io while building an internal classification, where the split by source type explained more about results than geography did.

Verticals that survive here

Low friction offers survive here, and considered purchases simply do not. Utility software, sweepstakes entries, dating registrations and licensed betting in permitted markets all ask for one small action from somebody who was not looking for anything at all, whereas a product needing comparison, or a form with eight fields, never recovers the cost of one window. The rule concerns the number of decisions sitting between arrival and payment. Count them. That count predicts results better than any list of approved categories a network chooses to publish in its own documentation.

Mobile and desktop are separate markets, since a desktop window waits in a taskbar while a mobile window becomes the visible tab at once. Split by device before geography, the discipline that also keeps any plan to buy website traffic legible, since one campaign covering both is two sharing a budget.

Testing sequences that shorten payback and stop a first campaign producing data nobody can read on popunder ads

Testing has to be sequenced because everything moves at once. Fixing the landing page first, then buying a set number of windows from each source, then adjusting bids only after the first blocking round keeps each variable readable. Changing two things in one day produces a result that explains nothing and costs the same as one that explains something, and most first campaigns skip the opening step, whatever the Buy Adult Web Traffic entry terms allow, then blame the sources. Sequencing costs far less than a week of unstructured spending on popunder ads.

Sample size is where most tests fail. Judging a source on two hundred windows produces a verdict that reverses itself the next day, and reverses again the day after that one. A thousand windows is a floor rather than a target for a first read, and two thousand is safer. Noise.

Steps that keep a first test readable

The sequence runs in four ordered steps, so load one offer across three to five sources at a fixed budget, hold bids identical so the comparison stays clean, block sub identifiers with no conversions once each has enough windows, then raise bids. Skipping the second step is why most first campaigns produce data nobody can interpret. The fourth step is where money gets made, since raising a bid on an uncleaned source buys more of the part that failed. Order matters.

Payback matters far more than daily return. A source breaking even on day one and returning half as much again across the following fortnight beats one that looks better immediately and produces nothing after it, which is why buyers judging popunder ads on day one cut the better source.