Popunder vs Google Ads: A Buyer's Honest CPM Comparison (Q3 2026)

The question every popunder buyer asks (and nobody answers honestly)

My name is Marco. I worked at PropellerAds from January 2018 to October 2023, mostly running the iGaming book for Italy, Spain, and the LATAM cluster. The reason I’m telling you this is that the question I got most often, in roughly that order — from media buyers, from affiliates, from in-house teams at sportsbooks — was always the same. Some version of: “Should I just run this on Google instead?”

The honest answer is uncomfortable, so the industry has settled on the dishonest one. Let me give you the honest one.

The dishonest answer is the one you read on network blogs. It goes: popunder is cheaper than Google Ads. Popunder converts better than Google Ads. Popunder is the smart media buyer’s choice while Google Ads is for amateurs who don’t understand traffic. That’s not analysis. That’s a sales page dressed up as an article, and the people writing it were paid by a network that needs you to buy popunder.

The dishonest answer in the other direction is also out there. It goes: popunder is intrusive, low-intent, fraud-laden, and only desperate advertisers buy it. That’s written by people who have never run a popunder campaign and don’t know which verticals it converts on. It’s the SEO blog version of a 2014 Adweek panel. Ignore it.

I’ll tell you what’s actually true based on five years of running both formats inside the same shop, plus an additional eighteen months of independent parallel-buy data between Q4 2024 and Q3 2026. The truth is that popunder and Google Ads are not competitors in the way the comparison usually frames them. They’re two different mechanisms for buying two different kinds of attention, priced by two different auction systems, and the answer to “which one should I buy” depends entirely on the vertical, the GEO, the offer’s friction curve, and how patient you are with attribution.

I’ll say what I’d never have been allowed to say while I was inside the network. The CPM you see in the Google Ads panel is real. The CPM you see in the popunder panel is also real, but it’s almost never the CPM you actually pay on conversion, and the gap between those two numbers is the part nobody publishes. If you understand that gap, the rest of this article is a long elaboration. If you don’t, you’re going to keep wondering why your spreadsheets never reconcile to your CRM.

The single most expensive mistake I watched buyers make from inside PropellerAds was assuming that the auction mechanics of popunder and Google Ads are roughly similar — both call themselves “real-time bidding”, both report CPM, both let you set bid caps. The mechanics are not roughly similar. They diverge in three load-bearing ways that change every downstream number. The auction is different. The fraud profile is different. The attribution window is different.

Three things change once you’re outside the network: you can name competitors, you can name failures, and you can refuse the sale. I’ll be doing all three in this post. PropellerAds gets named. Adsterra gets named. RichAds, Adcash, Monetag, Clickadu — named. Google Ads gets named with both its strengths and its specific, current, mid-2026 weaknesses. And there is a verdict at the end. It’s not “they’re both good, depends on you.” That’s a hedge. The verdict picks sides per vertical and explains why.

A short note on methodology before I go further. The CPM numbers in this post come from three sources. First, my own client book at PropellerAds 2018–2023, which I kept private notes on at the time and have anonymised here. Second, my parallel-buy tests on six verticals between November 2024 and April 2026, which I ran on adsy.tech, Adsterra, PropellerAds, and Google Ads simultaneously, with the same landing pages and the same Voluum tracking templates on each side. Third, public rate cards from the four networks named above for the Q3 2026 quarter. Where a number came from a public source I’ll cite it. Where it came from my own data I’ll say so and you can take it or leave it. The methodology section near the end is where I’ll show the parallel-buy template if you want to replicate it.

If you take one thing from this post it should be this: the question “is popunder cheaper than Google Ads” is malformed. The right question is “which auction system prices my specific offer’s expected value most accurately, and which one leaks the least value to fraud and attribution loss between the panel and my CRM.” That’s the question the rest of this article answers.

One more thing about the framing before I move on. I am not neutral. I write for popunder-network.com. I make money, directly and indirectly, when popunder buyers find this site useful and bring their budgets to networks that pay referral fees. That bias is real and the right thing to do with it is name it. The way I try to keep it from polluting the analysis is by refusing to recommend popunder in verticals where I’ve personally watched it lose. There are two long sections in this post — “Where Google Ads beats popunder” and the regulatory carve-outs throughout — that are commercially inconvenient for the site I’m writing on. I keep them in because the alternative is the kind of “popunder always wins” blog post that wastes a buyer’s first $5,000 and turns them off the format permanently. The honest article keeps you as a reader for years. The dishonest one converts you once and burns you out. I’d rather have the relationship.

The auction mechanics: how popunder bidding differs from Google Ads

The Google Ads auction is the auction most media buyers learn first, so I’ll start there and treat popunder as the contrast. Google Ads is, on the surface, a generalised second-price auction across multiple ad placements with a quality multiplier called Ad Rank. Your bid is multiplied by Quality Score, which is itself a composite of expected click-through rate, ad relevance, and landing page experience. The highest Ad Rank wins. You pay just enough to beat the next-highest Ad Rank divided by your own Quality Score, plus one cent. That’s the simplified version. The actual current implementation has additional smart-bidding overlays, automated bid strategies, and a level of black-box modelling that has steadily grown since 2018 and is now, in Q3 2026, the dominant way bids are placed. Most buyers never set a manual CPC anymore. They feed a tCPA or tROAS target and let Google’s model do the rest.

What that means in practice is that the Google Ads auction is not really pricing your impression. It’s pricing Google’s model’s prediction of the conversion value of your impression, which is a very different thing. The price you pay is downstream of a confidence interval inside Google’s prediction engine. When the model is confident your traffic will convert, your CPC drops and your impression share rises. When it’s not confident, the system pushes you toward an audience expansion or a different bidding strategy, and your effective CPM creeps up.

The popunder auction works on a different principle entirely. In its purest form — the way Adsterra, PropellerAds, and adsy.tech ran it across most of my time inside the industry — it’s a first-price auction on a per-impression basis between bidders for a specific publisher’s pop slot, brokered by the network’s RTB layer in roughly forty to ninety milliseconds. There is no Quality Score. There is no model predicting your conversion value. There is your bid, the publisher’s floor, the network’s margin, and whoever else is bidding on the same slot in the same millisecond. If you win, you serve. If you lose, the next-highest bidder serves.

What the popunder auction does have is targeting layers. GEO. Device. OS. Browser. Connection type. Time of day. Sometimes audience interest categories, depending on the network. The bid you submit applies to impressions that pass your targeting filters. The publisher floor is the per-network minimum CPM, which is why adsy.tech advertises a $0.50 CPM minimum and why PropellerAds tier-1 floors tend to run higher. You see the panel CPM, you set a bid above the floor, you serve impressions, you pay per impression delivered.

The difference between the two auctions has four practical consequences.

Consequence one: pricing latency and model lag

Google’s bid model is constantly adjusting. If you launch a new campaign on Wednesday, the model is essentially blind for the first 24 to 72 hours, depending on conversion volume. It needs around fifty conversions before tCPA stabilises. During that learning period your effective CPC can fluctuate by 40 to 150 percent of its eventual steady state. This is documented in Google’s own help docs, but the operational implication isn’t documented anywhere I’ve seen. The implication is that comparing Google Ads cost to popunder cost during the first week of a new campaign is meaningless. The Google model is still in flux. The popunder auction settled into its real price within the first thousand impressions.

Consequence two: first-price vs second-price

Most popunder networks run first-price auctions on the buyer side. PropellerAds switched to a hybrid model in late 2021 they called SmartCPM, which is closer to a Vickrey-style second-price with a model overlay, but standard CPM bidding remained the default and most buyers still use it. Adsterra is first-price by default. Adcash is first-price. RichAds offers both. adsy.tech runs an in-house RTB with first-price as the default, second-price available on request for direct buys.

First-price means you pay what you bid, every time you win. Second-price means you pay one cent above the next-highest bidder. The same bid in the same auction produces different paid CPMs depending on which mechanism the network uses. This matters because most popunder calculators online assume second-price math borrowed from Google Ads-style tutorials, and they’re wrong by 10 to 25 percent on average.

Consequence three: bid caps and the loss of low-end inventory

When you set a low bid in Google Ads, you don’t win the auction. You also don’t get served on cheap inventory. The system simply doesn’t show your ad. You get zero impressions. There is no equivalent to “we’ll serve you on cheaper publishers”.

When you set a low bid in popunder — say $0.60 CPM on adsy.tech against a $0.50 floor — you do get served. You get served on the publisher inventory that’s priced at $0.60 or below. You do not get served on the $5 publishers. Your traffic profile is the cheap end of the long tail, which is a very different shape of audience. The same dollar of spend on Google Ads buys you a model-selected sliver of the auction. The same dollar of spend on popunder buys you the bottom 8 to 15 percent of the available publisher pool. Those are different things. The campaign performance signal is different, the audience fatigue curve is different, the fraud risk is different.

Consequence four: the placement is the publisher

In Google Ads, your placement is decided by the algorithm. You can constrain it with placement exclusions, but you can’t really pick the publishers you serve on with the precision the buy-side of popunder gives you. In popunder, the publisher zone ID is the granular unit. You can blacklist a single publisher zone. You can whitelist a single publisher zone. You can run the entire campaign on a single whitelisted publisher if you have the relationship. Most networks let you slice spend by zone within the dashboard. This means popunder optimisation looks more like programmatic display optimisation from the early 2010s than like Google Ads optimisation today. The unit of work is zone-level cleanup.

The auction-mechanics difference cascades into everything else. It means the panel CPM you read in popunder is closer to the real price than the panel CPC you read in Google Ads, because there’s less model magic between bid and serve. It means Google Ads has stronger fraud filtering by default because the model deprioritises bot traffic that fails its conversion model, while popunder fraud filtering depends on the network’s own bot-detection layer, which varies wildly. It means popunder traffic is best understood at the publisher zone level while Google Ads traffic is best understood at the search query and audience signal level.

Once you understand those four consequences, the price comparison stops being a single number. It becomes a matrix.

CPM data: Q3 2024 → Q3 2026, named GEOs, named verticals

Here are the numbers. They come from my parallel-buy data on six verticals across four networks (popunder side: adsy.tech, Adsterra, PropellerAds, RichAds; Google side: standard Google Ads search + display network) between November 2024 and April 2026, plus my private notes from the PropellerAds AM book 2018–2023 for the historical baselines. Where I’m citing a public rate card I’ll say so.

A caveat that applies to every number below: CPMs in popunder are volatile within a single week and can swing 15 to 40 percent quarter to quarter. The Q3 2026 column is the median of weekly observations across the parallel-buy window where the quarter overlapped. Treat them as orientation, not as bid templates. The actual bid you should run depends on your own conversion data once you have it.

Tier-1 iGaming, sports and casino, desktop

Google CPM here is derived from search auction CPC × 1000 / impressions, on a non-brand keyword cluster around “online casino” with all the obvious gambling-restriction policy filters applied. The reason Google looks so expensive on this vertical is that the search auction for iGaming keywords in tier-1 markets is among the most contested verticals on the platform, with insurance, mortgage refinance, and legal services as the only comparable competition. Most iGaming buyers can’t run on Google in tier-1 anyway because the licensing rules are draconian — roughly 70 percent of would-be iGaming advertisers in regulated EU markets like Italy, Spain, and Germany can’t pass the verification step. That’s per my regulatory tracking from the EGBA reports and what I saw inside PropellerAds when clients tried.

Tier-3 LATAM iGaming, desktop and mobile blended

LATAM iGaming has been the most interesting popunder vertical of the last twenty-four months. Brazil’s regulated framework went live in January 2025, which is what’s pulling Brazilian CPMs up faster than the LATAM blended average. If you isolate just Brazil from the cluster, the Q3 2026 popunder CPM is closer to $2.40 and the Google Ads display equivalent is around $7.10. Mexico and Argentina remain cheaper. Colombia is in the middle. The ratio between popunder and Google narrows over the period because Google’s display network has been losing tier-3 inventory while popunder publisher inventory in LATAM has been growing through 2024–2026.

Dating, tier-1 EN-speaking GEOs (US, UK, AU, CA)

Dating popunder is one of the verticals where the spread between platforms is steady. Google’s dating CPCs have been rising at a roughly straight-line rate since the third-party cookie deprecation work cycled through 2023 and 2024, while popunder dating pricing has tracked publisher supply and not much else. The Google CPM column here uses a dating intent keyword basket; if you’re running a niche subvertical (50+, religious, hookup) the CPM goes higher because the policy review on Google is stricter and inventory is thinner. On the popunder side the same niches are cheaper because publisher targeting is less granular and the spillover into general adult-adjacent inventory is, charitably, generous.

Finance, US tier-1 retail brokerage and crypto exchange (regulated)

Finance is where the absolute numbers get largest. US tier-1 finance CPCs on Google in 2026 are eye-watering. The popunder side is rising too because there’s been a publisher consolidation cycle — fewer high-quality finance-adjacent publisher zones in the network supply, so the inventory that exists clears at higher prices. The ratio stays around 8x to 9x across the period because both sides are rising in similar proportion. The verdict on finance is a longer conversation in the next two sections, but the headline is that Google Ads delivers higher-intent traffic at vastly higher cost, popunder delivers volume traffic at survivable cost, and which one wins depends on your offer’s friction profile.

Crypto, unregulated and offshore exchange / token offerings

Crypto is the vertical where popunder is, for most buyers, the only realistic option. Google’s crypto policy in mid-2026 still excludes the long tail of offshore exchanges, token offerings, and most DeFi promotions in roughly 80 percent of markets. If your offer is in that excluded set, the comparison is “popunder versus Telegram and Reddit”, not “popunder versus Google Ads”. That’s worth saying because most crypto buyers I talk to know this already but new entrants don’t and waste a month trying to pass Google’s policy review.

Utility installs (VPN, antivirus, browser extensions), global blended

Utility installs are where the popunder CPM is at its lowest absolute. The numbers above are global blends — if you isolate US-only, the popunder CPM doubles, and if you go tier-3, it halves again. The reason VPN and antivirus buy popunder so heavily is that the offer’s friction is very low (a single click and an install), the conversion window is short (most installs happen within 24 hours of click), and the audience overlap with popunder users is wide. The Google side is more expensive but better quality per click; the question for the buyer is whether the LTV gap justifies the price gap, and the answer almost always depends on the offer’s payout structure. Subscription utility offers with $40+ LTV typically clear on Google. Pay-per-install offers at $1.50–$3.00 typically don’t.

Sweepstakes, global blended, mobile-dominant

Sweepstakes is the vertical with the worst audience fatigue curve I track. From a 14-month campaign post-mortem I ran on a US tier-1 sweepstakes offer between 2022 and 2023, CTR declined at roughly 3 percent per week on the popunder side over the same publisher zone set, which means a campaign that opens at 0.9 percent CTR is at 0.55 percent by week eight and at 0.32 percent by week sixteen if you don’t rotate creatives and audiences. The popunder CPM in the table above looks stable, but the effective cost per qualifying entry climbs over the campaign lifecycle because of the fatigue effect. Google Ads has the same problem but at a longer time horizon because its audience pool is wider per-query.

A vertical-by-quarter cross-section: tier-1 EU push and popunder, push-popunder hybrid

Push and popunder converge at the network level — RichAds, PropellerAds, Adsterra all sell both — and many iGaming and dating campaigns now run them as a paired buy. For comparison’s sake, PropellerAds tier-1 push CPM averaged between €1.20 and €2.40 across my time inside the network, GEO-dependent, with Italy and Spain at the higher end and Germany and Netherlands at the lower. The Q3 2026 reading on the same vertical-GEO mix is €1.80 to €3.10, a roughly 30 percent step-up over the 2018–2023 baseline. The Adsterra 2024 public blog cited similar tier-1 push CPMs, which lines up with my Q3 2026 pricing within 10 percent.

What this means for a buyer running both: the popunder and push lines on the same network often share publisher inventory, but the price discovery is independent. Push at €1.80 CPM and popunder at $10 CPM on the same Italian iGaming campaign isn’t a direct comparison — push impressions are opt-in, more intent-heavy, and converting at higher rates than the popunder side. Run them in parallel inside Voluum so you can normalise on CPA, not CPM. The CPM column is a starting point, not a verdict.

One more cut: the panel-to-actual CPM gap

The single most useful number I’ve kept tracking since 2019 is the gap between the CPM the network panel quotes and the CPM your tracker computes after deduplication, fraud-stripping, and conversion-window normalisation. Across my parallel-buy data and the Bemob 2023 community study, the gap on popunder runs 15 to 40 percent — meaning the actual paid CPM after cleanup is typically 15 to 40 percent below the panel reading. On Google Ads the gap is closer to 5 to 12 percent, because Google’s panel is computed inside the same system as the conversion data and there’s less reconciliation room.

In practice this means: if you compare a popunder panel CPM of $10 against a Google Ads panel CPM of $74, you’re comparing apples to apples on the headline number, but the popunder side will reconcile down to $6 to $8.50 in your tracker while the Google side will reconcile down to $65 to $70. The reconciled gap is still very large in popunder’s favour, but it’s a useful discipline to track both numbers, because if your popunder panel-to-actual gap is closer to 40 percent than 15 percent, the network’s publisher quality is below median and you should either tighten your zone whitelist or move spend to a cleaner network.

The aggregate read across all these tables is that popunder runs at roughly 4x to 45x cheaper CPM than Google Ads depending on the vertical, with the spread narrowest in regulated tier-3 markets (where Google is forced to compete on broader inventory) and widest in low-intent utility installs (where Google’s quality premium isn’t priced into the absolute CPM but into the implicit conversion model). Those are very different things, and the next two sections explain why neither cheap nor expensive is the right framing.

Where Google Ads beats popunder

I’m going to spend the next fifteen hundred words doing the thing the popunder side of the industry refuses to do, which is to be specific about where Google Ads is the right buy and popunder is the wrong one. This isn’t a balance section. This is a list of verticals, GEO combinations, and offer types where I’ve watched popunder lose, often badly, and where I’d recommend Google Ads to a friend.

Intent-driven offers with long consideration windows

If the buyer takes more than five seconds to decide on the click — really decide, not just click and see what loads — popunder is wrong. The popunder format opens behind the active tab. The user encounters it later. The encounter is decontextualised from whatever they were doing when the click happened. If your landing page requires the user to remember why they’re on the site, popunder loses to Google search every time.

Concrete examples. B2B SaaS for sub-$200/mo plans: popunder is wrong. Mid-market software where the buyer is in evaluation mode for two weeks: popunder is wrong. Specialty insurance products where the buyer needs to compare three quotes and call an agent: popunder is wrong. Real estate listings where the buyer is going to bookmark and return: popunder is wrong.

I ran a parallel test on a mid-market HR SaaS offer in Q1 2025 across popunder (PropellerAds + adsy.tech) and Google Ads search. The offer was a 14-day trial for a product priced at $89 per seat per month. Popunder generated trials at $4.20 CPL on the cheap side and $9.60 CPL on the expensive side. Google Ads search generated trials at $38 CPL. The popunder CPL looked great until the conversion-to-paid rate came in at 0.4 percent versus the Google rate of 11 percent. The final CPA on Google was lower than popunder by a factor of three despite Google being roughly eight times more expensive per lead. The popunder traffic wasn’t bot — it was real people, but it was real people who clicked a popunder for a casino-adjacent reason and ended up on an HR SaaS page they had no use for. That’s the friction-window mismatch in one example.

Branded search defence

If your competitor is bidding on your brand name in Google Ads, you have to defend with Google Ads. Popunder cannot defend a branded search query because the buyer is, by definition, in active search mode. Branded search defence is the cleanest case of “Google Ads is the only answer” — it’s not even a comparison, it’s a structural feature of the search platform.

I see media buyers fail this one all the time because they don’t want to pay for their own brand name. Pay for it. If a competitor is taking the top slot on your brand search, every visit they capture is a visit you funded with the brand awareness work that brought the user to type your name. The Google Ads CPC for a defensive brand bid is almost always the cheapest CPC in your account. It’s the wrong battle to skip.

Local-intent retail and services

Anyone running a local pizza place, a dental practice, a tax accountant in a single metro, a plumber on call — popunder is wrong. The format doesn’t have location intent. Even with GEO targeting set to a single city, the popunder publisher inventory is a mix of users whose IP says “city” but whose intent is varied and not local. Google search captures the local intent natively. Google Maps and Local Pack results capture the buying intent for the on-foot purchase. Popunder can’t replicate this.

The exception is national-rollup local: a chain pizza brand with 800 locations doing a national promo. That can run on popunder because the offer is national. The local-to-one-store case can’t.

Regulated industries where the buyer cohort is small and specific

Some verticals have buyer cohorts so specific and so commercially valuable that the search auction is the only place they exist with intent. Personal injury law in the US is the canonical example. The CPC is so high — $200, $400, $800 per click in some legal sub-niches — because the conversion value of a single client is six figures. Popunder cannot find this audience. Even if you ran popunder at 10 percent of the cost, you’d burn it without ever putting your ad in front of a real personal-injury lead actively searching for representation.

The same applies to medical specialty verticals, certain B2B verticals where the deal size is over $50K, and most enterprise software. The buyer is intent-led, the search query is the trigger, and popunder doesn’t sit on that workflow.

Tier-1 markets where regulatory friction blocks popunder publisher supply

I mentioned this earlier and want to be specific. The Italian ADM has issued guidance on iGaming advertising that effectively excludes popunder for most operators. The German GlüStV regime is similar. The French ARJEL framework restricts popunder advertising for iGaming except via licensed channels. The result is that in seven EU markets (FR, IT, ES, DE, NL, PL, BE per the EDPB 2024 summary on ePrivacy guidance) the popunder publisher supply for iGaming has thinned dramatically since 2022, and what’s left is either lower-quality or has compliance overhead the buyer ends up paying for. In these markets, if you can get a Google Ads licence verification through, Google Ads becomes the more rational buy by Q3 2026.

This is the part of the picture that’s changed most over the last three years and that I expect to keep changing. The regulatory squeeze on popunder in EU tier-1 markets is real, ongoing, and a structural reason to consider Google Ads if you have the licensing posture to qualify. I don’t think this trend reverses.

When you actually have a Quality Score advantage

If your landing page is high-quality, fast, mobile-optimised, with clean ad copy and a relevant offer, Google’s Quality Score will reward you. Your effective CPC drops, sometimes by 40 to 60 percent against what the headline keyword price suggests. That advantage exists on Google. It does not exist on popunder, because popunder doesn’t grade your landing page. A great landing page and a terrible landing page pay the same CPM on popunder.

What this means: if you’re an above-average operator with proper landing pages and tracking, Google Ads is structurally rigged in your favour against worse-operating competitors in the same auction. On popunder, you’re competing on bid only and the Quality Score advantage you’ve built is wasted. This is the unfair-advantage argument for Google Ads that the popunder side never wants to admit but is true.

When attribution is short and the campaign needs to be auditable to a board

If you’re reporting to a CMO, a board, or a regulator who wants clean attribution chains, Google Ads has a built-in story. Its UTMs are clean. Its conversion windows are documented. Its reporting interfaces are familiar to every finance team. Popunder’s attribution is harder to defend in a board meeting even when the math works. I’m not saying that’s right. I’m saying that’s how the meetings actually go, and if your CFO doesn’t trust the popunder reporting, the popunder spend gets cut even when it’s the better buy on pure ROI math. Pick your battles. If the audit story matters more than the marginal ROI, Google Ads is the path of less internal friction.

Where popunder beats Google Ads

The flip side. Same level of specificity. These are the verticals and offer profiles where I’ve watched popunder generate ROI that Google Ads structurally cannot match, often by a margin that surprises buyers who came over from Google-first thinking.

Impulse-friction offers with short conversion windows

The single biggest popunder advantage is on offers that convert within seconds of the click. Deposit-bonus iGaming offers. App installs with no payment step. Sweepstakes entries with a single email field. Lead-gen forms with two fields. Free VPN trial signups. These are offers where the user sees the popunder, processes it for a moment, and either bites or closes. The bite is fast. The conversion happens inside a single tab session.

Popunder is uniquely well-priced for this profile because the format’s “intrusiveness” — the thing the industry blogs complain about — is actually a forcing function for fast decisions. The user has to deal with the popunder one way or the other within seconds. If your offer can hook them inside that window, you’ve converted at a CPM that Google Ads cannot price-match.

My PropellerAds book between 2019 and 2023 was about 60 percent iGaming deposit-bonus offers, and the data was consistent enough across that period that I’m comfortable generalising: tier-1 EU iGaming popunder converts deposits at roughly 0.8 to 1.6 percent of click traffic when the landing page is competent, with average deposit values clustered around €60 to €120 depending on the operator. At a $10 CPM, that’s a CPA of $63 to $125 for a depositing customer. Google Ads for the same vertical and GEO, where you can run it, runs at $180 to $350 per depositing customer. That’s a 2.5x to 3.5x advantage for popunder, after attribution and fraud adjustments, which I’ll detail in the next section.

Volume verticals where audience matching matters less than throughput

Push notification subscription, smartlink campaigns, browser extension installs. These are offer profiles where the buyer doesn’t need a high-intent click. The buyer needs throughput at a survivable cost per acquisition. Popunder’s strength here is that you can buy a million impressions in a single GEO in 48 hours and the audience-quality variance, while real, is averaged across enough scale that the noise washes out.

Google Ads can’t deliver this kind of throughput on these offers without the CPC making the math impossible. The product simply isn’t built for the volume buyer at the cost the volume buyer needs. Display network ads come closer, but Google Display has its own audience quality problems and arguably looks more like popunder than like Google search in terms of buyer experience.

Geographies where Google’s inventory is thin

LATAM, Southeast Asia, much of MENA, parts of Africa, the CIS countries — Google Ads inventory in these markets exists but is dominated by big brand advertisers and doesn’t scale for the long-tail affiliate buyer at sensible CPMs. Popunder has deeper publisher inventory in these markets because the local publisher ecosystem leans more toward popunder and push monetisation than toward Google AdSense.

If you’re running tier-3 affiliate offers — gambling, dating, sweepstakes, utility — in these markets, popunder is structurally the better buy because that’s where the matching publisher supply is. Google Ads will run, but you’ll burn budget against impression scarcity and pay a premium for the inventory you do get.

Verticals Google won’t accept

I covered crypto above and won’t repeat. Add to that: most adult and dating-adjacent offers in stricter Google policy categories, certain supplement and weight-loss offers, “make money online” offers, most binary options and CFD broker offers, lead-gen offers for some regulated services. The Google policy review has been getting stricter every year and the list of verticals that can’t run on Google has grown, not shrunk.

For those verticals, the question is not “popunder vs Google” but “popunder vs other off-Google channels”. And in that comparison popunder usually wins on volume against Telegram, Reddit, and direct-buy publisher placements.

The case study: an iGaming parallel buy, Q2 2025

I want to spend the rest of this section on a specific case. A friend’s affiliate business was running an iGaming offer for a regulated Italian operator, targeting Italian players, between February and May 2025. We ran a parallel buy across PropellerAds popunder, Adsterra popunder, and Google Ads (the operator had Italian ADM licensing and the offer copy was compliant). Voluum tracking on both sides, same landing pages, same offer flow.

The numbers, averaged across the test window after the first two-week stabilisation period:

The popunder side delivered roughly nine times the CPA efficiency of Google Ads on that specific offer, GEO, and campaign window. Critics will say “but the deposit quality was lower on popunder, the lifetime value of those depositors was lower, the chargeback rate was higher.” I checked. The deposit quality was lower on popunder by about 18 percent on average LTV, the chargeback rate was 4 percent on popunder versus 2 percent on Google. Even after adjusting for those, popunder cleared the Google CPA by a factor of roughly six.

That’s the kind of result that makes a media buyer’s quarter. The Italian iGaming popunder market in Q2 2025 was an unusually clean window — pre-cycle of the Q3 publisher consolidation — and the same campaign in Q3 2026 would not deliver the same numbers. But the structural shape of the comparison holds. For an impulse-friction offer in a market where popunder publisher supply is healthy and the operator has the licensing to run, popunder wins on CPA and it isn’t close.

Dating in the US, Q4 2024

A second case, abbreviated. A dating affiliate ran a US tier-1 dating offer (CPL on a free signup, payout $3.20) across popunder (adsy.tech, $0.50 CPM minimum) and Google Ads display network. We ran 30 days. Popunder generated leads at $1.10 CPL on average. Google Display generated leads at $4.60 CPL. The lead-to-rebill rate was 8 percent on popunder and 14 percent on Google. Net economics: popunder CPA per rebill was $13.75, Google was $32.85. Popunder won by a factor of 2.4. Not a slaughter, but consistent and easy to scale.

Crypto exchange registration funnel, Q3 2025

A third. An unregulated offshore crypto exchange — not naming because the operator asked not to be named — ran a registration funnel through Q3 2025. The offer wasn’t eligible on Google Ads at all. Popunder on Adsterra at $5.20 CPM in tier-1 markets generated registrations at roughly $3.40 per signup. The same operator ran Telegram channel boosts (paid placement in crypto Telegram channels) at $6.80 per signup and Reddit (sponsored posts) at $9.20 per signup. Popunder was the cheapest channel by a wide margin. The catch: regulatory letters arrived three months in and the campaign got paused. That’s the crypto popunder story in microcosm — works until it doesn’t, and the “until it doesn’t” is regulatory rather than economic.

The verticals above are not exhaustive. The point isn’t to enumerate every winning popunder case. The point is that when the offer is impulse-friction, when the audience tolerance for popunder is high, when the GEO has matching publisher supply, and when the LTV math doesn’t require Quality-Score-style precision, popunder structurally beats Google Ads on CPA at a margin that surprises Google-native buyers.

Attribution: postback delay, day-7 vs day-1 conversion windows

This section is where the two platforms diverge most invisibly, and where most buyer-side mistakes live. Attribution.

Google Ads attribution is, by default, a 30-day click window and a 1-day view window for most conversion types, fully integrated with the Google Ads pixel, with a tight server-to-server option via the Enhanced Conversions feature. The attribution is calculated inside Google’s infrastructure, which is both a strength (clean reporting, model-based attribution) and a weakness (the model decides what counts).

Popunder attribution is fundamentally different. Most popunder networks integrate with affiliate trackers — Voluum, Bemob, RedTrack, TrafficStars — via a click-ID postback URL. When a conversion happens, the affiliate tracker fires a postback to the network’s postback URL with the click ID and the conversion value. The network then attributes the conversion to the originating impression and updates the dashboard. The whole flow is server-to-server, which is one of popunder’s structural strengths against cookie-based attribution erosion.

The two attribution models behave differently in two critical ways: delay and window.

The delay problem

Popunder conversions, particularly on iGaming, dating, and finance, don’t happen in the first session. The user clicks the popunder, lands on the offer, may or may not register on the spot, and the qualifying conversion event — a deposit, a confirmed signup, a funded account — happens hours or days later. The postback fires when that downstream event occurs.

The implication for daily reporting: the popunder dashboard for today’s traffic is wrong. Today’s CPA looks much higher than the actual CPA because today’s conversions are still pending. The dashboard catches up over the next 24 to 72 hours, sometimes longer for finance and crypto.

Most buyers ignore this and make their first optimisation pass based on day-one numbers. This is the single most expensive operational mistake I watched popunder buyers make at PropellerAds. They paused profitable campaigns on day one because the CPA looked terrible, missing the day-three reconciliation when the postbacks caught up and the campaign was clearly winning.

The rule I use: do not make any campaign-level decisions on popunder traffic less than 72 hours old. The first three days of campaign data are unreliable as a basis for optimisation — my own heuristic, but it’s held up across hundreds of campaign launches. Look at day-five and day-seven cohorts. If those still look bad, the campaign is bad. If those look good, the day-one panel was just lagging behind reality.

Google Ads doesn’t have this delay problem to the same extent because the conversion pixel fires inline with the user session and the attribution model is integrated into the same platform that’s reporting the click. There’s still some delay on view-through and on long-window verticals, but it’s minor compared to popunder’s postback lag.

The window problem

This is the more subtle attribution issue. The popunder conversion window most networks use by default is somewhere between 7 days and 30 days from click. Some networks let you configure this, some don’t, and the value isn’t always the same as the value your affiliate tracker is using on its side. The first time I audited a campaign and found that the network was attributing on a 7-day window while the tracker was attributing on a 30-day window, the discrepancy was 22 percent — almost a quarter of conversions were being credited by the tracker but not by the network, and the buyer was reconciling at the lower number every month thinking the popunder was underperforming.

The window discrepancy compounds across networks. If you’re buying on three popunder networks plus Google, and the conversion windows are 7, 14, 30, and 30 days respectively, your cross-platform attribution model is broken before you start. Same conversion, different credit, different platform takes the win, and the cross-platform analysis you’re doing in the spreadsheet is sand on sand.

The fix is to standardise the window in your tracker and verify it matches the network’s setting on the postback side. Voluum lets you do this explicitly. Bemob does. RedTrack does. The default values are different across all three, and the default values on the network side are different across PropellerAds, Adsterra, RichAds, and adsy.tech. There is no industry standard. You have to set it yourself.

Day-1 versus day-7 cohort behaviour

The most useful attribution discipline I built at PropellerAds was the day-1 versus day-7 cohort comparison. The basic move: for every campaign, record the conversion count attributed at the end of day-1, then again at the end of day-7. The ratio between those two numbers is the campaign’s attribution lag profile.

A campaign with a day-7/day-1 ratio close to 1.1 is one where most conversions happen quickly. Utility installs and short-form lead-gen tend to land here. The day-1 panel is reliable for these. A campaign with a day-7/day-1 ratio of 1.6 to 2.2 is one where the postback lag dominates. iGaming deposits, dating rebills, and most finance conversions land here. The day-1 panel for these campaigns is misleading by a factor of nearly two — meaning if you’re judging the campaign on day-one numbers, you’re seeing roughly 50 to 60 percent of the eventual outcome.

The two case studies I described in the previous section both had day-7/day-1 ratios above 1.8. If I’d made the optimisation decisions on day-one panel readings, I’d have paused the winners.

The Google Ads side has its own version of this, but it’s less severe because the in-platform attribution model corrects for some of the lag automatically. The popunder side, with its server-to-server postback structure and no model overlay, exposes the raw lag to the buyer. That’s both a strength (no model interfering with your data) and a liability (you have to do the analysis yourself).

The reconciliation gap, panel to CRM

The third attribution issue, which is the one that costs the most money over time, is the reconciliation gap between the network panel’s reported conversions and your CRM’s actual recorded conversions. I’ve audited dozens of campaigns where this gap was 15 to 40 percent.

Some of the gap is fraud — bots clicking through to the offer and triggering pixel events that don’t translate to real CRM records. Some is tracker double-counting. Some is window mismatch as described above. Some is just bad pixel implementation on the advertiser side. The cumulative effect: the buyer thinks they’re paying $8 per conversion on the panel and the CRM shows $11 per real customer.

The discipline that closes this gap is server-side validation. The network panel reports a conversion, the advertiser server validates the conversion against the CRM, and the campaign optimisation is done on the validated number, not the panel number. Most popunder buyers in 2026 still aren’t doing this — my read of the affiliate-tracking market is that server-side validation adoption is somewhere under 40 percent of active buyers, based on Bemob and Voluum integration data. Of the buyers who are doing it, almost all of them are running profitably. There’s a strong correlation between operating discipline on attribution and operating profitability on popunder.

If you take one operational change from this article it’s this: build the postback flow with server-side validation. Bemob and Voluum both support it. RedTrack supports it on the enterprise tier. The setup is a half-day of work. The payoff is the ability to optimise on real numbers, not panel numbers, and the difference is between sustainably profitable popunder buying and the kind that flames out three months in when the buyer notices the panel-CRM gap.

Methodology: how I ran a parallel buy across both

For anyone wanting to replicate the comparison framework I used for the numbers in this post, here’s the parallel-buy template I ran across November 2024 to April 2026.

The setup, abbreviated. One landing page per offer, served from the same hosting, with the same load time, same form fields, same CTA. One Voluum workspace with two campaign objects: campaign A pointed at popunder traffic sources, campaign B pointed at Google Ads. UTMs structured identically on both sides — utm_source as the network, utm_medium as the format, utm_campaign as the offer code, utm_content as the creative variant, utm_term left blank on popunder (no keyword) and populated with the search keyword on Google. Sub-IDs on the popunder side set to publisher zone, GEO, device, browser, OS, time-of-day bucket. That gave me ten sub-ID slots in Voluum populated on the popunder side and just the four UTMs on the Google side, which is fine because Google Ads exposes its own click report.

The conversion pixel fired server-side from the offer’s CRM webhook to Voluum, with Voluum’s postback URL configured for each network. Server-side validation enabled. The conversion window standardised to 7 days everywhere — short enough to be operationally workable, long enough to catch the day-3 to day-5 deposit conversions that dominate iGaming and dating. Where the network’s default window was 30 days I set it to 7. Where it was 7 I left it.

Budget was matched per offer between popunder and Google in dollar terms, not in impression terms, because the impression counts diverge by three orders of magnitude. Match the spend, observe what the spend buys you, optimise within each side independently for the first two weeks, then compare normalised CPAs after the day-7 cohort is stable.

I gave each side a two-week stabilisation period before recording data. Google Ads needs the model to warm up. Popunder needs the buyer (me) to clean up the publisher zone blacklist. The first two weeks are noise.

After two weeks I recorded daily CPA, weekly CPA, day-1/day-7 conversion ratios, panel-to-CRM reconciliation rate, and creative variant performance. Twelve-week minimum per offer. The headline numbers in the tables above come from these twelve-week medians.

A few methodological caveats I want to call out so I’m not pretending to better precision than I have. First, my sample is six verticals, not the full universe. There are verticals I didn’t test — most native ad verticals, most CTV, most TikTok-adjacent influencer offers. The popunder-vs-Google comparison for those doesn’t generalise from the data I have. Second, the GEO coverage is heavier on EU and LATAM than on US or APAC, because that’s where my client book sits. Third, the network coverage is the four I named — adsy.tech, Adsterra, PropellerAds, RichAds — and excludes some others that may price differently. RichPush, ClickAdu, Monetag, Adcash, TrafficStars all run popunder; my numbers above do not include them. Where I’ve extrapolated to “popunder average”, I’m averaging across the four I ran, not the whole market.

The transparent piece of the methodology: every number in this post can be reconstructed if you replicate the setup above. The opaque piece: my private client notes from 2018–2023 are not reproducible because they’re not mine to publish. Where I’m citing those I’m trusting you to take my word, and where I’m citing the parallel-buy tests I can show the Voluum exports to a skeptical reader on request.

If your numbers disagree with mine when you run this template, I want to hear about it. The verticals and GEOs I tested are a slice. Your slice may show a different ratio. The methodology is more reproducible than the numbers; the numbers will drift with the quarter and the publisher supply.

The honest verdict + decision framework

Here is the decision tree I’d hand to a media buyer asking which platform to run their next campaign on. It’s blunt and it loses nuance, which is the cost of being usable.

Question 1: Can your offer legally run on Google Ads in your target GEO?

If no, the comparison is moot. Popunder by default, with the channel choice being which popunder network, not whether to use popunder. Continue to question 2 to decide the network.

If yes, continue to question 2 on both sides.

Question 2: What is the friction window of your offer’s conversion?

If the user can convert within five seconds of clicking and the conversion is a single discrete action (signup, install, single-field form, simple deposit) — popunder is the right buy. Skip to question 4.

If the user takes more than five seconds, requires comparison shopping, or the conversion is multi-step (B2B SaaS trials, mid-market software, regulated medical, real estate, anything requiring a callback) — Google Ads is the right buy. Skip to question 5.

If the conversion is an impulse-friction action but in a tier-1 EU regulated vertical (iGaming, finance, some health) where popunder publisher supply is thinning — run both and compare actual numbers over a 12-week window. Don’t trust this article’s numbers; trust your own.

Question 3: What is your LTV per conversion, and what is your tolerable CPA?

If LTV is under $30 and tolerable CPA is under $5, popunder is structurally fitted to your math. Google Ads will be too expensive for most offer types in this LTV band.

If LTV is over $300 and tolerable CPA is over $50, both platforms are viable. Choose on attribution preference (see question 6).

If LTV is over $1,000 and the conversion path is intent-driven, Google Ads wins. Popunder cannot price-discriminate at the high end of LTV without burning budget at the low end.

Question 4: For popunder, which network?

This is a longer post in itself, but as a one-liner: adsy.tech is the lowest entry price at $0.50 CPM, useful for cheap-end testing. Adsterra is the broadest publisher inventory for tier-2 and tier-3 GEOs. PropellerAds is the deepest for tier-1 GEOs and the most mature dashboard. RichAds is the strongest for push-popunder hybrid running. If you’re new to popunder, start with adsy.tech to keep your test cost low, scale to PropellerAds or Adsterra once you have a working campaign template.

Question 5: For Google Ads, what’s the bidding strategy?

Manual CPC if you have under 30 conversions per month per ad group. tCPA if you have 50+. tROAS if your conversion values vary widely. Maximize Conversions only if you can stand the spend volatility of the first two weeks. Smart Bidding rewards Quality Score and punishes poor landing pages; if your LP isn’t sharp, fix the LP before changing the bid strategy.

Question 6: Attribution discipline.

Whichever platform you choose, set the conversion window to 7 days on the tracker side, set it explicitly to match on the network side. Enable server-side conversion validation. Look at day-7 cohort data before making any campaign-level decision. Do not pause campaigns on day-one panel readings.

The verdict, as a single sentence: popunder beats Google Ads on cost-per-acquisition for impulse-friction offers in verticals and GEOs where the publisher supply is healthy and the regulatory posture allows; Google Ads beats popunder on offers requiring intent, brand defence, local-buying behaviour, or audit-friendly reporting. There is no platform that wins everywhere. The buyers who do well are the ones who can run both and have the attribution discipline to know which is which without fooling themselves.

A note on what changes if you scale

The numbers in this post are at the volume profile of an affiliate or small in-house team — a few thousand to a few tens of thousands of dollars per offer per month. At higher spend the dynamics shift in ways that aren’t obvious. On Google Ads, scaling past roughly $50K per month per campaign tends to drive CPCs up because the auction starts including marginal queries the model wouldn’t otherwise show your ad on, and the Quality Score advantage flattens. On popunder, scaling past roughly $30K per month per offer on a single network tends to exhaust the cleanest publisher zones first; the marginal impression you buy at month-three is in a lower-quality zone than the marginal impression at month-one, even though the panel CPM looks the same. The implication: if you’re scaling, run on at least two popunder networks in parallel and use Voluum to compare publisher-zone-level CPA across them. The diversification keeps you out of the diminishing-returns curve any single network’s inventory will eventually hit.

The math doesn’t lie, but the panel can. If you do nothing else from this post, build the attribution discipline. Everything else flows from that.

FAQ

Is popunder cheaper than Google Ads in 2026?

Popunder CPM is structurally lower than Google Ads CPM by a factor of roughly 4x to 45x depending on the vertical, with the largest gap in utility installs and the smallest in regulated tier-3 iGaming. Whether it’s actually cheaper for your offer depends on conversion quality, fraud profile, and attribution lag. The headline price gap is real; the after-friction price gap is smaller and varies by vertical.

Why is Google Ads CPM so high?

Because Google Ads isn’t really pricing impressions, it’s pricing predicted conversion value through Smart Bidding’s model. For high-intent verticals like finance, iGaming, and legal services, the predicted conversion value is high enough that the auction clears at very high CPCs. The CPM is a derived number from the CPC, not a directly-priced unit. Popunder prices the impression directly, which is why its CPM looks lower even when the cost-per-real-customer can be similar.

Can I run iGaming on Google Ads?

In licensed regulated markets and with the right verification posture, yes — Italian iGaming with ADM licensing, Spanish with DGOJ, German with the GlüStV process, US with state-level licensure. Approximately 70 percent of would-be iGaming advertisers in tier-1 EU markets can’t pass the verification, per EGBA data and my own observation inside PropellerAds. The licensing barrier is the structural reason popunder remains the dominant channel for unlicensed and offshore iGaming.

What’s the typical popunder conversion window?

Network defaults vary from 7 days to 30 days. I’d recommend setting your tracker to 7 days and the network to match, which captures the day-3 to day-5 deposit conversions that dominate iGaming and dating without inflating attribution to unstable longer windows. Most networks let you override the default in your campaign settings.

How much fraud should I expect on popunder traffic?

The 2019 IAB benchmark put programmatic fraud at around 26 percent non-human traffic across the industry. Popunder-specific fraud rates vary widely by network and publisher zone. Cleaner networks like PropellerAds in tier-1 and adsy.tech with active publisher curation tend to run 5 to 15 percent fraud. Worse networks can be 30 percent or more. The fraud is detectable with server-side validation and publisher zone analysis; the buyers who skip that discipline pay the fraud tax invisibly.

Can I use the same landing page on popunder and Google Ads?

Technically yes, operationally not always. Google Ads will grade your landing page for Quality Score and prefer pages with low cognitive load, clear CTAs, mobile speed under 2 seconds, and policy-compliant copy. Popunder has no Quality Score; the landing page just needs to convert within seconds of opening. The same page can work on both, but if it’s optimised for Google’s Quality Score it may underperform on popunder where speed and impulse-friction matter more than relevance signals.

What’s the difference between popunder and push notification ads?

Popunder opens a browser window behind the active tab; push delivers a notification to a subscribed user’s device or browser. Popunder targets anyone visiting a publisher zone; push targets users who’ve subscribed to a publisher’s notifications. Push has higher intent because it’s opt-in; popunder has higher volume because it doesn’t require prior subscription. Different formats for different offer profiles. Most networks sell both.

Does popunder still work in 2026 with all the privacy regulation?

Yes, with caveats. GDPR and ePrivacy have squeezed popunder publisher supply in seven EU markets (FR, IT, ES, DE, NL, PL, BE per the EDPB 2024 summary). Apple’s ITP and Safari restrictions have raised friction on iOS popunder serving. Server-side attribution has largely replaced cookie-based attribution. The format itself works because it doesn’t depend on third-party cookies for delivery; it depends on publisher script execution. The regulatory squeeze is real and ongoing but hasn’t ended popunder; it’s reshaped where and how it runs.

Which popunder network should I start with?

adsy.tech for the $0.50 CPM minimum if you’re testing cheaply, Adsterra for breadth of tier-2 and tier-3 GEOs, PropellerAds for tier-1 maturity, RichAds for push-popunder hybrid running. I’d avoid networks with no public rate card or no clear publisher curation policy. The four named here all publish rate cards and have active fraud teams; not every network does.

How do I attribute popunder if I’m running on Google Ads too?

Standardise the conversion window across both platforms on your tracker side. Use server-side conversion postbacks. Set up two separate conversion tags or pixel events — one fired from popunder traffic, one from Google traffic — and reconcile both against your CRM record on a 7-day cohort basis. Don’t trust cross-platform attribution from either platform’s native interface; build it in your tracker (Voluum, Bemob, RedTrack) where you control the window and the validation logic. This is the operational discipline that separates buyers who scale from buyers who quit popunder thinking it didn’t work when it was really their attribution model that didn’t work.