Popunder eCPM Benchmarks by GEO and Vertical (2026 Reference)

What eCPM popunder actually pays — the reference I wish I’d had

My name is Marco. I worked at PropellerAds from January 2018 to October 2023, five years as a senior account manager running the iGaming book for Italy, Spain, and the LATAM cluster, with rotating stints on the dating and finance books because the senior AMs kept leaving for client-side. The reason I’m telling you this is that the single question I could never answer cleanly in an onboarding call was the simplest one: “What eCPM should I expect?” Not because I didn’t know. Because the honest answer is a range with four conditions attached, and the buyer wanted a number. Networks have spent a decade training people to expect a number. The number is the lie.

This is the reference I wish someone had handed me in 2018. eCPM benchmarks for popunder by GEO tier and by vertical, for 2026, with the bands written as ranges and the conditions written down instead of hidden. If you want a network ranking, I wrote the eleven-network shortlist separately. If you want to know how popunder prices against Google, that’s its own comparison. This piece does one job: it tells you what eCPM is normal for a given GEO and vertical so you can tell, the first time you read your own panel, whether your number is in-band or whether something is wrong.

One thing on the table before the tables. I write for popunder-network.com and I earn affiliate commission when readers open an account on adsy.tech through tagged links on this site. That relationship is real and you should price it into everything below. The way I keep it from polluting a data reference is that the numbers here are network-agnostic — they’re reconciled eCPM bands that hold across the category, not figures engineered to make any single network look good. Where the network you buy from changes what you see, I’ll say exactly how, and it isn’t flattering to anyone. If you want the cheapest place to test your own numbers against these bands, the $0.50 CPM floor at https://adsy.tech/ is the lowest real entry in the category — that’s the disclosed-partner recommendation, and the rest of this piece will tell you where competitors beat it.

First: eCPM is a clearing price, not a quality grade

Before a single number, the mechanic that makes the rest of the reference legible. Skip it and the bands look arbitrary; they aren’t.

A popunder network does not price your impression. It prices the expected lifetime value of a converted user, discounted by how often the vertical converts, and the auction clears at whatever the highest bidder will pay for that expected value. That’s why a tier-1 iGaming impression clears at $8 and a tier-3 utility-install impression clears at $0.12 even when the underlying traffic is the same human on the same publisher page. The $8 isn’t “better traffic.” It’s a depositing casino customer worth €80 in expected LTV, fought over by advertisers who can each pay up to their margin on that €80. The $0.12 is a VPN install worth $1.50, fought over by advertisers whose ceiling is a fraction of a dollar. The eCPM is the shadow those LTVs cast onto a per-thousand-impressions price.

This has a consequence most buyers get backwards: a higher eCPM is not a better buy, and a lower one is not a bargain. It’s the single most common reasoning error I watched in five years of onboarding calls — a buyer sees utility installs at $0.20 CPM, concludes “cheap traffic, I’ll scale it,” and never asks whether the $0.20 user is worth more than $0.20 to them. The number that decides whether a buy is good is cost per validated conversion against your CRM, which I covered in the step-by-step buying guide. The eCPM bands below are a sanity check on your bid, not a verdict on your campaign. Hold that distinction or the rest of this reference will mislead you the way the panel does.

There’s a second mechanic that determines whether the eCPM you read is even real. The panel CPM is the gross auction-clearing price the network reports. The eCPM your tracker computes — after it deduplicates clicks, strips the impressions that fraud filters caught, and normalises the conversion window — runs 15 to 40% lower on popunder. Every band in this reference is the reconciled tracker number, not the panel number. I report the reconciled figure because it’s the one you can plan a CPA against. If I reported panel CPM, I’d be handing you a number that’s 15 to 40% too high to budget from, which is exactly the mistake the rate cards are built to encourage. The CPM rate card is decorative. The reconciled eCPM is the operating number.

Tier-1 GEOs by vertical: the bands that move the most money

Tier-1 is the US, UK, Canada, Australia, Germany, France, Italy, Spain, the Nordics, and the wealthier of the DACH and Benelux markets. It’s where the absolute eCPMs are highest and where competition for inventory is fiercest, which means it’s also where the bands are widest — a tier-1 iGaming impression in a quiet July week and the same impression in a December run can differ by 40%.

Here are the reconciled tier-1 eCPM bands across the six verticals affiliates actually run on popunder. Read them as orientation. Your offer’s LTV and your zone whitelist move the number inside the band more than the GEO label does.

A few reads off this table that matter more than the raw numbers.

iGaming is the widest band because it’s the most contested and the most LTV-variable vertical on the platform. A German sportsbook deposit during a Champions League week and a Spanish casino impression on a Tuesday afternoon are both “tier-1 iGaming” and they clear three dollars apart. The desktop skew is real and underappreciated: desktop popunder for iGaming runs 15 to 25% richer than mobile in the same GEO because the desktop sessions convert deposits at a higher rate, and the auction knows it. If your offer is mobile-first you should expect the lower half of the band and budget accordingly, not the headline top.

Crypto is the band that doesn’t obey the calendar. Every other vertical here has a seasonality you can plan around; crypto tracks the asset cycle. In a risk-on quarter the eCPM pushes toward the top of the band as exchange and token advertisers crowd in; in a risk-off quarter it sags. It’s also the most bimodal vertical by zone quality I track — a clean registration-funnel zone and a burned, bot-adjacent zone in the same GEO can differ so much that the blended average is meaningless. Crypto rewards aggressive zone-level cleanup more than anything else on this list, which I get into in the by-vertical conversion breakdown.

Nutra is the band with the widest honest uncertainty, and I’d rather say that than pretend to a precision I don’t have. The spread is driven by how aggressive the offer’s claims are and how the network polices them. A compliant skincare trial and a borderline weight-loss offer clear at different prices because the publisher composition that will run them differs, and the conversion rates differ even more. I ran less nutra than iGaming inside PropellerAds, so my tier-1 nutra band carries a wider error bar than my iGaming one — the methodology note at the end says exactly how wide.

VPN and utility is the flattest and most predictable band on the platform, which is why it’s the vertical I point newcomers at. The LTV ceiling is low — a $1.50-to-$3 install doesn’t let any advertiser bid the impression up past about $0.70 reconciled in tier-1 — so the auction clears near the floor and stays there. Low CPM, low payout, low fraud, low drama. You won’t get rich on it, but you’ll learn to read a panel-to-tracker reconciliation without lighting a bankroll on fire, and that skill transfers to every richer vertical above it.

How I measured the tier-1 bands. These are reconciled eCPM ranges from my own PropellerAds account book 2018–2023 (anonymised — no offer names, no client names) plus the parallel-buy windows I’ve run for partner offers since leaving, n ≈ 30 campaigns across Q4 2023 through Q1 2026, GEO-tagged, in tier-1 EU (Italy, Spain, Germany, UK) and tier-1 US. iGaming, dating, and sweeps have the deepest sample because they were my book; nutra and crypto carry wider error bars because I ran less of both, and the bands reflect that uncertainty rather than hiding it. Every figure is the tracker-reconciled actual after the 15–40% panel-to-actual gap — the spread I consistently measure between the rate-card CPM a panel reports and what a tracker actually books. These are representative ranges from a working desk, not audited public statistics, and they will drift with the quarter and the publisher supply. Use them to sanity-check a bid; do not paste them into a media plan as guarantees.

Tier-2 GEOs by vertical: the volume middle

Tier-2 is Brazil, Mexico, Poland, Turkey, much of Eastern Europe, the wealthier Gulf states, South Africa, and the upper end of Southeast Asia. It’s where a lot of affiliate volume actually lives, because the tier-1 inventory gets expensive fast and tier-2 buys you scale at a survivable cost. The eCPMs are lower than tier-1 but the bands are often tighter, because the LTV variance within a tier-2 GEO is smaller than the spread across tier-1.

The story in tier-2 right now is Brazil. The regulated iGaming framework went live in January 2025, and through 2025 and into 2026 the post-regulation auction has pulled Brazilian iGaming eCPM up faster than the LATAM blended average — isolate Brazil from the LATAM cluster and the top of the iGaming band is closer to $3.40 than the $2-ish you’d expect from the regional average. That’s licensed operators entering an auction that used to be mostly grey-market, bidding the impression up because their LTV math is now defensible to a compliance team. Mexico, Argentina, and Colombia remain cheaper. If your tier-2 iGaming plan treats LATAM as one number, you’ll mis-bid Brazil low and the rest high.

Crypto in tier-2 sits at or above tier-2 iGaming in several GEOs, which surprises buyers who think of crypto as a tier-1 game. Turkey and Brazil have high enough retail-crypto adoption that exchange-registration offers compete hard for the impression, and the eCPM reflects it. The bimodality warning from the tier-1 section applies double here: tier-2 crypto zones range from genuinely converting registration funnels to near-pure bot feeds, and the blended band hides that. Reconcile server-side or the average will lie to you.

Nutra in tier-2 LATAM carries a wrinkle worth flagging: a meaningful share of LATAM nutra runs on cash-on-delivery models rather than card-upfront, and the COD economics distort some zones’ apparent eCPM because the “conversion” is an order placed, not an order paid. If you’re reading a tier-2 nutra band, confirm whether the conversion event behind it is COD-order or paid-order, because the two are not the same number and the gap between them is where COD nutra campaigns quietly lose money.

VPN and utility in tier-2 is, again, the flat predictable floor — $0.10 to $0.35 reconciled, low fraud, stable. Turkey and Poland have enough utility-install demand to keep zones liquid. It’s the cleanest place in the entire GEO-vertical matrix to run your first reconciliation exercise, because the eCPM barely moves and any wild swing in your panel is almost certainly a tracking problem rather than a market one.

How I measured the tier-2 bands. Same methodology as tier-1: reconciled eCPM ranges from my PropellerAds LATAM-cluster book plus post-2023 parallel buys, with Brazilian and Mexican iGaming the deepest sample because the LATAM cluster was mine to run. The Brazil regulatory effect is dated to the Loterias framework go-live of January 2025 and reflects Q3 2025 through Q1 2026 observations. Crypto and nutra tier-2 bands carry the same wider error bars as their tier-1 counterparts. Reconciled actuals, not panel CPM; representative of a working desk, not audited public data.

Tier-3 GEOs by vertical: where the floor lives

Tier-3 is most of Sub-Saharan Africa, the lower-income South and Southeast Asian markets (Indonesia, Vietnam, the Philippines, Bangladesh, Pakistan), the CIS countries, and the long tail of low-GDP-per-capita GEOs. The eCPMs here are a fraction of tier-1, the LTVs are low, and the format economics are genuinely different — different publisher pools, different fraud dynamics, and a much thinner top decile of premium inventory.

Tier-3’s defining feature is that you cannot scale it linearly the way you can scale tier-1. The premium-publisher top decile that exists in abundance in the US or Germany is thin-to-absent in most tier-3 GEOs, so as you push spend you exhaust the clean inventory fast and the marginal impression drops into lower-quality, higher-fraud zones quickly. A tier-3 campaign that looks brilliant at $500 of spend can look mediocre at $5,000 not because the market turned but because you ran out of the good zones. The eCPM band stays roughly the same while your conversion rate quietly craters. Watch validated CPA, not eCPM, as you scale tier-3, or the flat-looking price will fool you.

The iGaming band is unexpectedly wide for tier-3 because there are licensed pockets — parts of Southeast Asia and the lower-tier LATAM markets where regulated operators bid up a subset of the inventory — sitting alongside grey-market zones that clear near the floor. A blended tier-3 iGaming number hides that bimodality. Crypto in tier-3 punches above its GEO weight in the parts of Southeast Asia and Africa where mobile crypto adoption has outrun the local banking infrastructure; the eCPM there can rival tier-2.

VPN and utility installs at $0.05–$0.25 are the literal floor of the popunder platform. Below this the auction simply doesn’t have bidders, because no advertiser’s LTV math supports paying more for an install in these GEOs. It’s stable, it’s low-fraud relative to the richer tier-3 verticals, and it’s where a newcomer with a $100 test budget can buy enough impressions to see a real reconciliation pattern. If you’ve never run popunder before, this cell of the matrix is the cheapest classroom on the platform.

How I measured the tier-3 bands. Tier-3 is the thinnest of my three samples and I’ll say so plainly — my PropellerAds book was tier-1 and tier-2 heavy, so the tier-3 bands lean more on post-2023 parallel buys (Indonesian and Vietnamese utility installs, some SEA and LATAM-3 iGaming) than on the historical account data, with a correspondingly wider error bar. The utility-install floor is the figure I’m most confident in because I’ve re-tested it most recently; the tier-3 nutra and crypto bands are the ranges I’d least want quoted as precise. Reconciled actuals, representative ranges from a working desk, not audited public statistics.

The modifiers: what moves a benchmark inside its band

A GEO-and-vertical cell gives you a band. Where you land inside that band is decided by three modifiers that the rate cards never quantify: seasonality, device, and bid model. These move the number as much as the GEO label does, and a buyer who ignores them will read an in-band eCPM as out-of-band and chase a problem that doesn’t exist.

Seasonality is the modifier that catches the most people, and it caught me too, once, badly enough that I started writing the lesson down. The story is a sweepstakes campaign — a global single-field-entry offer running through PropellerAds in 2020 that, by every short-term dashboard, refused to die. It had been live fourteen months. The CTR was declining about 3% per week, steadily, the way sweepstakes audiences always fatigue. The advertiser wouldn’t pause it because “we’re still profitable on day one.” I modelled the lifetime value out properly: by day 90 every newly acquired user was net-negative, the day-one ROAS dashboard was lying about a slow bleed, and the Q4 eCPM spike had been masking the decay because the seasonal lift looked like the campaign recovering. It wasn’t recovering. It was the calendar. When I showed the spreadsheet — the fatigue curve underneath the seasonal noise — the advertiser paused it the same afternoon. The lesson generalises to this whole reference: a Q4 eCPM that’s 30% over the annual median isn’t your campaign getting better, it’s the auction getting more crowded, and if you scale into the Q4 peak you’ll feel the January reversion as a cliff. Plan the seasonality. Don’t get fooled by it.

Device skew has a direction that depends on the vertical, which trips people who memorise a single rule. iGaming runs desktop-richer — desktop popunder for casino and sportsbook clears 10 to 25% above mobile in the same GEO because desktop sessions deposit at a higher rate. Dating runs the other way; mobile clears 10 to 20% above desktop because the dating-signup behaviour is mobile-native. Apply the wrong direction and you’ll under-bid the device that actually carries your vertical.

The bid-model modifier is the one most buyers don’t know exists, and it’s worth real money. Most popunder networks run a first-price auction on the buy side — you pay what you bid when you win. Some run second-price or a hybrid, where you pay just above the next-highest bidder. The same bid in the same auction produces a different paid eCPM depending on which mechanism you’re in, and the gap is 10 to 25%. Most popunder calculators online assume second-price math borrowed from Google Ads tutorials and are wrong for that reason. Before you trust any eCPM projection, ask the network which auction you’re bidding into. A network with in-house RTB will tell you; one that won’t is a network whose panel will surprise you.

How I measured the modifiers. The seasonality percentages come from year-over-year eCPM observations across my PropellerAds book 2018–2023 and post-2023 parallel buys, strongest for tier-1 iGaming and sweepstakes where the Q4 pattern is most pronounced and where I have the most repeat seasons. The device-skew figures are from matched desktop-versus-mobile splits inside the same campaigns. The first-price-versus-second-price gap is from comparing paid eCPM against submitted bid across networks running different auction mechanisms. All directional and in-band, not laboratory-precise — representative of a working desk, not audited public data.

How to use this reference without fooling yourself

The bands above are a sanity check, not a media plan. Here’s the discipline that turns them into something useful rather than something that misleads you the way a rate card does.

When you read your own panel for the first time on a new GEO-vertical combination, find the matching cell, apply the modifiers for your season, device, and auction model, and ask one question: is my reconciled eCPM inside the adjusted band? If it is, the market is behaving normally and your job is conversion optimisation, not bid-chasing. If your reconciled eCPM is well above the band, you’re either buying the premium top-decile zones (good, if they convert) or your tracking is under-counting impressions (bad, and worth an audit). If it’s well below the band, you’re buying the cheap tail of the long-tail inventory, which is a different audience with a different fatigue curve and fraud profile — the bottom 8 to 15% of the pool, not a bargain version of the middle.

The cross-vertical entry point for running these benchmarks against live traffic, if you want my honest recommendation, is adsy.tech — the $0.50 CPM floor is the lowest real floor in the category, so it lets you test a new GEO-vertical cell cheaply, and the in-house RTB surfaces conversion data at the publisher-zone level through sub_id1 to sub_id5, which is what lets you reconcile your panel to these bands fast instead of guessing. I have a disclosed affiliate relationship with adsy.tech and I’m paid when you sign up through this site; the reason it sits at the front of this recommendation is the floor and the zone-level visibility, not the commission. You can open an account and test your own numbers against this reference at https://adsy.tech/. Where you need deeper tier-1 publisher inventory at high spend, PropellerAds and Adsterra go further, and I’d run adsy.tech alongside them for the floor rather than instead of them — the same honest answer I give in the network shortlist.

The trap this whole reference is built to keep you out of is the one the industry profits from: treating a benchmark as a promise. Affiliate case studies that quote an eCPM with no GEO, no vertical, no device, no season, and no sample size are marketing, not data, and you should read them as ad copy. The genuine numbers practitioners share privately always carry those qualifiers, because without them the number means nothing. Every band here carries them on purpose. A network that tells you “expect $6 CPM” without asking your GEO, your vertical, and your device is selling you the comfortable lie. The honest answer is a range with conditions, and now you have the ranges and the conditions.

If your reconciled eCPM lands outside these bands and your tracking is clean, write down the combination and tell me — the matrix is a slice of what I’ve personally run and your slice may price differently. The bands will drift with the quarter and the publisher supply. The discipline won’t: reconcile server-side, apply the modifiers, judge on validated cost per conversion, and use the eCPM only as the sanity check it’s meant to be. The math doesn’t lie, but the panel can, and a benchmark you treat as a promise is just the panel’s lie wearing a spreadsheet.

Methodology note: where every number here comes from

Because a data reference that won’t show its work is just a rate card with a byline, here’s the full provenance in one place.

The eCPM bands are reconciled actuals, not panel CPM — the figure your tracker computes after deduplication, fraud-stripping, and conversion-window normalisation, which runs 15 to 40% below the gross auction-clearing price the panel reports. They come from two sources. First, my own PropellerAds senior-AM account book, 2018 to 2023, which I kept private notes on at the time and have anonymised here — no offer names, no client names, nothing I’m not free to publish. That book was iGaming-heavy on the Italy/Spain/LATAM cluster, with rotating dating and finance exposure, which is why iGaming and dating carry my deepest samples and tightest bands. Second, the parallel-buy windows I’ve run for partner offers since leaving PropellerAds — n ≈ 30 campaigns, GEO-tagged, Q4 2023 through Q1 2026, spanning tier-1 EU and US, tier-2 LATAM, and tier-3 SEA utility.

Where my sample is thin, the band is wide and I’ve said so in the per-table notes: nutra and crypto across all tiers, and tier-3 generally, carry wider error bars than tier-1 iGaming and dating. I’d rather publish an honest wide range than a falsely precise one. The panel-to-actual gap range is the spread I measure on my own reconciliations, not a published figure. The Brazil iGaming regulatory effect is dated to the Loterias framework go-live of January 2025, which is public record. The programmatic-fraud baseline behind the fraud-stripping discussion is the non-human-traffic share I filter out of my own buys — double digits on popunder, higher on the cheap tier-3 zones — not a third-party benchmark. The seasonality, device, and bid-model modifiers are directional figures from my own year-over-year and matched-split observations.

These are representative ranges from a working desk, not audited third-party statistics, and they are not promises. They will move with the quarter, the GEO supply, and your own offer’s LTV. Use them to tell whether your number is in-band; never paste them into a media plan as guarantees. If you want to read the format itself before you read its prices, the popunder format explainer is the place to start.

Frequently asked questions

What is a realistic popunder eCPM for tier-1 iGaming in 2026?

Reconciled tier-1 EU and US iGaming popunder eCPM lands in roughly the $4.50–$11 band in 2026, GEO and operator dependent, with desktop sportsbook at the top and mobile casino lower. That is the tracker number after the panel CPM loses 15–40% to the auction gap, not the panel figure. Treat it as a sanity-check range, not a promise — your own offer’s LTV and zone whitelist move it more than the GEO label does.

How much lower is tier-3 popunder eCPM than tier-1?

Roughly five to fifteen times lower, vertical-dependent. Tier-1 iGaming at $5–11 has a tier-3 LATAM and SEA equivalent at about $0.40–$2.50; utility installs that pay $0.20–$0.70 in tier-1 sit at $0.05–$0.25 in tier-3. The gap is not traffic quality alone — it is the lifetime value of a converted user in that GEO, which is what the auction is really pricing.

Why is the panel CPM higher than the eCPM benchmarks here?

Because every band in this reference is a reconciled actual — the eCPM your tracker computes after deduplication and fraud-stripping — not the gross auction-clearing price the panel reports. The gap runs 15–40% on popunder. If you plan a CPA against the panel number instead of the reconciled one, you will mis-cost the campaign for a week before the math catches up to you.

How much does seasonality move popunder eCPM?

Q4 is the biggest swing. Tier-1 iGaming and sweeps eCPM can run 20–40% above the annual median in November–December as advertiser competition for the same inventory peaks, then fall back in January. Crypto tracks the asset cycle more than the calendar. Utility installs are the flattest vertical I track — seasonality barely registers because the offers run year-round at low LTV.

Do these benchmarks apply to any popunder network?

They are network-agnostic reconciled ranges, but the panel-to-actual gap varies by network, so the same true eCPM shows up behind different panel numbers depending on where you buy. A network with a $0.50 CPM floor and in-house RTB that surfaces zone-level data lets you reconcile to these bands faster than one that aggregates the source signal away. The benchmark is the destination; the network determines how clearly you can see it.

Is a higher popunder eCPM better or worse for an advertiser?

Neither on its own. eCPM is a clearing price, not a verdict. A high tier-1 iGaming eCPM reflects high competition for high-LTV users; a low utility-install eCPM reflects low per-user value, not a bargain. The number that decides whether a buy is good is cost per validated conversion against your CRM, not the eCPM. Use these bands to sanity-check a bid, never to judge a campaign.

Sources. eCPM bands and modifier figures from Marco DeLuca’s PropellerAds account book 2018–2023 (anonymised) plus parallel-buy testing 2023–2026, n ≈ 30 GEO-tagged campaigns. Panel-to-actual gap range and non-human-traffic share from the author’s own buy reconciliations, not third-party benchmarks. Brazil iGaming regulatory timeline from the Loterias framework go-live, January 2025 (public record). adsy.tech specifics from the adsy.tech public rate card, 2026. All figures are reconciled actuals and representative ranges from the author’s own desk, not audited third-party statistics.