Ad network eCPM benchmarks 2026, by format and GEO

What ad formats actually pay by GEO in 2026

Methodology in the appendix. The benchmark below. If a number surprises you, the methodology will explain why; if the methodology has a hole, the address is in the footer.

The short version: in 2026, eCPM is set by format and country before it is set by anything you do. Across my parallel buys, a Tier-1 popunder cleared roughly twenty to forty times a Tier-3 push impression, and the highest-paying format on a given GEO routinely paid three to six times the lowest. The table below maps those ranges so you can tell whether your own eCPM, or a network’s quote, sits in the right neighbourhood — it is a market benchmark by format and GEO, not a price any single network guarantees.

I’m James. Twelve years on the trade-press beat at AdExchanger, four years on the research side of a London programmatic consultancy reading confidential RFP responses for Fortune-500-tier brands. This is the data asset the rest of this site links back to, so it is the one page where I show the most working. The reason I publish a format×GEO table instead of a network-by-network one is structural: an honest eCPM is a property of the inventory and the auction, not a fixed attribute of a logo, and the moment you publish “Network X pays $0.47” you are either quoting a rate card or fabricating a number. I will not do either. What I can publish — because I measured it across identical buys — is what each format paid across each GEO tier in aggregate. That is a modest thing to claim. It is not nothing.

The benchmark table — format × GEO, tracker-reconciled

Six formats down the side, three GEO tiers across the top, two to three example countries per tier. Every cell is a tracker-reconciled eCPM range in USD, computed after perimeter fraud filtering, aggregated across my parallel-buy test set for the Q4 2025–Q1 2026 window. These are not rate-card prices and not network-specific — they are what the format paid into that GEO tier across my buys, full stop. Read the methodology note directly below before you read a single figure as gospel.

The single most useful number in that grid is not any individual cell — it is the ratio between Tier-1 and Tier-3 on the same row, which runs from about fifteen-to-one on banner to past thirty-to-one on push. That ratio is the thing most “CPM by country” listicles flatten into a single global average, and the average is useless: a media buyer who plans a Tier-3 push campaign against a blended global push eCPM will mis-budget by an order of magnitude. Plan against the cell, never the row average and never the table average.

Methodology note — read this before you cite a figure

These ranges are illustrative eCPM bands from my own parallel-buy testing, not vendor rate cards and not guaranteed prices. Specifically: the window is Q4 2025 through Q1 2026; spend scale was roughly $2,000–$4,000 per format-GEO cell over rolling fourteen-day flights; the vertical mix was weighted toward mainstream direct-response and DTC offers (no adult, no high-risk verticals, which price differently); figures are eCPM reconciled server-to-server in Voluum against advertiser back-office data, computed after perimeter fraud filtering on a 7-day click window. Sample depth varied by cell with fill — the Tier-1 cells rest on more impressions than the Tier-3 ones, and the ranges widen where the sample thinned. Fourteen-day flights do not support seasonality claims, so the bands are point-in-time, not a forecast. Where I could not run a clean matched buy into a specific country, I have folded it into the tier band rather than reporting it as a standalone figure. Nothing in this table is attributed to a named network’s pricing, because no honest eCPM can be.

Why Tier-1 pays so much more than Tier-3

The eCPM gap between a US user and an Indonesian user is not a quirk of supply — it is a direct readout of advertiser willingness to pay, and that willingness tracks the conversion value behind the click. A Tier-1 user converts into a basket, a subscription or a deposit large enough to support a high bid; a Tier-3 user, on average and across the mainstream verticals I tested, does not, so the auction clears lower even for technically identical inventory. The roughly twenty-to-forty-times Tier-1-to-Tier-3 spread I saw on popunder is the auction pricing that difference in downstream value, nothing more mysterious than that.

This is why “cheap Tier-3 traffic” is a category error as often as it is an opportunity. The eCPM is low because the expected value is low; the buy only wins if your offer monetises Tier-3 users better than the average advertiser bidding into that auction, which is a real edge for some affiliate verticals and a trap for most DTC ones. The table gives you the going rate; whether that rate is a bargain depends entirely on your own conversion economics, and the only way to know is to reconcile against your tracker, not against the eCPM.

One caveat I have to state plainly: tier boundaries are fuzzy and the example countries are representatives, not the whole tier. Germany and the US both sit in Tier-1 but a German push eCPM and a US push eCPM can differ by half within the band, because GDPR-shaped consent flows and opt-in rates change the addressable push audience. Read the tier as a neighbourhood and the country examples as landmarks inside it.

Why the format changes the eCPM as much as the GEO does

Hold the GEO constant and the eCPM still moves three to six times across formats, because each format buys a different quantity of user attention. An interstitial takes the full screen; a popunder takes a full tab; both command a high bid because the impression is hard to ignore. A banner sits in a slot the user has been trained to skip, so it clears at the bottom of the range almost everywhere — in my Tier-1 data, banner eCPM was routinely a quarter to a third of interstitial eCPM for the same GEO and vertical.

Push and in-page push are the interesting middle. Push reaches a user who opted into a notification channel, which is why its Tier-1 eCPM holds above display despite a tiny creative; in-page push mimics the format inside a webpage without the opt-in, and it prices slightly below true push in my set for exactly that reason — the consent signal is weaker, so the bid is too. Native splits the difference between attention and intent: it earns a higher eCPM than banner because it blends into content, and that same blending is why a share of native clicks are accidental, a point I have hammered elsewhere and will only flag here.

The operational consequence: do not compare your push eCPM to someone else’s native eCPM and conclude one network is underpaying you. They are different products at different points on the attention curve. Compare like format to like format, in the same GEO tier, or the comparison tells you nothing.

What destroys eCPM accuracy — fraud, MFA and viewability

A reported eCPM is only as trustworthy as the impression count underneath it, and the impression count is exactly what fraud inflates. Made-for-advertising sites exist to manufacture impressions and clicks that never had a chance to convert; the ANA’s programmatic transparency work put MFA inventory at roughly 6.2% of programmatic spend in 2024, down from about 15% in 2023 — genuinely improving, but still enough to distort an eCPM read in the long tail where the cheap inventory lives. An eCPM measured against unfiltered impressions and one measured against filtered impressions are not the same metric, even from the same campaign on the same day.

This is why every figure in my table is computed after perimeter filtering, and why I treat any benchmark that does not state its filtering posture as unusable. The same caution applies to viewability: a banner impression that renders below the fold and is never seen still counts in a naïve eCPM denominator, dragging the number down in a way that has nothing to do with what the inventory is worth to a buyer who measures viewable impressions. Two numbers, two definitions; if a source will not tell you which one it is reporting, the number is decoration.

The practical defence is the same one I recommend for choosing a network at all: buy where you can exclude inventory at the source level. A format×GEO eCPM you cannot decompose to the publisher or sub-source is a number you cannot defend, because you cannot see which slice of it is real traffic and which slice is the long-tail MFA inflating your denominator.

How to read this table for your own buy

Find your format’s row and your GEO tier’s column, read the range, and treat the midpoint as your null hypothesis. Three outcomes, three actions. If your live eCPM lands inside the band, the buy is behaving like the market and your job is to optimise on conversion, not to renegotiate price — the price is normal. If it lands well above the band, audit fill and whether you are paying a placement premium you did not intend; a high eCPM with thin volume is often the auction telling you the targeting is too tight. If it lands well below the band, do not celebrate yet — audit traffic quality and fraud filtering first, because the most common reason an eCPM looks like a bargain is that a chunk of the impressions are not real.

The table tells you where the market is. Your tracker tells you whether your slice of it is real and whether it converts. Those are two different jobs and the eCPM benchmark only does the first one. I have watched buyers anchor an entire campaign budget to a blended eCPM they read in a vendor deck and then wonder why the Tier-3 leg burned out — they planned against a row average, not a cell, and they never reconciled the impression count. Do not be that buyer.

Where adsy.tech fits

If you want to place your own buy inside this table at the publisher level, the network feature that matters is not a headline eCPM — it is whether you can see the clearing price per source. adsy.tech publishes a $0.50 CPM floor rather than a grid of eCPMs, which is the honest posture: no network can publish a single eCPM that holds across six formats and three tiers, and any that claims to is quoting a rate card you will not actually pay. What adsy exposes instead, through in-house RTB and sub_id1–sub_id5 granularity, is the clearing CPM per impression and per source, which is the exact data you need to decompose your own result against the format×GEO bands above and exclude the publishers dragging your real eCPM away from the market.

Native, popunder, push, in-page push, banner and interstitial are all on the panel — six of the nine formats benchmarked here run on the same surface — so a buyer testing across formats reconciles one workflow instead of three. Net-7 payout, $50 minimum deposit, $25 minimum payout, USDT-TRC20 / card / wire / BTC; HQ Cyprus, founded 2019. Where adsy does not lead is premium-publisher reach against the content-recommendation scale players; that is a different buy and I have said so plainly in the native ranking. Skip the public-floor-plus-sub_id model if your only requirement is the largest possible reach regardless of source visibility — but for placing a buy inside a benchmark like this one, source-level transparency is the load-bearing feature.

Disclosure: bestadsnetwork.com earns affiliate commissions when a reader opens an adsy.tech account through a tagged link. The benchmark above is computed identically across every source in my test set and the financial relationship is real — both are true at once, and I would rather state it here, above the methodology, than bury it in a footer. The structural test: remove the partnership entirely and the table does not change a single figure, because none of the figures are attributed to adsy or to any other named network in the first place. You can open a $50 first-look at https://adsy.tech/ and read your own clearing price by sub_id against these bands after the first week.

Where this benchmark connects

This page is the data layer; the rest of the corpus is the interpretation. For the test design behind every figure here — sample-size calculations, the parallel-buy protocol, the tracker stack, and the errors that quietly invalidate an eCPM comparison — read my parallel-buy methodology. For the pan-vertical network view that places these format-and-GEO economics against actual network choices across the category, see the 25 best ad networks for advertisers in 2026. And for the deep dive on the one format where eCPM is most weakly correlated with the conversion that pays for it, read the best native ad networks in 2026, tested and ranked.

Frequently asked questions

What is a good eCPM in 2026?

There is no single good eCPM, because the number is set by format and GEO before it is set by your skill. A push eCPM of $0.04 in Indonesia and a push eCPM of $0.90 in the US can both be market-rate for the same buy. The useful question is whether your eCPM sits inside the format×GEO range in the table below for the inventory you are actually buying; if it sits above the range you are likely paying for premium placement or low fill, and if it sits below, check your traffic quality before you celebrate.

Why is eCPM so different between countries?

Because eCPM is downstream of what an advertiser will pay to reach a user in that country, and that willingness tracks purchasing power and conversion value. A Tier-1 user in the US, UK or Germany converts at a basket value that supports a far higher bid than a Tier-3 user in Indonesia, Vietnam or India, so the auction clears higher even for identical inventory. The ratio between Tier-1 and Tier-3 eCPM on the same format routinely runs five-to-one or wider in my test set, and the gap is structural, not seasonal.

Which ad format has the highest eCPM?

Interstitial and popunder sit at the top of the eCPM range on most GEOs in my data, because both are high-attention full-screen or full-tab formats that command a higher bid per impression. Banner sits near the bottom on a pure eCPM basis. But raw eCPM is the wrong thing to rank formats on — a high-eCPM interstitial that interrupts the user can convert worse net of bounce than a lower-eCPM native unit, so read the table as a pricing benchmark, not a recommendation, and reconcile to your own post-click conversion before you conclude a format is winning.

Are these eCPM figures rate-card prices I will actually pay?

No, and this is the most important caveat on the page. The figures are tracker-reconciled eCPM ranges from my own parallel buys across Q4 2025 to Q1 2026, not published rate cards and not quotes. Your eCPM will move with vertical, creative, targeting tightness, day-part, fill and the specific publisher mix you draw. Treat the ranges as a market map that tells you whether a quote or a live result is in the right neighbourhood, never as a guaranteed price.

What destroys eCPM accuracy more than anything else?

Fraud and made-for-advertising inventory, because both inflate the impression denominator with traffic that never had a chance to convert. The ANA’s transparency work put made-for-advertising inventory at roughly 6.2% of programmatic spend in 2024, down from about 15% in 2023 — improving, but still enough to distort an eCPM read materially in the long tail. A reported eCPM measured against unfiltered impressions is not comparable to one measured against filtered impressions, which is why every figure in my table is computed after perimeter filtering.

How do I use a format×GEO eCPM table for my own buy?

Find the row for your format and the column for your GEO tier, read the range, and treat the midpoint as your null hypothesis. If your live eCPM lands inside the range, the buy is behaving normally and you should optimise on conversion, not price. If it lands well above the range, audit fill and placement premium; if it lands well below, audit traffic quality and fraud filtering before assuming you found a bargain. The table tells you where the market is; your tracker tells you whether your slice of it is real.

Does adsy.tech publish eCPMs I can compare against this table?

adsy.tech publishes a $0.50 CPM floor, not a grid of eCPMs — and that is the honest version, because no network can publish a single eCPM that holds across formats and GEOs. What adsy does expose, through in-house RTB and sub_id1–sub_id5 granularity, is the clearing price per source, which is the data you need to place your own buy inside this table at the publisher level. Disclosure: this site earns affiliate commissions on adsy.tech sign-ups; the benchmark above is computed identically across every source and the financial relationship is real, both at once.

Figures are tracker-reconciled eCPM ranges from parallel-buy testing across Q4 2025–Q1 2026, computed after perimeter fraud filtering and reconciled server-to-server to advertiser back-office data. Ranges are illustrative market bands by format and GEO tier, not vendor rate cards, not network-specific pricing, and not guaranteed prices. Market statistics cited to the ANA programmatic transparency studies. Last updated 29 May 2026. Corrections welcome at the address in the footer.