Revenue per session is total revenue for a period divided by the number of visits recorded over the same period. Expressed per thousand visits it is normally called session RPM, and the sum is trivial: revenue divided by sessions, multiplied by 1,000. A site earning 500 dollars across 60,000 sessions has a session RPM of 8.33 dollars, to use the worked example published by Search Engine Journal. The metric exists because the numbers publishers traditionally optimised measure an advertising slot rather than a reader. A page can be loaded with units until the yield per page climbs, and it will climb; what that does to the visit is invisible at that resolution.

Two numerators are in circulation. Publishers count monetisation revenue, mostly programmatic and direct advertising, increasingly subscriptions, affiliate commissions and commerce. Retailers count transaction revenue, which turns the same ratio into a conversion metric. The denominator is identical in both cases, and that is where most of the trouble lives.

How the number is assembled

The formula hides an integration problem. Revenue sits in the ad server or the payment system; sessions sit in the analytics platform. Neither knows about the other, so the datasets have to be stitched, usually by date and property, sometimes by landing page or traffic source. Clickio sells that join as a product, combining its own data with AdSense and Google Ad Manager earnings alongside Google Analytics session counts.

What counts as a session is set by the analytics tool, not the publisher. In Google Analytics 4 a session begins when a user views a page or opens an app with no session already running, firing a session_start event that generates a session identifier (ga_session_id) and a session number (ga_session_number). It times out after 30 minutes of inactivity by default, adjustable between five minutes and seven hours 55 minutes.

Ad servers ran a parallel version. Google's AdSense glossary defines an ad session as a visit during which one or more pages carrying ads are viewed within a certain period, and ad session RPM as estimated earnings divided by ad sessions, multiplied by 1,000. Coverage was partial: a companion metric, ad sessions measurable, reports the share of AdSense for content traffic the system observed. Those pages now carry a notice that session-related metrics are being retired from September 2025.

The useful property is that it decomposes. Session revenue is approximately page RPM multiplied by pages per session, which turns any monetisation change into an arithmetic trade. Blockthrough set out the case cleanly: a publisher with a page RPM of 6 dollars adds units on its highest-traffic pages and lifts page RPM to 8 dollars, while average pages per session fall from four to two. Page RPM is up a third, session RPM has dropped from 24 dollars to 16 dollars, and revenue is down 33%. That asymmetry is the argument for the metric.

Ezoic markets the same idea under EPMV, short for earnings per thousand visitors: total earnings divided by visitors in thousands, so 6,500 dollars from one million visitors gives 6.50 dollars. Its documentation calls page-level RPM a false positive, inflatable by adding ads while overall earnings fall.

Where the metric came from

The denominator predates the metric by two decades and rests on a citation most of the industry has never checked. The 30-minute timeout traces to a 1995 paper by Lara Catledge and James Pitkow, who measured a mean gap of 9.3 minutes between user events, added 1.5 standard deviations to reach 25.5 minutes and proposed that as a boundary. A later Wikimedia Foundation paper noted the threshold had since been smoothed to 30 minutes.

Session-denominated revenue reporting arrived much later. Universal Analytics shipped Per Session Value, transaction revenue divided by sessions, giving e-commerce operators the figure years before publishers had one. Advertising-side adoption followed in the second half of the 2010s, pushed hardest by Ezoic, whose case appeared in The Drum on February 13, 2018 under the claim that publishing was finally adopting EPMV, or session revenue.

Universal Analytics also restarted a session at midnight and whenever a visitor arrived through a new campaign source. Google Analytics 4 does neither, so historic and current session counts are not the same quantity even for identical traffic.

The most recent move came from the supply side. On June 11, 2026 Teads launched EngageOS, a feed operating system that merges editorial recommendations and advertising demand into one real-time auction and optimises for revenue per session rather than revenue per impression. Where a conventional stack treats an internal link and an ad slot as separate inventory, the orchestration engine forecasts both and awards the placement to whichever carries the higher predicted session revenue. Magnite is the launch demand partner, connected through Magnite Demand Server to Prebid Server-compatible supply-side platforms.

Why publishers reached for it

Session yield became urgent as the denominator started shrinking. Teads told investors on November 6, 2025 that publishers across its network had lost 10% to 15% of pageviews in the third quarter, blaming AI summaries and changes in discovery, and that higher revenue per thousand impressions had not made up the difference. Chartbeat data reported in April 2026 showed small publishers losing 60% of search referral traffic over two years, with medium-sized publishers down 47%. Index Exchange recorded year-over-year declines in ad opportunities across 69% of publishers on its platform through 2025, averaging 14%.

An impression-denominated metric cannot see that problem and can obscure it. During the January 2026 Google serving failures, publishers reported impression RPM falling from roughly 7.25 dollars to 4.00 dollars while coverage dropped from 89% to 79%, a slot-level view of an outage whose commercial size resolves only at the visit level. Alphabet's network advertising revenues across AdSense, AdMob and Google Ad Manager fell 1% to 7.4 billion dollars in the second quarter of 2025.

Vendors now sell yield per visitor rather than traffic growth. Ezoic reported uplift rising across four measurement windows to 11.3% in tests spanning more than 900 sites between January and July 2026, and a 23.6% increase in network EPMV in the first quarter of 2026. Both figures are company-reported and unaudited.

For buyers the metric prices traffic sources. Academic work covered in October 2025 used revenue per session as a headline measure in a study of 973 e-commerce sites with 20 billion dollars in combined revenue, which found ChatGPT referrals underperforming Google's paid and organic channels on conversion rate and revenue per session. Vendor findings put AI-referred conversion anywhere from below parity to eleven times search, and the spread across published benchmarks makes any single figure close to useless for one site.

Limitations and disputes

The metric is not standardised, not accredited and not comparable between companies. Networks calculate it against different denominators, and Gourmet Ads has argued that session-denominated reporting is chosen precisely because sessions are fewer than pageviews, producing a larger headline rate without any change in performance, which it calls dishonest reporting and an obstacle to comparing networks.

The denominator itself is a convention. Ringside Data published an analysis on August 7, 2026 testing the 30-minute rule against twelve months of event data across leisure and travel, direct-to-consumer e-commerce and business technology. Around 95% of gaps between consecutive events fell under two minutes in the consumer segments and 73% in the business segment, with the quietest part of the distribution sitting between 15 and 20 minutes everywhere. Moving the cutoff between 15 and 60 minutes shifted session counts by low single digits; dropping to five minutes inflated them by 10% to 17%. Re-running the 1995 method on modern data produced six different answers depending on where the tail was truncated.

Measurement coverage is the other soft spot. Consent refusals, ad blocking, cookie deletion and cross-device behaviour suppress session counts while revenue keeps arriving, inflating the ratio; AdSense acknowledged as much by shipping a measurability metric beside its session figures. The ratio can also improve for reasons nobody would celebrate, since shedding low-value traffic lifts revenue per session while total revenue falls.

Not the same as

Page RPM divides earnings by pageviews rather than visits. It answers a question about a page, and its blind spot is the reason revenue per session exists.

eCPM and CPM are priced against impressions. CPM is what a buyer pays for a thousand impressions; effective CPM is what a seller realises across mixed pricing models. Neither sees what a visitor does next.

EPMV is Ezoic's name for the same construct, calculated against visitors rather than sessions. Session RPM, EPMV and session revenue are used interchangeably in practice, which is reason enough to check the denominator before comparing two numbers.

ARPU divides revenue by users over a period, usually a month. A user generates several sessions, so it answers a retention question rather than a yield question.

Recent developments

EngageOS is in testing with Penske Media, The Arena Group and Scripps, and Teads chief executive David Kostman described significant yield lifts during those tests on the second-quarter call at which the company withdrew its 2026 guidance. The thesis carries a dependency. In the complaint filed against Google on August 3, 2026, Teads stated that prospective publisher clients had told it EngageOS was not a viable full solution without access to AdX or Google Ad Manager, and warned that Google could cut its revenue in retaliation.

Google has moved the other way. Retiring AdSense session metrics from September 2025 removed the only session-denominated revenue figure a publisher could read straight out of a major Google monetisation product, leaving the calculation to third parties and manual joins at the moment the industry has most use for it.

Timeline

  • 1995: Catledge and Pitkow measure a 9.3-minute mean gap between user events and propose 25.5 minutes as a session boundary
  • 2005 onwards: Google Analytics adopts a 30-minute default session timeout, later carried into Universal Analytics with Per Session Value
  • February 13, 2018: The Drum publishes the case for EPMV, or session revenue, as the publishing industry's primary monetisation metric
  • 2020 to 2022: Session RPM spreads across publisher ad tech vendors, with competing claims about its honesty
  • September 2025: Google begins retiring session-related metrics from AdSense reporting
  • November 6, 2025: Teads reports 10% to 15% pageview declines across its publisher network and says higher RPM did not compensate
  • October 2025: Research on 973 e-commerce sites uses revenue per session to compare ChatGPT referrals with established channels
  • April 2026: Chartbeat data shows small publishers down 60% in search referrals over two years
  • June 11, 2026: Teads launches EngageOS, built to optimise predicted revenue per session, with Magnite as demand partner
  • August 3, 2026: Teads sues Google and discloses that EngageOS is not considered viable by prospective clients without AdX access
  • August 7, 2026: Ringside Data publishes an empirical test of the 30-minute session rule across twelve months of event data

Summary

Who. Publishers and their yield teams calculate it; publisher ad tech vendors including Ezoic, Clickio and Teads promote it; e-commerce and conversion teams use the transaction-revenue version; Google reports the underlying session counts in Analytics and is withdrawing its own session metrics from AdSense.

What. Total revenue divided by visits over the same period, reported per thousand sessions as session RPM, and decomposing approximately into page RPM multiplied by pages per session.

When. The 30-minute session convention dates from 1995 research; Universal Analytics carried a per-session revenue metric; advertising-side adoption ran from about 2016, with Teads making predicted session revenue an auction input in June 2026.

Where. Web and app publishing, principally in open-web display and native monetisation, and in e-commerce analytics where the numerator is transaction revenue.

Why. Impression and page metrics reward ad density and cannot detect the engagement that density destroys. With pageviews falling across the open web, the value extracted from each visit has become the number that decides whether a publishing business works.