For publishers
A pageview count does not say which piece held attention past the first scroll, or whether the person who read it last week came back for this one. Media and content sites usually find out from the comments, not the data.
Start freeArticle performance is one flat number
A pageview count and an average time on page do not say whether anyone read past the first paragraph, or which piece is actually working.
Your busiest traffic is the hardest to see
Content sites carry more ad-block usage than most categories, and a lot of analytics scripts sit on the same blocklists as the adverts next to the article.
Nobody can say whether a reader came back
Cookie-based tools need a consent banner to answer it, and most cookieless tools reset every visitor at midnight either way.
Per-page engagement and scroll depth, tracked automatically
Every pageview records how long the tab was actually visible and how far down the page the reader got. The Optimise screen ranks pages by the biggest week-over-week fall in either number, so a piece losing readers partway through shows up on its own.
Channels classified without a cookie
Referrer and UTM parameters resolve into organic search, paid search, social, referral, email, affiliate or direct, with the source of a paid click id recorded and the id value itself never stored.
A first-party proxy, so fewer readers get blocked
Documented proxy setups for Cloudflare Workers, Nginx, a Next.js rewrite or a Vercel rewrite put the collection request on your own domain, which is what most ad-block lists match on. It reduces the loss; it does not remove it.
Segments on your own properties
Send category, author or content type as a property on the pageview and filter every report by it, the same mechanism an agency or shop uses for its own properties.
Three things this page implies are not quite what is built. There is no 'finished reading' metric: scroll depth is captured automatically on every pageview and is exactly what feeds the engagement-drop panel above, but completion is not inferred from it. Want that, send your own event at whatever scroll or time threshold you define. Who comes back depends on the collection mode, and cookieless, the default, rotates its visitor hash every day by design, so a reader who returns tomorrow looks new rather than returning, and the Retention screen says so rather than drawing a chart it cannot support. Cookie mode is meant to fix that with a durable per-visitor id, but the SDK does not yet set or read that cookie automatically, so today the only thing that actually reconnects a reader across days is your own identify() call, and whether that needs a consent banner is a question for your counsel, not one a toggle answers. Ad-blocker loss is reduced by the first-party proxy option, never eliminated: nothing here measures what a blocker stops before it reaches the server.
It is on the changelog with everything else we have not done. A roadmap item described as a feature is how software gets bought and then resented.
Two minutes to install, no card, and it runs alongside whatever you use today.
Start free