How to Measure Time on Page in PostHog

PostHog doesn't give you a single, clean "time on page" number — and neither does any analytics tool, honestly, because true per-page time depends on the gap to the next pageview, which the last page of a visit never has. The practical measures are session duration — the spread between a session's first and last event, computable in HogQL with dateDiff — and PostHog's native web analytics. Both are honest proxies for engagement.

Why "time on page" is slippery

To know how long someone spent on page A, you need the timestamp of their move to page B. For the last page they view — which includes every bounced single-page session — there's no "next" event, so the time is unknown. Tools paper over this with estimates. Rather than trust a fuzzy per-page number, measure session duration (how long the whole visit lasted) and lean on scroll depth for how engaged a single page was.

Compute session duration in HogQL

Group events into sessions, take the spread between the first and last event with dateDiff, and average it:

SELECT
  round(avg(duration_seconds), 1) AS avg_session_seconds,
  count() AS sessions
FROM (
  SELECT
    properties.$session_id AS session,
    dateDiff('second', min(timestamp), max(timestamp)) AS duration_seconds
  FROM events
  WHERE timestamp > now() - INTERVAL 30 DAY
    AND properties.$session_id != ''
  GROUP BY session
)

Illustrative sample output:

avg_session_seconds	sessions
96.4	8,120

This is honest: it's the average spread of activity within a session, not a per-page figure. On a landing-page-heavy site where most sessions are a single page, it's a decent stand-in for time on that page. Single-event sessions naturally show a duration of zero, which keeps bounces from inflating the average.

The native path: PostHog web analytics

If you don't want to write HogQL, PostHog's Web Analytics section reports session duration and engagement out of the box, and session replays let you watch how long real users actually linger and where they stall. Use web analytics for the aggregate number and replays to understand why it's high or low.

A better single-page engagement signal

For one page, scroll depth usually beats time. PostHog records the max scroll of a pageview on the $pageleave event, so you can ask "did people actually read this?" directly:

SELECT
  round(avg(properties.$prev_pageview_max_scroll_percentage) * 100, 1) AS avg_max_scroll_pct,
  count() AS pageleaves
FROM events
WHERE event = '$pageleave'
  AND properties.$pathname = '/'
  AND timestamp > now() - INTERVAL 30 DAY

A long time on page with shallow scroll often means confusion, not engagement — which is exactly the nuance a raw time number hides.

Use it as a proxy, not a target

Don't optimise for time on page directly; a confused visitor stays a long time too. Read session duration and scroll depth together, against conversion, to judge whether engagement is healthy. If you'd like those signals turned into a weekly conversion fix shipped as a Pull Request, that's what Velyr does.