Traffic
Time on page
Also called dwell time.
Time on page is the elapsed time from one pageview to the next hit in the same session. It is a gap, not a stopwatch.
How it is measured
Subtract the pageview's timestamp from the one that follows it. The last page of a session has no successor, so tools drop it or score it at zero, which makes the figure biased toward people who kept clicking.
Tarsier sends a ping every 15 seconds while the tab is visible and one when the page is hidden, which gives even the last page a value. It still describes visible seconds on that page, not reading.
Worked example
A tutorial page stays open for 12 minutes while the reader switches to a terminal and runs its commands. The next hit arrives 12 minutes later, so the report says 12 minutes. The reader actually spent about 90 seconds on the page itself.
The author reads this as proof that the tutorial grips people. A scroll-depth check shows 70 percent never got below the first code block.
How it differs
Time on page is gap-based. Engagement time counts only visible seconds. Average session duration spans the whole visit. Time on page is the right tool for relative comparison between two similar pages and the wrong one for claiming attention.
Common errors
Believing the last page has a value. Reading a mean pulled up by a few abandoned tabs. Equating a long gap with reading. Comparing a short post to a long guide. Forgetting that a tab left open skews everything.
In practice
Compare pages of the same type with medians. When the question is whether anyone actually read the page, move to engagement time or scroll depth.