This is where interaction events come in. As mentioned in the original post (and on Google Answers for impact of interaction events on session duration), if a page sends an interaction event back to the GA servers (in GTM parlance, Non-Interaction Hit set as False), it becomes possible to capture time even on exit pages. Each such event acts as a fresh timestamp, helping GA measure the difference between the last event and the time stamp at page load.
A simple way to understand this is by looking at your data. Here’s a custom report to look at exit pages and their Time on Page values. Notice that this uses the ga:timeOnPage metric, which is not available in the default reports in GA.
Here’s the URL of this shared custom report that you can directly apply to your data.
To look at only the exit pages, I applied the following filter on the data:

Here are the results for a week’s data on a client’s website:

All of these five pages had a 100% exit rate over this period. It’s not surprising to see the Avg. Time on Page (1) as 0 for all of them. However, three out of these five have non-zero Time on Page values (2). And the column marked 3 gives us the modified value of Avg. Time on Page as described in the original post on this topic.
Here are the key findings:
- Exit pages can record a Time on Page.
- Use interaction events to help Analytics receive information about time spent on the page.
- Consider a timer event to periodically fire these interaction events.
- Implement the modified Avg. Time on Page as a calculated metric to extract a more accurate value.