Web Analytics: More on Time on Site Calculations

Following Avinash’s great post I’ve already mentioned yesterday, I re-read the entire post and more than 3o interesting comments.

The main problem as Avinash explains, is that we can’t actually calculate the time on page and time on site where we don’t have an “exit” mark.

This basically means, that most of our “bounces”, “zero” time on site and “short visits” (depending on your software verbiage) are related not only to those who close their browser right after entering your page, but to those who viewed 1 page, perhaps even for a while – but didn’t go any further.

Well I say delete them!

The problem with zero time on site

The zero time on site is screwing up the entire time on site statistics of your pages.
Here’s an example for a website we manage where the average time on site is 91 seconds.

Average Time on Site - All VisitorsAverage Time on Site - All Visitors

Looks fine to the untrained eye: 91 seconds and 2.2 pages per visit – So is this the correct number?

Time on Site – the accurate numbers

To examine the correct way to calculate, I’ve created to segment using ClickTracks Analytics for users who spent a maximum of 2 seconds on the website AND for people who spend a minimum of 3 seconds on the website.
Here’s what I’ve got now:

Average Time On Site - CorrectedAmazing!
The users who spends 3 seconds and more, actually spent 354 seconds and viewed 5.6 pages from my website – this is almost 4 times more than the original number!!!

It means that real users, who do not close their browser window, go to the next tab or simply stay only on one page – spend a much larger portion of time on the web site, in average. This means that the engagement we’re getting from them is so much better than we thought 10 minutes ago!

Any thoughts?

3 Responses so far.

  1. Wow, that’s a big difference! But note also that whether a person visited just one page or many, their time will always be underestimated because you can’t see the time on the last page. So the people who just visited one page didn’t really spend 0 seconds there, but equally the people who visited more than one page didn’t really spend an average of 354 seconds, but 354 seconds plus whatever time they spent on the last page.

  2. Ophir Cohen says:

    Hi Stephen,
    Thanks for the comment. You are soooo right. I think I asked you last time we met what would happen if we triggered a sort of a “refresh” dunction which would enable to count the last page as well and you convinced me that it would most probably just bombard the servers… :-) However in Avinash’s post the discussed an “onunload” script that may do the trick. I wonder, why not do it?

  3. [...] the latest posts on time on site on this blog and on Occams Razon by Avinash Kaushik, I received a few questions and I also saw a [...]

Leave a Reply


*