Thursday, December 12, 2013

Analytics is undercounting engagement of your users

Versión en español aquí.

The two most important traffic measurements for news entrepreneurs are NOT unique visitors and page views. Those numbers can mislead you. They count people who arrive at your website by accident or search, glance at a page and leave.

As Ken Doctor has so eloquently put it:
"Unique visitors are a great dumb count. As I’ve noted, it’s as if in the print world we counted the everyday subscriber — consuming 5 hours a month of a news publication — the same as someone who, standing on a Midtown corner on a windy day, happened to catch a sheet of flying newsprint as she held up her hand to hail a cab."
By contrast, the two measures that should really matter to you are:

  • engagement -- how long a visitor is on your site per visit and how many pages they view
  • loyalty -- how many times they return per day, week or month
Most visitors to your site will be fly-bys. This is true even for the best-known news sites. The Pew Research Center’s Project for Excellence in Journalism studied the top 25 news sites and found that 77 percent of their traffic came from users who visited just one or two times a month. Only 7 percent were so-called power users who visited more than 10 times a month. 

Users with high engagement and high loyalty (power users) are important to identify because they are the ones who are most likely to pay for premium services, buy tickets to your events, patronize your advertisers and recommend you to others. These are your real customers, not the fly-by visitors. 

A troubling discovery

So I was surprised to realize recently that Google Analytics undercounts reader engagement because of the way it calculates the bounce rate -- more specifically, the length of a bounce visit. (A bounce is a visit in which the visitor sees one page and leaves without taking any other action on the page.)

Here's how it works: Let's say you tweet out a link to your site or post the link on Facebook. Someone clicks on the link, comes to your site, looks at the recommended page and doesn't click on anything or take any other action. Even if they read the entire page, Google counts the time spent as zero

You don't believe me? Let me quote Google
"Since bounced visits only consist of one interaction, Google Analytics does not have a second interaction to use for the calculation of visit duration or time on page. These visits, and the one page view included in the visit, are assigned a visit duration and time on page of zero."

Wow. I can see now why so many of the visits to my blogs are measured by Google Analytics as lasting only 0-10 seconds on the engagement report. 

Consider the example below of a Google Analytics report on engagement for a cultural website. This site specializes in long-form journalism where a visitor could easily click on a long article, read it in its entirety and leave without taking any other action. That visit would be counted as a bounce, and the visit's duration would be measured as lasting for 0 seconds.
Click on graphic to see larger image.

The graphic above shows that 79% of the visits to the site are measured as lasting for 0 to 10 seconds. How many of those were "bounce visits" that lasted longer? We don't know. But surely some were. So the true level of user engagement has to be higher than what is shown in the Analytics report. 

Now I understand why, when I dig down into the details of my traffic reports to look at very small breakdowns of visitors (say, readers of one particular article from one country on one particular day) the total time spent by all four or five visitors might be zero.

Still useful

What this means for you is that the engagement report's first visit-duration segment -- visits of 0 to 10 seconds -- is not very helpful to you. However, the other duration segments of the report are still useful. They provide some indication of the percentage of your users who are spending significant time per visit on your content. You can track the change in those percentages over time to see if you're getting better at holding users' attention.

As a blogger, I am consoled to know that some of those visits in the 0-10 seconds category might have actually lasted longer. How much longer? I don't know. So I will focus on the other duration segments of that report and work on increasing user engagement in those.

If you want to go deeper into this topic of how Analytics calculates time, check out this blog entry of Google's Justin Cutroni. 



  1. Very interesting. Can I use this information for my feasibility paper for my Business Management class? My topic is about company websites, how it can be used to turn page visits to actual sales. I'll be sure to insert your website as reference.

  2. Sure. Keep in mind that new technologies, like that used by news application Circa, are trying to deal with this quirk in Analytics.