Those who’ve used Google Analytics or similar platforms will be familiar with concepts like Session Duration, Watch Time, and Exit Page. This blog post focuses on Session Duration.
The truth – sadly – is that the descriptive statistics provided by most SEO and diagnostic systems are enormously skewed. This isn’t the “fault” of the creators of these tools; they, understandably, want to create a nice gradient between people who visit the site briefly and those who absorb more content. The truth, meanwhile, is that for most pages on your site, this distribution is probably brutally bimodal: some people leave right away, others (usually fewer) remain for 100x as long.
The accuracy of this information is problematic before it is ever interpreted by these systems, however. Its initial mode of reporting from Google or others is vulnerable to certain “biases.” The two things marketers and others are most likely to examine are among the most vulnerable.
Session time is a strange statistic in that, at first glance, it should itself be enormously indicative of – perhaps even dispositive of – user preferences. But in an age of multitasking and tabbed browsing, computers look at things even when we don’t. So how do we know which session times are “real” and which one are overly-optimistic situations where some unseen tab spent an hour on your page? And it gets worse – the session keeps “running” on that page until the user “leaves.”
Let’s start with things that are unlikely. It’s unlikely that someone read the webpage, watched the video, or listened to the song if that person was on the page for less than ten seconds. Let’s also say it’s unlikely any significant number of people visited a particular page on your site for exactly thirty minutes (this, or 1,800 arbitrary units is the default timeout in Google Analytics). It’s even less likely that a user spent hours or more on a single page, no matter how interesting that page’s content.
But somebody between ten seconds and half an hour is paying attention. That’s what lies between.
And that user requires more nuance to analyse. As early as 2007 – with patents filed as early as 2010 – Haystack began analysing time on page relative to amount of content on that page. With audio and video files (media files), this is particularly important and what we call the consumption ratio. To quote our applicable patent, this involves “comparing the length of time spent by the user viewing the video file against a total playing length of the video file to determine a viewing ratio.” (U.S. Pat. 9,594,809).
This is a rare area where media files are actually easier to deal with than text (this is not, for reasons we’ll discuss in a future blog post, true in most search contexts). This is because some people are faster readers than others, or people able to skip ahead on a page and find the nugget of information they need, but it takes everyone exactly the same amount of time to listen to a three-minute song or watch a five-minute video. And this is why consumption ratio – our statistic – is both more accurate and more actionable when analysing user attention to content than Google’s Session Duration, Watch Time, or Time on Page.