Average time on site and average pageviews revisited

I want to take a closer look at two common metrics: average pageviews and average time on site. What I want to find out if these average metrics adequately describe the underlying data.

To start, lets work wth non-bounce visits. It’s a fact of life that most sites have a ton of bounces, i.e. visits with one pageview and 0 time on site. I just want to look at visitors who have shown a little bit of engagement.

Here is the overview data:

I have 2 data points, 9.36 average pageviews and 7min 13secs time on site. But we don’t know anything about how this data is distributed. Does this mean that most visits had just over 9 pageviews? Or maybe one visit had 100 pageviews and 20 others only had 5 pageviews?

Fortunately, we do have a couple of reports that give us a distribution for time on site and pageviews per visit, called Length of Visit and **Depth of Visit, **respectively.

1) Length of Visit

This is really nice. The data almost looks like a normal distribution. What this tells me is that indeed most people spend 3 to 10 minutes on the site. “Most” is called the mode in statistics. The mean or average is also somewhere in there at 7mins and 12secs. We can also say that the median time on site is in that time bracket. The median is the mid point of your data, so 50% of the data is less than the median and 50% is greater than the median. In this case, the median is probably closer to 3 minutes than to 10 minutes, but it’s in this band.

Bottom line is that using the mean or average is a good statistic for time on site.

2) Depth of Visit

Wow, this distribution looks completely different than Length of Visit. The average is 9.36, but most visits in fact have just 2 pageviews. Also, the median is 5 pageviews. So, 50% of visits have fewer than 5 pageviews and 50% have more than 5 pageviews.

In fact, this looks more like a power law to me, aka long tail distribution.

Since you can now set engagement goals in Google Analytics, which goal value should you choose for pages/visit? Or, would you?

Something to think about, but I will suggest an approach in an upcoming post.