Google Revisits its Visit Definition: What is the Impact on Your Data?

by Lawrence on Tuesday, 16th August, 2011

The visits metric is central to web analytics. However, Google has recently made a change to how it defines a visit within its Google Analytics package.  Previously Google deemed a session ended when one of the following conditions was met:

• More than 30 minutes have elapsed between pageviews for a single visitor.
• At the end of a day.
• When a visitor closes their browser.

Now this has changed, so that a session now finishes when either:

• More than 30 minutes have elapsed between pageviews for a single visitor.
• At the end of a day.
• When any traffic source value for the user changes. Traffic source information includes:utm_source, utm_medium, utm_term, utm_content, utm_id, utm_campaign, and gclid.

The change took place last Thursday (11^th Aug), with the understanding that none of the data held in your account prior to that date will be changed retrospectively – only data from this date forwards will be affected.

By using a changing traffic source to denote a new visit, this will likely increase the reported number of visits received by a website, without making any change to non-visit based metrics, such as page views.  It will be more likely to affect websites or visitor segments where visitors are prone to revisit the site.  This means when making any comparison between visits or visit-based metrics (such as average time on site or bounce rate)before and after the change you won’t be comparing like with like, and as such need to exercise extreme caution.

We’ve carried out an analysis of the data for some of our clients to determine what the effect of the changes is, and the results have been very interesting.  For some there’s been little change in their metrics, whilst others have seen an increase in visits.

 The client above exhibits the changes that you’d expect from the Google visits update, with an increase in visits after the change, but no jump in non-visit based metrics, creating a knock-on effect for metrics which are based on visits, such as bounce rates and page views per visit.  In fact for this client the impact of the change is massive and far higher than Google’s expected circa 1% change!

So, aside from having to explain to stakeholders why a jump in bounce rate or fall in page views per visit is caused by a definition change rather than visitor behaviour, what other issues does this change cause?  As the main change to the definition of a visit is based on traffic sources, those sites where a visitor might use more than one source in quick succession to visit a site will see the largest changes.  This will likely affect sites with high competition as visitors jump between sites, refine their search terms or look for money-saving sites to lower their costs.

Users of Google Analytics may wish to review their campaign tracking methodology in light of these changes to ensure it accurately reflects the attributes of the campaigns pushing traffic to their site.  Furthermore, it’s now more advisable than ever to avoid using the Google Analytics campaign tracking parameters to monitor internal campaigns (as used by some people) and use event tracking instead.

Above all though, the key thing is not to panic.  This is a one-off change to a definition of a metric, so provided you record this within the annotations section of your account and provide the necessary caveats to any reporting you provide, there should be no issues.  It’s rather like when a business changes accounting procedures; they have the same customers who spend the same amount, but certain figures are recorded in a different way.

Is there a better way?

A few blogs on this topic have questioned why Google didn’t keep the old visits metric and introduce a new one in order that historical data could be viewed without a step-change to a clients’ data.  This would make sense for many clients.  In fact, approx 11 years ago RedEye had a neat solution where it reported on visits and click-throughs to provide a work-around for this issue.

Visits worked in the same way as GA used to work and the click-through would be incremented each time a media placement was clicked on.  So if I arrived at the site 10 times in a visit we would report 10 click-throughs against the appropriate media placements and 1 visit.