Google Analytics

Understanding the Impact of Session Calculation Changes in Google Analytics

Posted at August 15, 2011 | By : | Categories : Google Analytics | 4 Comments

On August 11th 2011 Google Analytics announced a change in how a session (i.e. “Visit”) is calculated.  The update was explained by as an “improvement” to how Google Analytics identifies sessions to bring GA “inline with the common definition of a visit.”  Personally, I am skeptical about whether or not this is a good change, however long-term I don’t think it is inherently “bad” – it just changes things, and a change in such a key component as how a session is identified is potentially disastrous from a reporting and data continuity perspective.  Keep reading for more analysis of this change and its impacts.

Reviewing the Change to Session Identification

The Google Analytics blog post about the updatedescribes what was changed:

Currently, Google Analytics ends a session when:

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

If any of these events occur, then the next pageview from the visitor will start a new session. In the new model, Google Analytics will end a session when:

  • 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_sourceutm_mediumutm_termutm_contentutm_idutm_campaign, and gclid.

As before, if any of these events occur, then the next pageview from the user will be the start of a new session.

Analysis of the Change’s Impact on Your Data

This change is one that will have an impact on your data that varies based on your users’ typical behavior patterns and the extent to which you use custom campaign tagging (i.e. utm_source=blah, etc…).

Increased Visits

First of all, the effect will be increased visits since the definition of a “Visit” has become narrower. Previously, if a user on your site came from Google Organic, browsed a few pages, and clicked the Back button, searched again, and later clicked another link, they would be tracked as 1 visit provided they weren’t gone from your site more than 30 minutes. NOW, with this change in effect, they would be counted as 2 visits. I’ve seen some sites that have rather high rates of this type of behavior pattern. Typically they are in industries where users conduct heavy and active research and click to lots of sites during one browsing session, such as Travel. The visit increase effect will be greater for those who use custom campaign tags, especially if you use them on INTERNAL site links. I.e. if you put utm_source, utm_campaign, etc… on links in your own pages TO your own pages, this change will basically nuke your data because every time someone clicks on one of those links they will get a new visit. If you have a lot of tagged marketing out there and users are likely to encounter that, i.e. responed to an email link, view a couple pages, go back to the email, click another link, that’s going to be 2 sessions when it was previously 1.

Impact on Other Metrics

When session definitions change ALL metrics related to sessions will change.  In Google Analytics that’s a lot of stuff.  A few key metrics that are impacted by this change are:

  • Return Visitors/New Visitors – with this change, the same visitors will end up having more sessions/visits, thus the number of visits attributed to “return” visitors will increase and the percentage of visits from “new visitors” will decrease.  You’ll see this on any report that uses the % New Visits metric.
  • Bounce Rate – bounce rates are likely to increase as the number of sessions that end up with only 1 pageview will increase.  If a user searches for “hotels in london” and clicks to your site, looks around the landing page, then clicks back to the SERP and searches for “hotels in chelsea, london” and clicks to your site again and browses around the site, they will now produce 2 visits, the 1st visit having bounced and the second visit not bouncing and being from a return visitor.  Previously this would have been reported as 1 visit that did not bounce.
  • Pages per Visit/Time on Site – these metrics will also change, perhaps drastically depending on your site’s typical user behavior.  As with Bounce Rate, if a user who once was recorded under only 1 session is now recorded under 2 sessions, then your Pages per Visit and Average Time on Site will be cut in half.  Before the change, if a user clicked on a link in your tagged email and browsed 5 pages, then went back to the email and clicked on a different link with a different tag and browsed another 5 pages, they would be reported as 1 visit with 10 pageviews.  Now, they’ll be reported as 2 visits each with 5 pageviews.
  • Conversion Attribution – since sessions are changing and traffic source identification is updated with each session, conversion attribution will change.  If you’re one of the lucky few who are part of the “Multi-Channel Analytics” beta, then this is really good news.  If you’re stuck on the outside with the old last-click attribution model, you’re going to see a swing, possibly a wild swing, in conversion attribution, again based on user behaviors for your particular site.  Long-term, this is going to be a good change, but until Multi-Channel is available for all, I believe it further muddies the already murky waters around conversion attribution.

What To Do Next

If you’re seeing a wild swing in your data starting the 12th of August, 2011 then you’re being affected by this change.  As the blog post stated, the change is unlikely to drastically impact most sites (however at the time of writing this Twitter is a bit on fire over the matter).  If you are seeing those big changes, the first thing you need to do is validate the integrity of your implementation.  Pre-existing problems in your implementation may be magnified by this change to a more noticeable level.  Also, if you are using campaign tracking parameters on internal ads or links that you’re going to drastically impact your data.  Campaign Tags should never be used for internal links - rather, use Custom Variables and/or Event Tracking for this. To check for internal problems give our free Analytics HealthCheck tool a spin, or if you think you need a more heavy duty forensic audit of your data quality, contact us and our crack team of Analytics Pros can assist. Best, -Caleb

Share

Comment

  • Jan

    August 17, 2011 at 3:21 pm

    By the way, one more metric which is being affected is
    conversion rate, due to the fact the number of visits went up. On our client
    site conversion rate dropped after 11th of August by 49.63% compare to previous
    week. So now it seems like a new beginning. Your data prior August 11th, can’t
    be used for comparison anymore. All the data seems pretty much useless now. Any
    idea how to deal with it? What is it you advise your clients, as what was true
    yesterday isn’t today.

    • Caleb Whitmore

      August 17, 2011 at 5:41 pm

      Jan – yes indeed, conversion rate will sink if visits increase, and this is an unfortunate piece of additional “collateral damage” from the update.

      I’m advising my clients to do what should be done anytime core data inputs change and impact other key metrics: explain the change, why it happened, what has changed, and how to live with the new reality.

      Specific to conversion rate, the best you’ll be able to do to create some semblance of sanity for before/after comparison is to wait a couple more weeks until you have more data and calculate the overall delta in visits.  If you’re up 30% on average, then a loose way to adjust historical numbers will be to increase prior visits by 30% and re-calculate conversion rates based on the “adjusted” visit number.  This is NOT going to be a very clean “scientific” way to adjust the data, but you don’t have many other options.  

      One thing to keep in mind: CONVERSIONS shouldn’t have changed, so you can always trend conversions and for historical comparisons, trend the %change in total conversions over time.

      I agree, this is frustrating.  The change has simply created a new reality for GA data… long-term that is probably going to be good, but it’s hard to accept in the moment.

      Best,

      -Caleb

  • Jan

    August 17, 2011 at 3:02 pm

    On our clients sites bounce rates went up between
    15-100%, and number of visits increased on some site by up to 45%. On my own
    site, my home page bounce rate went up from 10% to 20%. I’m bit puzzled now, as
    what ‘good’ or ‘benchmark’ bounce rate going to be. Is the bounce rate going to
    lose a meaning as a metrics?

    • Caleb Whitmore

      August 17, 2011 at 5:35 pm

      Jan,

      Thanks for reading, and posting some feedback.  ”Good” has always been subjective to your site.  I think overall this is going to lower bounce rates across the board from averaging around 50% to maybe 45%.  But, the biggest change will depend on YOUR site and YOUR customer behavior as it relates to browsing patterns.

      On the thought of Bounce Rate, I would argue that overall Bounce Rate has been and still is broken unless you are measuring the entirety of user experience.  If I go to your site, scroll around and read all the content on the landing page, click on things that don’t load a new page like outbound links, downloads, or script objects, and then leave, GA before and now will consider be as having “bounced” when in *reality* I was far from bouncing.  Check out http://www.analyticsengine.net for a tool that will help solve this by making GA track more comprehensive interaction and user behavior.

      Lastly, you have a unique opportunity to learn about user behavior with this change.  Try comparing the days following the update to the same days of the week prior to the update, then look at, say, keywords or landing pages.  Where you see big changes in bounce rate highlights keywords or landing pages where *user behavior* involves entering and leaving but coming back later through some other traffic source.  This is the only time you’ll really be able to see this.

      Hope this helps,

      -Caleb

Leave a Comment