This article is more than 1 year old
Google squashes Google Analytics bug
Traffic report snafu nipped in bud
Google has released a fix for Google Analytics that corrects a traffic reporting bug introduced by an update to the service last week.
On August 11, Google changed the way Analytics calculates user sessions on websites. But the change also added a bug that misrepresented traffic in some situations.
Since the update arrived, UK-based webmaster Neil Mordecai noticed that his site was vastly under-reporting the number of new users and over-reporting the number of returning users. "Google Analytics have released a buggy version of their system. Since last week...they've been counting on two hands," he told The Register on Wednesday.
"Before the 11th August 'first visit' traffic was typically 80% of the total traffic with 'same day returns' being less 10% - this pattern extends long into the past. After 11 August, 'first visit' traffic dropped well below 50% and 'same day returns' rose to well over 50%."
When we contacted Google about the apparent bug, a company spokeswoman said she would look into the matter, and later in the day, Google updated the blog that announced last week's change, saying that it had found and fixed a bug in the update. The fix, Google said, was issued on Tuesday.
After the blog post was updated, the company spokeswoman alerted us to the update, and she later reiterated that Google engineers were aware of the bug before we notified the company.
Google's description of the bug is similar to what Mordecai observed. The company indicated that in some cases, Analytics was incorrectly reporting fewer visits from new users and more visits from returning visitors on sites using multiple trackers, which it said was an unsupported configuration. The company also pointed to a traffic reporting error that occurred in certain other situations. The details are here.
The original Analytics update is still in place. In essence, Analytics now decides a user session is over when any traffic source value for the user changes. Previously, user sessions ended only when more than 30 minutes elapsed between a user's pageviews, the day ended, or the user closed his browser. ®