During recent testing we saw an abnormal increase in direct visits around the same time as the launch of one of our A/B tests using. Drilling into the landing page dimension we could see that most of these new direct visits were going to the variant page. This threw a HUGE question up as to the impact of the Website optimizer redirect that allocates visits to the original and variant page.
The redirect causes the source cookie information to be overwritten as a direct visit, (obviously because the GA tracking code has not had a chance to fire on the original page.)
The simple solution is to move the Google Analytics tracking code ABOVE the Google Website Optimizer control script. Problem solved! Obviously this is not best practice as far as GA code placement is concerned but the negative impact on data quality justifies this for the period of time that tests are being carried out.
An extra tip is to add a ‘no robots’, ‘no follow’ tag to the variant page(s). Google does state that indexing should not be affected, but better to be safe than sorry!