We recently completed a complex Google Analytics installation for an ecommerce site. The new site launched with a campaign to drive traffic through a variety of tagged banner ads from an established site. At launch, direct traffic implausibly spiked and equaled banner traffic. Skip to the last paragraph to see the solution, or follow our investigative process in detail, below.
We had to first address whether the direct traffic could, in fact, be ‘real’. We doubted the possibility, given the lack of any concurrent marketing or PR initiative to promote the site. The banner ad campaign represented the only marketing effort in place. The direct traffic also showed no qualitative difference (lower bounce rate, longer time on site, higher conversion) that we would normally expect for direct traffic compared to marketing campaign traffic.
Next, we looked for missing or incomplete campaign tags in the banner ad links. All were tagged correctly. The implementation involved a series of redirects (an ad server on the source site as well as a vanity URL on the destination site) which can be notorious for dropping campaign tags. In this case, however, all redirects functioned correctly and passed all campaign query string parameters to the destination URL.
Lastly, we checked the URL rewrite implemented to translate the indecipherable URLs generated by the content management into more logical page names for Google Analytics. Again, we verified that _trackPageview passed the full translated URL with all campaign query parameters intact.
At this point – stumped – we turned to the data, to find a clue about this ‘direct’ traffic. We discovered that the ‘direct’ traffic was exclusively from Internet Explorer, while the campaign traffic was dominated by Firefox, Safari and Chrome. We immediately QA’d with Internet Explorer and discovered that even though all of the campaign parameters remained intact on the landing page, IE did not pass the UTMZ cookie data which contains all of the campaign variables.
In Firefox, Google Analytics passed this:
Data from a similar click in IE looked like this (no utmz, only the visitor and session information in the utma cookie):
At this point, we turned our attention to a customization (_setCampaignCookieTimout) we added to shorten the life span of the campaign cookie from 6 months to 2 weeks for campaign attribution, on the theory that it must be incompatible with the leading versions of Internet Explorer. After removing the line of custom code, banner traffic jumped to 98% of traffic, as anticipated.We QA'd in IE and could then see the campaign variables passing to Google Analtics: