Estimating Google Organic Search Visits Hidden by iOS 6

iOS 6 Safari Search Bar

As we pointed out earlier this quarter, the vast majority of Google search traffic from devices running iOS 6 is being passed without a referrer as a result of Apple's move to default its Safari search box to Google's Secure SSL search. Without the referrer, analytics packages are unable to correctly attribute organic search traffic as such, and the traffic is instead seen as direct.  This presents a significant challenge for SEOs looking to accurately measure and assess the efficacy of their efforts over time. With mobile generating a 25% or higher share of search visits for many sites, and iOS making up the majority of that, organic traffic comparisons from year to year, or even month to month, have already been thrown off considerably by iOS 6. Unless Apple changes its default search option or Google employs a workaround to deliver a referrer scrubbed of the search query -- as it does for secure desktop searches -- the extent to which organic search visits are "hidden" by iOS 6 will only grow as its adoption rate increases. This leaves SEOs little recourse but to work to provide estimates of the true volume of Google organic search visits from iOS 6, and in turn, the full picture for the organic channel.  Here are some tips to do just that.

Determining The Percentage of Google iOS 6 Traffic Without a Referrer

How we go about estimating organic iOS 6 traffic will depend on the information we have at our disposal.  We'll outline two common scenarios below, both of which will first require knowing the approximate percentage of Google iOS 6 organic search traffic that does not have a referrer.  Again, this is the segment that will be seen as a direct visit by analytics packages, even if it came through Google organic search. To get this number, we can use our results from paid search as a close proxy.  Most sophisticated paid search programs employ a redirect for tracking purposes.  Upon an ad click, this redirect can capture the browser's user-agent, which tells us about the device and its OS, and parameters in the ad URL that indicate the source of the click, e.g. Google.  We can therefore determine what percentage of the time an iOS 6 user from Google paid search did not pass a referrer. Our original post found this number to be 82% for an aggregate sample of sites.  In a follow-up analysis we found the median result to be similar at 79% with a mean absolute deviation of 7%.  If you or your paid search manager cannot determine the value for your site, you could employ 80% as a reasonable stand-in.

Case 1:  We Can See Google Organic Traffic by OS Version (Google Analytics/RKG Attribution)

For sites employing analytics packages that can deliver results specific to the operating system version -- iOS 5 vs iOS 6, and so on -- the calculations to estimate the number of visits improperly attributed to direct due to iOS 6 are fairly simple. If we know that, say 80%, for a nice round number, of iOS 6 search visits do not pass a referrer, we can say that only 20% of Google organic iOS 6 visits are accurately attributed.  With that knowledge, we take the number of iOS 6 visits recorded for Google organic and multiply it by 5 to get an estimate for the true volume, according to the following formula:
(1/(1 - % iOS 6 Search Traffic Without Referrer)) * (Number of Recorded iOS 6 Google Organic Visits) = Actual iOS 6 Google Organic Visits
To calculate just the number of Google organic visits hidden by iOS 6, we can subtract the recorded visits from the figure we obtained above or follow a slightly different path from the beginning using this formula:
((% iOS 6 Search Traffic Without Referrer)/(1 - % iOS 6 Search Traffic Without Referrer)) * (Number of Recorded iOS 6 Google Organic Visits) = Hidden iOS 6 Google Organic Visits
Users of Google Analytics can create a custom report to obtain the number or iOS 6 visits that were correctly recorded for Google organic search:

In the example above, we are applying filters to limit visits to the Source / Medium of "google / organic", the Operating System of "iOS", and the Operating System Version of "6\.".  Note that for the Operating System Version we are using a Regex, which will capture all versions of iOS 6.  Currently, it is common to see iOS 6.0, 6.0.1 and occasionally 6.1.

This example also includes a keyword-level breakdown for those SEO programs where it is important to include that granularity.  A quick aside, if you run this report, you should not see any (not provided) queries, because that issue is confined to desktop where Google employs the workaround to deliver referrers scrubbed of the search query mentioned above.

Case 2: We Can Only See Google Organic Traffic at the OS Level (Adobe SiteCatalyst)

In cases where we cannot see how much traffic was recorded for specific versions of iOS, our estimates require another assumption and get a bit more complicated.  Specifically, we will need to estimate the breakdown of the iOS bucket, or in other words, the iOS 6 adoption rate. Unfortunately, this is a moving target, so even if we have a strong method of determining this figure, we will need to continue to update it as more and more iOS users move to the latest version. As with determining the percentage of missing referrers, paid search data is likely going to be our best bet for estimating iOS 6 share of Google organic search visits.  RKG data shows that iOS 6 adoption is at 67% for the median site in December to date.  This percentage is pretty consistent across sites.

For this equation, let's go ahead and define some variables:

iOSA = Actual iOS Google Organic Traffic iOSR = Recorded iOS Google Organic Traffic AR = iOS 6 Adoption Rate MR = Percentage of iOS 6 Search Traffic without Referrer
First let's consider what constitutes the total amount of traffic that was properly recorded for iOS Google organic:
iOSR = iOSA * AR * (1-MR)  + iOSA* (1-AR)
The piece on the right -- iOSA* (1-AR) -- is the traffic we see from earlier versions of iOS.  So if iOS 6 adoption (AR) is at 67%, 1-AR is 33% and analytics programs properly record 33% of actual iOS Google organic (iOSA) as Google organic.  The left side of the equation is similar, but adds in the missing referrer percentage for iOS 6 to account for a portion of that piece of the pie being labeled as direct. The problem is we don't know the value of iOSA.  With a little algebra, we can solve for it though:
iOSA = iOSR/(AR*(1-MR)+(1-AR))
So if we can get a value for iOSR from our analytics program, we can plug that into the equation, along with the values we previously determined for iOS 6 adoption rate and percentage of iOS 6 traffic missing referrers to get actual iOS 6 Google organic traffic.  Subtracting the recorded value from the actual value will give us the amount that was hidden. To get recorded iOS traffic for Google organic out of the Adobe SiteCatalyst UI, we take the following steps: 1. Under the Reports navigation, click Mobile and then Operating System: 2. On the next page, click the icon just to the left of the iOS listing to bring up the Breakdown By menu: 3. Select Traffic Sources then Search Engines Natural: 4. In the filter bar, enter "Google". 5. Download the report and sum the values.

Conclusion

As we indicated up front, iOS 6 visits being improperly attributed as direct is already having a major impact on how organic search traffic levels appear in analytics packages.  A typical site that generates 25% of its traffic from mobile is already seeing its recorded Google organic search visits running about 12% lower than the actual volume.  If iOS 6 adoption were 90%, that figure would increase to 16%.  That's assuming mobile traffic share remains flat, which will certainly not be the case. While Google may step in and employ a workaround to deliver a referrer for secure mobile searches, we simply cannot wait for that to happen when such a significant amount of traffic is affected. To try these calculations for your own site, you can download this Excel spreadsheet and plug in your iOS data: Estimates for iOS 6 Google Organic Traffic (Excel)
Join the Discussion