The Android world is certainly less uniform and it’s possible that some may pass referrer information, or that Google/Android may have been able to change something recently. Sadly, the amount of traffic from Android 4+ devices is small making due diligence here more difficult.

Perhaps certain phones aren’t passing referrer information?

In you’re first image after the “Reclaiming Stolen Search Traffic”, should the segment not be iOS 6.X instead of just all of iOS? This would take into account a sort of “adoption rate”.

Interesting article, however I’m in agreement with Christian Oliveira that this may not be the best formula for calculating the lost iOS 6 numbers.

I used a different formula to calculate the average percentage of direct traffic that was being contributed by iOS devices, then deducted that percentage from the affected months leaving me with a pretty close approximation of the increase due to the ‘bug’.

Obviously the formula I’m using won’t be as applicable for sites who are making use of strategies that will significantly vary their direct traffic numbers during the affected months.

The accounts I’m using this formula for are, thankfully, not using any particular tactics that would cause such deviation from the ‘norm’ (in terms of their direct traffic). So I hope that the figures I’m getting are quite realistic.

I do the same you do but instead of get the “real” percentage from months prior to iOS 6 launch, I get it from the iOS 5 visits on the same period, because I think that the direct/organic percentage can vary a lot through time (if you are improving in SEO but direct doesn’t grow for example). We can do this at least while we have enough iOS5 visits.

Also taking one or several month of data in sites with big traffic implies that you see sampled data, so accuracy decreases.

As you say in the update it’s the same in Android, so I hope they solve this quickly

As if (not provided) were not enough