A few weeks ago Google started to encrypt search data for logged in users. This essentially means that website owners will see no keyword data for visitors referred from Google.

We didn’t think too much of it at the time, given that just 0.68% of our visits were affected in the period immediately after this initiative came into play.

However, it seems that things have significantly worsened in November, at least for Econsultancy. 

Why keyword data matters

I have been looking at some of our top performing pages, to try to understand what makes them consistently popular. Ultimately my goal here is to identify certain words and phrases in search queries, in order to optimise our content (and therefore our business) strategy for 2012.

Knowing what phrases perform the best is crucial to making our content and our website better. Which keywords ultimately lead to sales? Which ones result in high levels of social sharing? Which ones have the highest bounce rates? 

Why Google has chosen to restrict our view is beyond me. It seems to go against its aims to improve the web experience.

Let’s look at the detail…

Page-level keyword analysis 

One of our most popular pages is a blog article called ’25 brilliant examples of Facebook pages’. Despite the fact that it was published in August 2010 it continues to attract thousands of visits and pageviews every month. Last month the post generated around 9,000 pageviews. 

Econsultancy has an internationalised website. We currently have US and UK versions of our site (econsultancy.com/us and econsultancy.com/uk). This is a little annoying when looking at the analytics (and also because the Twitter counter only tracks US retweets, rather than grouping in the UK action too), but it is revealing too, as we will see.

Below are two charts for the US and UK versions of this article. The date range covers the past month, during which time Google started to hide keyword data for logged in users. 

US page

Suddenly we’re seeing that ‘not provided’ sits at the top of our referring keywords for US visitors, representing 431 pageviews of the total 3,651 (that’s about 12% if my mental arithmetic is correct).

UK page

You can see that this issue isn’t quite so serious in the UK, with 82 in the ‘not provided’ field against a total of 2,419 pageviews (around 3% or thereabouts).

So the first takeaway, based on this small sample, is that the problem is magnified by a factor of four in the US, compared with the UK. I guess our US visitors are four times more likely to be logged into Gmail, or some other Google account.

November spawned a monster

What happens if we restrict the date range to just November? That’s when we think Google rolled this out on a widespread basis. The data below shows that the story considerably darkens… 

US page


A staggering 362 pageviews out of a total 1,138 are showing no keyword data. That’s around 33%. One in three search referrals from the US do not pass on any search query data. My jaw has dropped to the floor.

UK page

Again, the ‘hidden’ number has jumped. 70 pageviews out of a total 839 have no search referral data attached to them. Around 8%.

The big picture

What happens if we look at the site through a wider lens? Google sends Econsultancy.com around 40% of its traffic, so what proportion of that is now camouflaged? Let’s look at the entire site (including US and UK versions, and all pages), and let’s restrict the date range to the first 11 days of November…


Wow. The hideous ‘not provided’ is now showing as our top search referral. It has usurped our brand terms. It represents more than 10% of our total keyword referrals. And I fear it will worsen, as Google convinces (or requires) more people to log in to its products.

I guess the caveat here is that Econsultancy is a B2B website aimed squarely at the digital industry, so we might attract a greater proportion of users logged into the likes of Gmail and Google Plus (which could become a big deal as far as the rankings go, though social search shouldn’t evolve at the expense of this kind of standard data analysis). But there is a problem here, any which way you look at it, and I know that we’re not alone. B2C sites will be experiencing similar issues, even if the numbers look different.

So now it’s your turn. How does this look for you? You can use Avinash Kaushik’s fantastic custom report for Google Analytics. Do let us know in the comments below, and it would be helpful to know what kind of site / audience you have…