Google announced on the 18th of October that keyword referral data would no longer be available for search queries made by signed in users on Google.com. This was due to their adoption of SSL i.e. the https:// protocol for reasons relating to privacy.
Ultimately, Google believes that users signed into their Google account deserve privacy with regards to their search activities.
This is a seemingly noble cause, though there has been much speculation as to whether it will continue once the premium version of Google Analytics has been launched. My personal opinion is that the scepticism is ill-founded due to the huge outcry such an obvious contradiction would cause – a company that commands such public interest as Google could not go back on its declared beliefs in pursuit of Analytics dominance…could it?
Whilst debate over Google’s intentions continues, the impact of the change has yet to be fully comprehended – and in some cases it has been miscomprehended.
How does it currently affect the UK?
As mentioned, the change was only announced by Google as being in effect for signed in users on Google.com. This led some to jump to the conclusion that it was only applicable for users in the US. However, the data I have access to seems to show that it is applicable to users signed in to Google.com in around the world – and has been since the 18th of October.
For example, one site with a significant traffic volume has received almost 6000 visits being specified as keyword ‘not provided’ since the 18th of October. These visits have originated from 134 countries.
Towards the end of last week, there were a number of reports that percentage of keyword ‘not provided’ visits was on the increase. Many webmasters in the US (where Google.com is of course very widely used) reported an increase over and above Matt Cutts’ initial prediction of below 10% of organic visits being affected.
The cause behind the increase is unclear. Looking into the data of the same site mentioned above, it can be seen that additional countries started referring keyword ‘not provided’ visits. Previous to the 31st of October, the visits could be attributed to 88 countries, including the UK. Prior to the 31st of October, the visits have been received from 134 countries.
However, the potential conclusion that the change has been rolled out to further countries is probably a red herring. It seems more likely that the SSL certificate was rolled out to a greater overall sample of Google.com users – no matter what their country of origin.
Again, using the same data as before, the number of encrypted visits originating from the UK increased markedly on the 31st of October:
But this was also the same for visits originating Australia:
Therefore in addition to the number of countries referring encrypted visits increasing, the number of visits from each country increased. Rather than new countries being added to the equation, it is more likely that the sample of users being affected was increased across all countries and this led to a greater probability that a user from a ‘new’ country would land on the site.
To conclude, it seems that the encryption currently impacts a sample of users around the world that sign in using Google.com. This sample was increased last week, meaning that it is applicable to a great percentage of users around the world.
It seems likely that the next stage of the ‘roll out’ will involve making encrypted search applicable to domains other than Google.com – including Google.co.uk.
What will the impact be to my website?
Currently, due to the encryption only being applicable to signed-in users on Google.com, the impact on your website is likely to be minor.
Prior to the 31st of October, the average percentage of affected organic visits across 11 websites (with primary UK focus) was just 0.3%. Following the probable increase in sample size on the 31st of October, this percentage has increased to 0.9%.
Webmasters in the US are experience percentages that around 10 time higher, so we should expect the same when the change is rolled out to Google.co.uk.
I believe the impact will eventually be seen to be widely contrasting according to the target audience of the website. For example, a B2B focused website targeting marketing professionals is far more likely to suffer from a significant signed-in user black hole than ecommerce counterparts.
Early indications (according to data that I have access to) support this notion. Though the proportion seems to be still comparatively very low in the UK, a brief study of 4 B2B websites and 4 B2C websites over for a week prior to the 31st of October shows that the percentage of ‘not provided’ searches for the former was just over double the latter – 0.65% compared to 0.32%.
Obviously this study has limitations, for example the comparative popularity of the websites in question, but the wide contrast definitely goes some of the way to show that B2B sites will end up being hit harder.
What can I do to retrieve the lost data?
Unfortunately, unless the premium version of Google Analytics will in fact contain this data, there will be no way to find out what key phrases signed-in users have used to find your website.
However, there are ways to estimate it.
The best way involves modifying your Google Analytics code to include a custom variable focused on recording the ranking of the key phrase referring the visit.
This will then allow you to discover the ranking of key phrase that referred the ‘not provided’ visit – plus the page that the visitor landed on.
This should at least allow you see when a ‘not provided’ term sends a number of visits to a particular landing page from a certain ranking position – and should allow you to attribute them to one of the small number of high traffic volume ‘short tail’ terms that the page ranks for.
For example, if the page is ranking 8th for a particular high volume phrase, and you receive 10 visits from a ‘not provided’ phrase with a ranking of 8th, then you can be almost certain it is the same phrase that is referring the visits.