Moz did their whiteboard presentations a few days earlier than normal, Danny Sullivan added his own analysis and SEOs weighted in with suggestions as to what people can do to deal with this gap in their keyword-based analytics data (the advanced filtering in Google Analytics as covered by Verve Search is perhaps the most useful suggestion I’ve seen). (If you’re interested in how we ever got to this point, then WebPage FX have today published a nice post explaining how it relates to the NSA and the Edward Snowden affair).
Some of you might recall that last year, around the first anniversary of (not provided) coming about, we carried out some research into the overall effect it was having on the data in Google Analytics account. We estimated (from a sample set of several hundred websites on our platform) that it was averaging ~20% around this time last year and the growth suggested we could be looking at ~40% on average by this time this year.This was looking like it had been on track to prove to have been a fairly accurate prediction; however, some SEOs have noticed a recent spike over the past month (largely it seems due to the latest version of Internet Explorer defaulting to SSL searches) and then this month, Google switched everything to https.
With all this is mind, we decided to run the research again to try and see what the growth had been up until August/September and what proportion of sites were actually showing as having (not provided) data showing at 100%.
A couple of caveats about the research:
- We only looked at sites we have been monitoring on the platform for the past year where we had GA data (obviously!)
- Not all sites have a complete year’s worth of data, as sites get added to or removed from the platform and because GA integration can be switched off by users
- Of those we analyzed, we were only interested in sites where (not provided) data was present at all – which turned out to be pretty much all of them
- The September numbers are not full month as ran the report on the 24th September (we’ll update it in October when we have final numbers for these sites for September)
- The percentage of (not provided) traffic is based on the following simple calculation: NP visits / Total Organic Visits (i.e. it isn’t just looking at organic traffic from desktop searches on a Google variant)
- The total data set covered 5,818 websites, but the highest record count was 4,727 for any one month
The results of this analysis are pretty self-explanatory.
It was obvious as soon as I started looking at this large Excel spreadsheet (and after I’d added some nice conditional formatting) that things had got significantly worse in last couple of months and some other things also stood out.
- As you can see from the above chart, there has been a significant growth in the number of sites where (not provided) now makes up more than 50% of their organic traffic data
- Looking at a larger sample size than currently provided on notprovidedcount.com (60) shows that the picture isn’t perhaps quite as bad as everyone is making out, but it is still very high – the September average is 57.9%, as opposed to the 70% currently showing on notprovidedcount.com.
- Only a few sites (36 in all or 0.006%) were already showing (not provided) data at 100% for September, but these were typically very low traffic sites
- There was no significant difference between ccTLDs (.fr, .de, .co.uk, .com and .es sites all showed NP averages in September of between 56-65%), supporting the assertion that Google’s latest change had been universally rolled out
- There was very little month on month growth in the NP average from October 2012 through to July 2013 (anything from 0.1% to 4.9%), but it then jumped in August (7.87%) and September 2013 significantly (14.76%)
So, it may be the case that all sites end up showing (not provided) visit levels at 100% next month, but we’re not quite there yet and we will only know when we update these numbers next month. So, stay tuned!
By: Matt O’Toole, Customer Experience Manager at Analytics SEO