How to Manage Keyword (not provided) for Google Search (Part 1 of 2)

Optimizing Your Website’s Google Analytics Reports for SEOs

Back in October 2011, we all noticed (not provided) starting to appear in our Google Analytics keyword reports. In the first few weeks only a small percentage of keywords were affected – this was the percentage of visitors to our Website who were logged in to Google at the time of a “secure” organic search (browser address bar reads https, for secure), which led them to our Websites.

Google then rolled out “secured search” across the USA and we saw the percentage of organic search visits with keyword (not provided) jump from about 2% to about 11%. Then in early March 2012, Google rolled out secure search to Canada and other countries – and depending on the percentage of your Website traffic originating from the USA, or from other countries, you would have seen the percentage of organic search visits with keyword (not provided) jump to about 20% of all your organic traffic. This growing trend towards obfuscation continues even today…  Some are even predicting that the growth in keyword (not provided) will amount to 40-50% of keywords from Google organic…

Yes, this is a very big deal, and there is a tremendous loss in digital intelligence and our ability to understand visitor intent (keywords are an important clue in understanding why visitors have come to the Website). In response there was a backlash from the SEO community who felt persecuted and threatened – but officially this was only a side-effect of Google prioritizing the privacy of its own visitors, and possibly preventing third parties from eavesdropping on search patterns for unfair advantage.

So, what should you do about it?

1)      Filter out (not provided) and (not set) from your Google Analytics reports. See the filter example here: https://www.cardinalpath.com/determining-what-to-write-about-with-ga-custom-reports/

2)      Realize that things are not as bad as they seem –meaning that if we assume people who are searching on Google will search the same way whether they are logged in to Google or not, then the keyword distribution will be very close to the same for both the “known keyword set”, and the “unknown keyword set” (aka “not provided”). By this law of averages, we can infer that the unknown set of keywords contains roughly the same proportion of branded keywords, brand and subject keywords, and the long tail keywords – which tend to be longer and more precise – and also tend to express the very precise intent of the individual visitor.

3)      Keyword reports can be amended to qualify the keyword numbers being reported. This means that you can project the known keyword densities into the unknown keyword set and add this together to get the total number of expected keywords that match a pattern, which will be correct within statistically acceptable limits that you specify in your report. So, the directionality of your keyword data, reports, and related recommendations are all still valid with this reasonable qualification. This thinking works where there are several (or many) instances of a known keyword. Note that because we are losing resolution along the entire curve this loss of data is most impactful in the long tail, because a data point may be lost altogether, and no longer represented in the data set at all. In truth, however, the ultra long tail contains one-offs and rarities that may have been difficult to action, or uneconomical to action in some cases. Again, where there is a clear pattern in the keywords, the pattern is likely to remain even with the keyword (not provided) issue.

4)      If you use paid search, all these keywords will come through into your Google Analytics reports, but don’t forget to link your Google Analytics to your Adwords first! ++ AND remember that paid search campaigns that target the long tail tend to be the most profitable, because there is usually little competition for rare and infrequent keywords, but when these precise keywords align closely with your business, products, and value propositions, the conversion rates will tend to be much higher! So we can use paid search intelligence to regain resolution in the long tail.

5)      Make sure your onsite search is prominent and very user friendly – don’t forget Google Analytics can be configured to capture 100% of the search that takes place on your site – and so we can start to rebuild our understanding of “visitor intent” here. Consider going “BIG” here! If you can detect that a visitor is from Google Organic and the search term is (not provided), then dynamically alter the appearance of the page to show a more prominent onsite search box!

6)      Remember that Google Webmaster Tools (GWT) will also provide keyword data that will allow you to figure out what some of the missing keywords for your top landing pages are. Spend time on the top landing pages on your site for organic traffic, or those strategic landing pages that lead to conversions (or should lead to conversions). This is where your time is best spent to uncover the most valuable insight. This GWT approach is described further in the Search Marketing Weekly post titled Summary: Google’s Not Provided Tips.

7)      Segment your audiences geographically to focus on your target audience – the following image shows the geographical distribution of online visitors with the keyword (not provided) – if your business has a national focus, then excluding international traffic may reduce the proportion of visits that are (not provided). To create this report in Google Analytics, go into your traffic sources, then organic search, then click on the (not provided) value, and finally choose the secondary dimension of “country”.

8)      Finally, one positive side effect of this growing data void is that now we will all know the proportion of visits to our Website who are logged in to Google at the time of an organic search, which led to a visit! Remember, just before Google’s stock price shot through the roof, Web masters and analysts were remarking that, “between 40%-60% of all Website traffic is coming to our site from Google…”. Perhaps the next big jump in Google’s stock price will be triggered by a similar realization that, “between 20%-30% of the visitors to our Website have Google accounts, and Google knows a thing or two about these audiences that will help advertisers get the right value propositions in front of the right individuals…”.

In part 2 of this blog post we will look at keyword reports that take statistical uncertainty of keyword densities into account. AND we will expand on any areas where there are lots of comments or questions!

Thanks for reading!