seo

How to Analyze Google Analytics (not provided) Data

It has been over three months since Google announced the (not provided) update that would protect privacy hide search referral keywords for organic traffic. Since (not provided) results only show up while logged in to Google, the increase to over 60 million Google plus users – and estimates of 400 million by the end of 2012 – indicates that the problem of search referral data is not going away anytime soon.

While Google’s own Matt Cutts estimated that “even at full roll-out, this would still be in the single-digit percentages of all Google searches on google.com,” SEOs and web marketers are seeing that percentage grow steadily past initial estimates. While this might hold true when looking at total Google searches, likely it is not true for you or your clients. If you don’t like this kind of uncertainty in a career, then maybe SEO isn’t for you. However, if you love a challenge and finding creative ways to overcome obstacles, read on.

Which (not provided) results are branded?

One of the most immediate problems that the Google update caused was the inability to differentiate branded vs. non-branded search traffic. Before, simply excluding the branded keywords in Google analytics was enough to see how much traffic was coming to the website directly from SEO. Now, you get something more like this when you exclude brand keywords:

This accounts for 14.6% of the total (non branded) search referral traffic for this particular site, or 373 keywords that remain behind the mysterious (not provided) veil. That means that only 2177 search referrals are known keywords. While not all of these 373 visits are branded, there are enough that are to skew the data making it harder to track growth over time of SEO efforts.

Before you go cursing the Gods of search, take a deep breath and count to ten. After you find your inner Zen, let us look at how to solve this problem. Follow these three easy steps:

  1. Keyword Landing Page – As you probably know, there is a little option to check the landing pages that each keyword was linked to from the SERPs.

Google Analytics Filter

  1. Filter not provided – Once you click this option and turn it on, scroll to the bottom and filter the results by typing in “not provided” (with our without the parenthesis).

  1. Look for “/” – Once you have your results filtered, you can easily see which (not provided) results went to the home page and which went to internal pages. The number of (not provided) results that went to the home page (showing up as “/”) indicates roughly the number of branded keywords.

Conclusion:

  • In the above example, there are 92 results that went to the home page which indicates that 281 of the (not provided) results are non-branded keywords. This means that there are now 2,458 referrals that are non-branded.
  • One draw-back to this technique is that you cannot conclusively say that the 92 results to the homepage are ALL branded. In this particular example this may not be a huge deal since 92 visits is only 3.6% (92/2550) of the total known non-branded keyword traffic. However, with a site that gets much more traffic, 3.6% could easily turn into hundreds or thousands of unknown keywords driving traffic.
  • One potential way to estimate the number of branded keywords that are coming from logged in Google users is to look at historical data. For example, if before the Google update your site received 30% branded search traffic, then you can take that number and multiply it by the (not provided) searches that went to your home page. In the above example, that would mean that roughly 30 visits of the (not provided) were branded.

Differentiating (not provided) keywords

Once you have discovered how many (not provided) search referrals are coming from non-branded keywords, you are still left with gaping hole in data. With a little bit of patience and creativity, you can segment (not provided) keywords into general categories of search traffic making it especially useful for any type of analysis that depends on categorical keywords rather than exact match keywords. This method will only work if the keywords that ARE provided by analytics are matching up nicely with the correct landing pages.

So you have a nice list of (not provided) keywords and which pages they landed on. At a quick glance, you can quantify roughly how many keywords of a certain category were searched while logged in.

In the example above, on the second and third rows one can tell that 32 visitors found the website by searching for the broad match keyword “zorro zoysia grass,” and 28 for “how to lay sod.” While there are a number of keyword searches that could land on /zorro-zoysia-grass-sc (like “zorry zoysia grass”, “zoysia grass”, or “zoysia sod”), you now know that over this period of time 32 keyword searches related to “zorro zoysia” landed on your website.

This is wonderful information by itself, however sometimes you need more than at-a-glance data. When you have hundreds or thousands of keywords and landing pages, having an automated approach is necessary for analyzing large datasets. An entire new post could be written on how to do this using Excel or Google Docs. However, if you read Richard Baxter’s post on how to do keyword research using categories, you can take a lot of the same logic and formulas to segment (not provided) traffic into keyword categories to measure the change in categorical traffic.

As Google Plus continues to grow, organic search referral data will continue to decrease. How will you adapt to the lost data of Google Analytics?

Related Articles

Leave a Reply

Your email address will not be published. Required fields are marked *

Back to top button