The Google Analytics (not set) and (not provided) Nightmare

In this article I explain the difference between (not set) and (not provided) in Google Analytics, share data from one of my sites showing the (not provided) keywords, and walk you through the changes we made that helped us resolve the (not set) issue in Google Analytics.

The impact of Google Analytics displaying (not set) and (not provided) has been a hot topic of discussion lately within the Search industry, and now things are about to get exponentially worse as Mozilla Firefox enabled Google’s HTTPS encrypted search as their default search service this week, and the change should affect regular users within the next few months according to this article (thanks @webaddict for passing it on). This could impact searches from up to 25% of Internet users who currently use Firefox as their default browser.

I recently did an analysis of one of the larger sites I work on, and the following surfaced. (Organic and Google total traffic numbers are removed for client privacy.) I found that (not provided) keywords do make up single digit percentages of Google traffic, looming around 4%, which is true to what Google’s Matt Cutts claimed back in November in a back and forth between him and SEOmoz’s Rand Fishkin. But I also discovered that the proportion of (not set) keywords is in the double digits and has been growing steadily and inexplicably since November.

Google Analytics (not set) and (not provided) Data

I’m sure other SEOs can relate – this is a tracking nightmare. We’re all left wondering, “Why are there so many (not set) and (not provided) keywords in my Google Analytics?” and with nowhere to turn to get answers – or data. My year-over-year comparisons have been rendered completely useless and unreliable, as we have no way of reliably knowing how my work is affecting the sites in the long run. There is obviously no way of knowing which keywords are impacted by the (not set) and (not provided) problem.  Yes, we can still track traffic, top content, engagement, and so on. But it will be an increasingly difficult struggle to identify which search audience and keyword traffic is most relevant and converts best on our site, so we can in turn continue to grow and develop content that is most relevant and useful for our users.

So what do (not set) and (not provided) in Google Analytics really mean? From what I’ve gathered, these are the current explanations:

(not provided) – This marker is a result of the Google encryption of key terms that drove traffic to your site if the searcher was a logged in Google user. Google announced in October 2011 that they wanted to “protect personalized search results” by encrypting those search terms – even though the searcher’s personally identifying data is not revealed to us in the Analytics console – and that SSL Search would become the default search experience for those users.  So while this traffic is reported as organic search traffic, you no longer get access to the query terms. Oh, and PPC AdWords users still gets to see their keywords; that data is unaffected.

(not set) – The Google Analytics blog in 2009 said that (not set) is “any direct visit or referral visit… because it does not have a keyword, ad content or any other campaign information associated with the visit.” This problem has been attributed to faulty auto-tagging on destination URLs and gclid redirection for keywords in paid campaigns. Hopefully one of you dear readers can explain this part to me, though: Many have said that (not set) refers to traffic coming from referrals or direct landings. However, I am struggling to understand why referral and direct traffic are coming from Traffic Sources > Search > Organic (excluding paid/PPC) and identifying their source as Google organic. Why is it in Search at all, instead of under Direct or Referral traffic? This has been identified as a common issue with AdWords traffic, but according to Analytics this is not paid traffic.

I welcome your insight and expertise in the comments, because I am honestly stumped on the (not set) issue.

UPDATE – SEPTEMBER 2012

Here are updated percentages of (not set) and (not provided) Google organic keyword traffic from November 2011 through August 2012. As you can see, it continues to climb and now it’s almost at 19% of organic Google keyword traffic. That means I can’t identify nearly a quarter of the keywords Google organic are sending traffic from. Still just on this one site (which runs both PPC and organic), and it spiked very suddenly in November from virtually none before-hand. It’s almost entirely desktop traffic, not mobile traffic.

Month NP % NS %
November 4.00% 13.53%
December 3.98% 14.62%
January 4.23% 14.98%
February 4.06% 15.37%
March 4.59% 16.27%
April 4.21% 13.92%
May 4.39% 16.15%
June 4.37% 16.35%
July 2.63% 16.80%
August 4.94% 18.79%

 

UPDATE – NOVEMBER 2012

We had a handful of findings and undertook some projects to clean up our Analytics in hopes of finding a solution to this (not set) and (not provided) problem. I’m thrilled to say that the (not set) keywords have plummeted to a comfortable zero. However, for the site referenced above, the (not provided) keywords almost immediately shot up almost as much as (not set) went down – roughly 20% – bringing the total amount of invisible Google organic keywords back to that 25% range we struggled with before.

So what that tells us is these terms may be interchangeable.  Though I’m still looking into that and how it times with the rise in full SSL browsers encrypting searches, and will update when I know more. Regardless, the “war on keywords” goes on as more full SSL browsers begin to surface, and as Google Chrome gets more and more locked down even for non-logged in users. It’s unfortunate, really.

Here’s what we did:

  1. Found and replaced all instances of duplicate Analytics code or any outdated Urchin Analytics code with the most current, up-to-date Analytics tracking code. (Truthfully, this should have been in good working order anyway, but we found several parts of the site that still had old or duplicate code on them.)
  2. Found that AdWords was connected to two separate Analytics accounts. We eliminated one duplicate account and ensured all AdWords channels were connected to the proper Analytics UA.
  3. Double-checked all ad tagging and made sure that all ads were pointing to landing pages with the proper www. version of the URL rather than the non-www. version and being forced through a canonicalization redirect.

All of these are important, but if you do nothing else, check into Item 1.  Ever since we scoured the site for outdated Analytics code and replaced it with the most current version, we haven’t seen a single “(not set)” keyword in GA under Search > Organic keywords. 

If this still doesn’t solve your problem, please check out the comments below for a very informative community discussion with lots of potential theories to consider. And if something else worked for you, please share in the comments below so the community can benefit from your learnings!

Heather PhysiocThe Google Analytics (not set) and (not provided) Nightmare

Comments 58

  1. Kyle K

    From my understanding (not set) can be from a couple different things.
    - Even though it is associated with AdWords tagging in the URL’s etc, if you block paid traffic, this keyword has lost any tracking associated with it, therefore it will not filter under Paid, its just a keyword, thus falling to the organic section by default.
    - Another instance is the gclid redirection. Under this instance, you can click on the “not set” entry (or filter all but that if you can’t click, which happens most of the time) and seclude that one set of data. Then select source and then medium from your secondary dimensions. You could also select landing if source or medium doesn’t give you any clues about the missing keyword. This should help you narrow down the source of this keyword so that you can pinpoint which keyword’s gclid isn’t behaving properly. Again this centers on AdWords.
    - This can also be a result that is the same as (not provided) issue. We have seen a couple clients that have one or the other or both.
    - Lastly, (not set) can be a result of Google Analytics having a bug or issue and the sourcing is lost. Since it’s a free service, they don’t love us. Haha.

    Those are about the only reasons I have found in researching the same problem in the past.

  2. Levi

    If you need 100% accurate data Google isn’t for you. However, if you measure website performance from a statistical perspective “not provided” isn’t too much of a problem.

    I tend to pull reports on keywords that I know and make generalizations about the entire site. If I know one sample of specific data is performing a certain way then the rest of the site should be the same. A good example would be political polling, the President’s national approval rating around 45% and it only took a little over 1,000 people for a statistician to get to a 95% confidence level.

    1. Post
      Author
      Heather Physioc

      Great insights, Levi. I have two follow-up thoughts.

      1) What if the demographic of registered Google users varies from those of non-Google users, and thus the data is skewed and the (not set) and (not provided) data is not in fact proportional to the entire site. It’s a good work-around, but there are too many variables without knowing what the Google user demo looks like compared to other demos in a given market.

      2) Couldn’t the size of the site also be a pain point? A small, 20-page, 500 VPM site could potentially suffer a lot more than a large, 11,000-page, 27,000 VPM site could theoretically get much better data to work with.

      I wouldn’t say I’m someone who needs 100% accurate data, especially in the volume I’m looking at, but I do like to minimize variables and get reasonably GOOD data. I just don’t think segregating an entire breed of user (Google users, and now Firefox browsers) gives us good data anymore.

    2. Caleb Whitmore

      Levi,

      It becomes a problem when your demographic is made up of 50% of people who are almost always logged into Google Services, such as with sites that attract a heavy IT audience that isn’t super Microsoft-focused.

      -Caleb

      1. Post
        Author
    3. Alberto Mesquita

      Well, the problem is when for example the average visit duration for the “keyword” ‘not provided’ is much higher than other keywords. The same happens with ‘not set’. Being so, what statistical conclusions can you infere if these parameters show you that the behaviour and eventually a lot of other things are different?

    1. Post
      Author
      Heather Physioc

      That’s an interesting possibility, Jaan. I have not heard about this. The site referenced in the graph above does receive traffic from more than 85,000 keywords. Wouldn’t that mean it’s pulling many, many more keywords than the first 50,000?

      1. Caleb Whitmore

        Jaan and Heather:

        “not set” has nothing to do with the 50k daily data table limit in the free GA version (premium goes to 1 million daily data table rows). When the daily data table limit is exceeded, etries beyond 50,000 are truncated into “other” and their data is still counted. I.e. the 50,001st unique keyword bringing traffic into the site on a given day would be changed to “(other)” in the GA reports and it’s session attributed to the same.

        “not set” comes in when there is literally “nothing to set” for the term portion of the UTMZ cookie. You’ll get organic showing “not set” when you have a visit from a google domain come in that has the referring markers of an organic click, but no keyword query present. This is different from “not provided” in that it’s essentially a false-positive.

        Hope this helps,

        -Caleb

        1. Post
          Author
          Heather Physioc

          Caleb, are you saying that these (not set) keywords that are listed under Google organic traffic, and are driving traffic to non-PPC landing pages on our site, may not in fact be a PPC misattribution error? I hope that’s what you’re saying, because I’m at a loss.

          I examined three other sites today (one of which also uses PPC, but two of which do not) of varying sizes and all three had a 0% (not set) rate for the entire 4-month period I measured. Only the site I showed stats for in this post has (not set) listed under Google organic traffic. So I’m stuck on – what’s different between this site, and the three others?

  3. Kate

    I am getting 40% as ‘not provided’ keywords in my “keyword conversions” custom reporting section. Is there any other free analytic software where i can track proper keyword conversions.

  4. Jacob

    I’ve got a (not set) when I look at which device my mobile traffic was using. Can anyone explain that? The previous discussions about (not set) and anything i’ve seen online talk about keywords. Any insights on (not set) devices?

  5. Michelle

    I have the same problem with the Google organic keyword (not set) increasing each month.  I use Urchin as well as Google Analytics, and although (not set) is showing an increase in both GA and Urchin, the % is much less in Urchin. Heather, it’s interesting that you have only one site that is experiencing this.  I would love to know the reason, especially if the % of (not set) keywords continue to increase.

  6. Megan

    I am actually having this problem with paid search. I can login to AdWords and see specific keywords, CTRs, etc. But I want to compare with site data such as bounce rate and time on page. But in Analytics, the ONLY keyword under the paid tab is (not set). How can this be, when I can easily get the data from AdWords? The Organic data is showing me keywords, although most of them are (not provided).
    This is a secondary site for a specific region. The site is Joomla and also bi-lingual. Can anyone help me out?

    1. Joseph

      Same Problem here, I am facing (not set) problem under the paid search after linking Google Adwords with GWA.

      And now I can’t compare the data I get from Google Adwords with Google analytics data.

  7. Troy Groberg

    I happen to have two analytics packages installed on my site. GA and Site Oracle. I find it interesting that Site Oracle has no problem showing me all the keywords and GA is giving me 17% (not provided). For people using the keywords report to do ballpark measurements I guess this isn’t a big deal but since I am testing new keywords I really like to know that I have good data, especially my paycheck is directly tied to the performance of these new keywords. IMHO this is Google demonstrating how they see SEOs as the enemy. The fact that we can still see the keyword data from the paid search and not the organic just shows that their excuses are bogus. What they want is to get us to focus less on organic, and more on paid. Duh, does that surprise anyone?

  8. Marcus

    Hey Megan

    This is just a configuration issue, an annoying one for sure but one that is easily sorted for the paid traffic – take a look here for that one: http://www.bowlerhat.co.uk/blog/adwords-ppc-keywords-not-set-simple-fix/

    With organic it gets a bit more complicated. Assuming the PPC traffic is tagged so Google is not mistaking paid for organic (if paid shows 0 visits this could be the case) then seemingly the (not set) keywords are coming from a variety of sources.

    I currently see (not set) at around 10% of visits and then comparing this to direct traffic I see that direct accounts for 90% of those (not set) keywords.

    Ultimately, I am not worrying about this yet. It may be a bit more scary next month when Firefox 14 gets more widespread adoption (currently seeing firefox at about 25% but 14 at only 2%) but I am seeing most firefox 14 searches as (not provided) rather than not set.

    Ultimately, not worrying about (not set) too much at the moment but (not provided) is proving to be a little more worrisome and despite the conspiracy theorising this does walk us ever more down the pathway of needing a paid search component and specifically where the privacy of those users that Google cares so much about is compromised for people paying for search clicks.

    1. Post
      Author
      Heather Physioc

      I am struggling to understand why referral and direct traffic are coming from Traffic Sources > Search > Organic (excluding paid/PPC) and identifying their source as Google organic. Why is it in Search at all, instead of under Direct or Referral traffic? This has been identified as a common issue with AdWords traffic, but according to Analytics this is not paid traffic.

  9. Chris

    I have a number of blogs that I monitor for clients. The volume of “not provided” is increasing at a very rapid rate for many of my sites. One example is a site that I manage for a company that sells walking poles. I’m seeing 22%. This is very concerning to me. They sky is not falling, but the transparency that helped businesses make better decisions about what to sell and how to market, benefiting themselves and the consumer, is disappearing. IMHO it’s a big step backwards.

  10. George

    We’ve been seeing increasing amounts of (not set) showing up for landing pages in GA. This really boggles my mind as I can’t find any references to (not set) in this context.

    I did discover that one of our GA event trackers was causing a lot of these (not set), but even after removing that tracker we’re still seeing issues in profiles (not the main profile of the site).

    It also seems to be related to “phantom visits” in our profiles where there are visits with no pageviews.

    Any thoughts?

  11. Debbie

    I also am seeing both (not set) and (not provided) in one of my clients’ Google Organic keywords report– both are listed as #1 and #3 respectively in order of top referring keywords. They do not have PPC running, so that cannot be the issue. I understand where the (not provided) comes from but still haven’t found a reason for the (not set) keywords.

  12. Mark Hughes

    Word of caution – this is speculative based on a hunch. However, (not set) looks to me like it could also refer to visits from ad groups or campaigns that you have since deleted. AdWords cookies are set to expire after 6 months, so if you have made any drastic changes to the set up of your campaigns, you might see a rise in (not set). At least I am, so interested to see if anybody else has experienced similar results?

    1. Steph Fabb

      Hi Mark,

      Have you been able to verify this at all?

      The site I am looking at is using the old urchin.js tracking code, I’m wondering if this might be compounding the issue…

  13. tek

    Heather, I feel your pain. I discovered your blog because I have been trying to understand this same issue. I manage a large site that gets nearly 4M organic visits/month of which around 600,000 are identified as (not set).

    Everyone always starts talking about PPC issues, but it is the organic search reporting where this (not set) issue resides. If it is simply a bug or limitation, I suppose I can live with that, but it’s difficult to accept when Google has published explanations for both (not provided) and (not set), neither of which explain this issue.

    1. Post
      Author
  14. Brock C

    I recently spoke with Google support about this issue. I am not quite experiencing quite the same results as noted above. First, my (not set) data is only from paid traffic, not organic. I only have (not provided) traffic within my organic results.

    So assuming you are working off of paid traffic, I couldn’t get much out of the rep to explain exactly why I was seeing this (not set) keyword (and by the way, its representing about 13% of our paid traffic). I received the same response you have seen in most of the threads. What she did help me confirm is that these “visits” are not being contributed to any clicks I am paying for. For example, on your left hand nav, open Advertising – AdWords – Campaigns. Above your line graph, under Explorer, choose “Clicks”. Hopefully, your (not set) keywords should be receiving zero clicks and impressions (though they are obviously receiving “visits”). If you are, I would check in with Google support.

    Granted, this doesn’t really help explain the (not set) issue, but is a quick check to make sure you are not paying for any of these visits. Overall, your total “clicks” in Analytics and AdWords should be fairly close to one another (within 5%).

  15. Peter

    Small business owner here…(not set) and (not provided) combine for 60% of all visits to our site per Google Analytics. Most of what I’ve read here is over my head but I’m grateful to learn that professionals are also confused by this problem. I’m curious to learn if this is a matter of Google being secretive or the technology not being able to determine the keyword.

    1. Post
      Author
  16. Amish Keshwani

    Hi,

    Just want to share my recent experience, I m working on one of the leading b2b portal since last couple of months & noticed that, now NP is increased to 24% of total organic visits. I have no clue what majority of traffic searching for when landed on my site.

    I heard that, by integrating Webmaster tool with Analytics, we can get trend on search queries but not sure. Love to know if you have tried it out or any other way as a solution to this issue ?

  17. Linda

    It’s becoming a nightmare not knowing what keywords/queries are been used to land on specific pages. Today, 80% of all website traffic is not set.

  18. Scott Bateman

    “Not set” traffic is all direct traffic. It can’t have any keywords if someone is using a bookmark to return to the page or if they know the domain name and are typing it directly into the browser’s address window.

    “Not provided” traffic is from browsers with aggressive privacy settings that block tracking software from collecting information from the users.

    For example, Chrome has these settings:

    - Block sites from setting any data
    - Block third-party cookies and site data

    It’s becoming more useful to track landing pages rather than keywords because landing pages reflect site activity and not user’s browsing information.

  19. Barbara

    I’m getting “not set” at a high rate in my paid search traffic. It has increased to 25%. If it is not associated with a keyword and coming from referrals how can it end up so high in my paid search?

  20. Sushil

    I am not sure but (not set) keyword can be the source what people getting though email or something. Like my site having same issue and i think it’s becoz of those subscriber, who has subscribe to my website’s article feed.

  21. Doug

    Year-to-date, (not set) and (not provided) account for 60% of our search results. Interesting to read this thread and see the varied insights on the cause(s) of this.

  22. annapaws

    Hiya, thanks for all the above, reading with interest….
    My web site was panda’d on the 28th Sep. Now interestingly here, is that the (not set) visits plummeted by 77%. I’m only a small web site and small business. This has had a devastating affect. Many thanks in advance if anyone can shine more light on the not set thing then. cheers. annapaws

  23. mikevallano

    Hi Heather,

    I’ve got a client’s site with 6% (not set) and it’s all under organic. It started in October 2011, when (not provided) rolled out, but since then they’ve had a total of 18 (not provided) visits–or about 0%.

    I noticed that they’re using the Urchin tracker for GA, and I’m wondering if anyone else noticing the (not set) under organic traffic is also using this old version of GA code.

    I work with a lot of sites, but this is the first one where I’ve seen (not set) vs (not provided). It’s also the only one I’ve seen using the old Urchin tracker, so I’m curious if there’s a correlation there.

    1. Post
      Author
      Heather Physioc

      Mike – I actually just discovered this on my own site I’m referencing above about 3 weeks ago and we just went through the whole site and made sure to get rid of any duplicate or Urchin code and I was just about to update this post to say that (not set) PLUMMETED IMMEDIATELY AFTER, so I’m optimistic that this has something to do with it. NICE observation, sir! Our next step is to ensure all our AdWords channels are plugged into the proper Analytics UA and then to double check all campaigns for proper tagging.

      1. mikevallano

        Very interesting! I’m hoping to get Async up on this client’s site in the next day or two, so it’ll be sure to let you know if that removes the (not set) visits.

        I’m also working on a more thorough Analytics audit for any new client to catch this and other issues more quickly.

        1. mikevallano

          Yes! Updating the code to Async has dropped the (not set) visits and I’m now seeing the appropriate amount of (not provided).

          Given your experience and mine I think it’s safe to conclude that the Urchin GA tracker will show (not set) visits in organic.

          1. Post
            Author
  24. Dana Tan

    Hi Heather,
    I’d like to comment on the thread between you and Mike regarding Google analytics outdated and Urchin code. I use the asynchronous code on small, very new site that I manage for SEO and in November 2012 (not set) accounted for 4% of my paid traffic, but a whopping 28% of my paid search revenue. I don’t know how I’m supposed to be able to look at an Adwords report and even begin knowing what keywords produced conversions and which ones didn’t. Google Adwords apparently has a transparency problem that could cause them to lose my paid search business completely if they don’t figure it out.

    Thanks for this great post Heather. I grew up in KC!
    Dana

    1. Post
      Author
      Heather Physioc

      Hey Dana! Nice to hear from a fellow KC native. Where are you at nowadays? So let me make sure I’m following here. You’re getting (not set) in the Google AdWords console, not Google Analytics?

  25. Kevin Mahoney

    We ran into the same issue and we fond out that our rewrite rule for legacy AdWords URLs was converting the GCLID to lowercase. With GCLID converted to lowercase, it broke the encryption that Google uses and so the keywords and campaigns could not be set. It took us 3 months of troubleshooting to figure this one out.

  26. Omar A

    Excellent explanation thanks.

    Our customers are demanding answers to the not set stats as it’s urgent for some cases to know the exact demography of the search locations for example.

  27. Mike Anecito

    Great post and set of replies from the readers!! Very informative.

    My question is related to the (not set) issue, but in our case the (not set) is the Network Domain, Service Provider and a number of other fields. We saw a spike of direct traffic to the home page with 100% bounce rate and I was going to go in and filter out the domain as I’ve done for other bots (that don’t play nice)…but, without a domain to enter, it appears my only option is to crawl through the log files and search for IPs.

    Any other suggestions on this new twist on the (not set)?

  28. Chewy McGillacutty

    So at the end of the day, does someone have a clear cut answer to why (not set) shows as a keyword in the organic search section of analytics? If it’s appearing as a keyword, how can there by no keywords associated with it!?

Leave a Reply

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