How to steal some 'not provided' data back from Google

Google's recent introduction of SSL encryption for search queries from logged in users means that a lot of valuable data has now gone missing from Google Analytics.

This post explains a useful Google Analytics hack to regain some of the insight we used to have, and improve life for you in light of the '(not provided)' issue.

First, here are three devastatingly important 'notes of caution':

  1. You need Google Analytics Admin access to apply the 'hack' included in this post.
  2. If you get the hack wrong, it can mess up your data a little bit while the hack is in place.
  3. Econsultancy & Dan Barker take no responsibility for your implementation of this little hack. (though feel free to get in touch with Dan on Twitter if you want to ask questions).

Background

If you haven't been keeping pace lately (and, with the changing of the clocks, and an utter dearth of new David Duchovny films, who could blame you?) here is the background on this:

Before October 18th 2011, whenever visits landed on your site from its search results pages, Google would happily tell you all of the keywords they'd searched for:

After October 18th, Google started hiding(!) lots of this information:

(CC images courtesy of jurvetson & marriageequality on Flickr)

In other words: Before, for every 100 visits from Google, very roughly speaking, you'd see what all 100 searched for; now, between 5% and 33% of those terms are hidden by Google.

You still get 100% of this info for 'paid search', but for 'organic search', much is hidden. How mean-spirited of them, eh? The worst case I've seen of this so far is a site with more than 10,000 '(not provided)' search terms each week.

For further background, an original announcement of this is here. & there's much discussion & opinion on this here and here, among many others.

How to fix this?

The bad news is you can't completely fix this. Sadly, the scallywags at Google have stuffed this data right into the back of the larder, though we keep our fingers crossed. 

The good news is, there are some things you can do:

  1. Firstly, the great Avinash Kaushik has put together a magical blog post detailing how to benchmark the effects of this, and to run some other, new reports to assist.
  2. Secondly, I've put together a nice little Google Analytics hack below to bring some useful data back into your regular Google Analytics organic 'keywords' report wherever '(not provided)' would now appear.

The hack: what it does

I'd said the hack puts 'some useful data back' into your keyword reports. What it actually does is this:

  1. Looks for '(not provided)' search terms.
  2. Where it finds them, it looks to see which page the visitor landed on.
  3. It then changes your keywords report in Google Analytics to show those two pieces of information (the fact that Google suppressed the keyword, and the landing page), rather than just the utterly anonymous '(not provided)'.

The hack: why it's useful

The hack will give you the following insight that '(not provided)' hides:

  • Get a much clearer picture of the intent of all of your (not provided) visits.
  • Get a vague idea of the search terms those (not provided) visitors may have used.
  • Quickly get a rough idea of the 'brand' vs 'non-brand' split, by simply seeing which (not provided) visits land on the homepage vs others.

I had originally been running additional reports to do some of the above, but I've found it far quicker & more useful to add that data back into keywords reports so I can analyse '(not provided)' data alongside my regular search data.

The hack: before & after

Here's a very quick example of the hack on a small amount of data.

The Keyword Report

The Google Analytics Keyword/Organic Report looks like this:

Not provided - before the hack

If we ignore everything else, and concentrate purely on the '(not provided)' section. Here's the information Google now gives us on these anonymous keywords:

Obviously, there's not a whole lot of insight we can gather from that. Are those brand terms? Are they generic? Are they long/short tail? There is no way of knowing.

Not provided - after the hack

Here's the information we now get about those '(not provided)' keywords:

Even from that small example, you can quickly see things like:

  • The vast bulk of 'not provided' data was from people landing on the homepage. (ie. a bulk will probably be 'head' terms).
  • 'Laptops' & 'cheap laptops' are next biggest landing pages among 'not provided' terms. I know these are likely to be driven very much by 'head' terms.
  • The rest is a little further down the tail of search.
  • The 'bounce' rate varies wildly among this, as opposed to the flat number we saw before.
  • We could also use this to see any e-commerce or goal data related to (not provided).

And that's purely an example on a tiny amount of data. On larger pools, there's a whole lot more insight you can gather.

The Hack: how to apply it

The hack uses a 'filter' to change the data Google Analytics records. If you're at all uncomfortable about this, you may want to test it on a 'duplicate' profile first.

To apply the hack, first we have to go to the 'filters' area.

Step 1: Add a new Filter

Here's a diagram showing how to reach the filter admin area in the 'New Version' of Google Analytics:

Step 2: Configure the Filter

To do this, we create a 'Custom Advanced Filter' (this sounds more scary than it is). Here are the settings to apply:

This tells Google Analytics:

  1. "Whenever you see a search term that matches '<any character>not provided<any character>', apply this filter".
  2. "Within this filter, we also want to use the address of the page reached where '(not provided)' was the referring search term".
  3. "When you apply this filter, discard the original '(not provided)' search term, and replace it with 'np -' followed by the landing page address".

Clicking 'Save' then applies that filter to the Google Analytics profile, and starts recording the data. (note: you can edit and remove filters at any time later).

Step 3: Test the Filter

To test the filter:

  1. Go to https://www.google.com (if it redirects you to Google.co.uk or another international site, you'll need to click the 'Go to Google.com' link at the bottom right of the page).
  2. Make sure your browser address bar contains 'https://'.
  3. Search for something you know you rank for. Preferably a page deep within your site.
  4. Click the link through from Google to your site.

If all has worked, after a couple of hours you should be able to view the keywords/organic report in Google Analytics and will see that you have an entry beginning 'np -' followed by the page you landed on.

Following that, you'll start to see reports like this:

Instead of just this!

Summary

So there we are. Not as good as the data we had before Google took it away, but far more valuable than the single '(not provided)' line we get as standard at the moment.

There are lots of possible extensions of this too. For example, you could clean up the URLs you see here if you have lengthy ones, use 'Page Titles' instead of URLs if more useful, etc.

Let us know what you think in the comments, and do share this via Twitter if you think others may find it useful.

Dan Barker is an E-Business Consultant and a contributor to Econsultancy. He can also be found on Twitter and Google Plus

Topics: Search Marketing
Add your own

Reader comments (64)

  1. Avatar-blank-50x50 Dan

    2:19PM on 23rd November 2011

    How is this any different by matching the keyword to (not provided) and choosing secondary dimension as landing page?

    Sorry dude, this seems pretty pointless and dangerous as it can screw up the data.

  2. Tim Leighton-Boyce Tim Leighton-Boyce

    Analyst at CxFocus

    2:35PM on 23rd November 2011

    I can certainly see a benefit in using this as opposed to just using the secondary dimension to show the landing page. The approach described here means that you would see this data mixed in with the other keywords.

    It's often the case that insights are to be gained by comparing these metrics in the context of others (eg in this case the real keywords which are available).

    Having said that, I think it is essential to do this kind of manipulation in only some of your profiles. It's absolutely critical to keep a raw unprocessed profile with no filtering, ignored parameters etc.

    You must have that available for cross-reference and sanity checking. In this case I would take a close look at the ecommerce data (if relevant) for 'not provided' in the raw profile and use that to confirm whether it matches the the data for the total of the manipulated keywords. You need to establish that the filter is also being applied to the attribution of ecommerce data. It may not be working quite as you would hope.

  3. dan barker dan barker

    E-Business Consultant at Dan Barker

    2:45PM on 23rd November 2011

    hi, Dan,

    thanks for the comment. I'd mentioned in the post:

    "I had originally been running additional reports to do some of the above, but I've found it far quicker & more useful to add that data back into keywords reports so I can analyse '(not provided)' data alongside my regular search data."

    That included search term vs landing page reports, just as you're talking about.

    It may not be clear from the post, but this does not touch 'actual' search data. It only affects '(not provided)' rows. Therefore, what I get is this:

    1. When the data I need *is* already there, it does nothing.
    2. Where Google has obfuscated the data, it adds in extra info that helps me to analyse the intent of the searcher.

    Whereas, using 2 dimensions side by side, I get a report which also splits up all of my *actual* search terms into tens of thousands of rows.

    I played around with this, found it much more useful, and that it saved lots of time for me in analysis (over custom reports/secondary dimensions/joining other reports together), and so thought I'd share.

    Play around with it for a couple of days & see if you find it useful. (&, as I say in one of the strong caveats in the post, try it on a duplicate profile if you have any concerns).

    Thanks again! :)

    dan

  4. Avatar-blank-50x50 Diego Ruiz

    2:49PM on 23rd November 2011

    Great post, thanks. Not too technical so we'll give it a go.

  5. Tim Leighton-Boyce Tim Leighton-Boyce

    Analyst at CxFocus

    3:08PM on 23rd November 2011

    As Dan has said several times already: you should always use more than one profile when you are using filters to manipulate data in this way.

    You may not want to add lots of profiles for different purposes, but you should always have at least one which is not filtered in any way.

    For an explanation of what I mean by an unfiltered or 'raw' profile, see here:

    http://www.cxfocus.com/index.php/set-up-google-analytics/2-profiles-google-analytics/

    Tim

  6. Avatar-blank-50x50 Rachael

    3:13PM on 23rd November 2011

    Love this! Makes sense for a secondary profile since it will cut down on fast access mode-affected data.

  7. Avatar-blank-50x50 Gerry

    3:33PM on 23rd November 2011

    Awesome idea, and really quite simple

    Would STRONGLY recommend doing this on a test profile first, everytime I mess with filters I get it slightly wrong and lose everything. I typically have 3 accounts live, test and nofilters ...

  8. Avatar-blank-50x50 Richard Baxter

    4:03PM on 23rd November 2011

    Thanks for sharing, Dan!

  9. Avatar-blank-50x50 AJ Kohn

    4:29PM on 23rd November 2011

    Maybe I'm missing something here but can't you get this information by viewing your (not provided) keyword traffic and selecting Landing Page as your secondary dimension.

    In my post (http://www.blindfiveyearold.com/not-provided-keyword-in-google-analytics) I show how to do the opposite and view Landing Page by Keyword (secondary dimension) but flipping this the other way around should give you what you're looking for.

    You may have to be using the NEW version of Google Analytics to do this but it should gives you the same information if I'm not mistaken.

  10. Avatar-blank-50x50 Jose Uzcategui

    5:33PM on 23rd November 2011

    I agree with @AJ and @Dan... this it seems the information is readily accessible with secondary dimensions if you drill down to "(not provided)" in the keyword report (organic in V5).

    @Dan, in your reply to @Dan you're describing introducing the secondary dimensions without drilling down to (not provided).

    perhaps we're missing something?

  11. dan barker dan barker

    E-Business Consultant at Dan Barker

    5:53PM on 23rd November 2011

    hi, AJ, thanks a lot for the comment. If you take a look at comments 2 & 3 hopefully they answer your question :)

    Thanks!

    dan

  12. dan barker dan barker

    E-Business Consultant at Dan Barker

    12:04AM on 24th November 2011

    hi, Jose, how are you?

    Here's one silly example of some of the value I've found in doing it this way.

    Let's say I have a page on my site targeting the phrase "cheap laptops" (and variations), with the url "/cheap-laptops".

    I want to answer the question "has my traffic improved/decreased for this term vs last week?"

    Here's what I do under the 'drill-down' method:

    1. Set the date range to this week vs last.
    2. Go to the organic keywords report.
    3. Do an in-report search to show only terms containing 'cheap' & 'laptops'.
    4. Look at the report, and keep that information in my head.
    5. Remove that search filter.
    6. Add in a secondary dimension for 'landing pages'.
    7. Add a search filter to show only terms exactly matching '(not provided)'
    8. Add in a second (AND) filter for 'landing pages' containing cheap & laptops.
    9. Remember all of the information from the first report, and combine it in my head with that second report.
    10. Make decisions & take action based on that.

    Do that for a dozen keywords for half a dozen sites & life gets boring quite fast.

    Here's what I can do under the method I described in the post here:

    1. Set the date range for this week vs last.
    2. Go to the organic keywords report.
    3. Filter to show only 'cheap' & 'laptops'.
    4. Make decisions & take actions based on that.

    Again, this is just one example. Give it a go on a few sites, see if you find value. I find lots of benefit in it, your mileage may vary.

    And thanks for reading!

    dan

  13. Avatar-blank-50x50 Hyderali

    7:45AM on 24th November 2011

    Thanks a lot Dan!!!

    On my way to test this & will let you know if its working fine.

    Thanks for posting such easy *hack* to know our keywords.Would love to know what other additional method you applied to get this report. I know they may be lenghty or very techy but would like to see them too.

    THanks.

  14. David Rankin David Rankin

    OWNER at Photogold ecommerce photography

    11:25AM on 24th November 2011

    I read a post somewhere that if you have SSL hosting on your domain that you will get all your keyword data. Is that true?

  15. dan barker dan barker

    E-Business Consultant at Dan Barker

    11:35AM on 24th November 2011

    @David,

    sadly not. That's one of the complaints actually - it would have been fairly straightforward for Google to have set it up like that. Sadly... they didn't!

    dan

  16. Avatar-blank-50x50 Paul Mitty

    12:05PM on 24th November 2011

    Hi Dan

    I was wondering what site the figures used in the example relate to ?

    Thanks

    Paul

  17. Avatar-blank-50x50 Cristina Chetroi

    1:07PM on 24th November 2011

    Why use profile filters when this data is already easily accessible in reports though second dimension or just by drilling in (not provided) and selecting landing page instead of keyword as your main dimension?

  18. dan barker dan barker

    E-Business Consultant at Dan Barker

    1:36PM on 24th November 2011

    hi, Cristina, if you take a look at comments 2 & 3, or my response to Jose, I think you'll find the answer.

    Paul: drop me an email if you like. dj.barker[at]barker.dj

    dan

  19. Avatar-blank-50x50 Kiril Bunin

    5:09PM on 24th November 2011

    It looks like a great idea but I am afraid it solves the problem which doesn't actually exist. I don't really understand quite a lot of things in this post. Would be thankful if you can clarify them:

    1. Since when do we call a rewrite filter a hack?
    2. How does landing page traffic data is "stealing" back the keyword data?
    3. And finally why secondary dimension or just landing page report with some advanced segmentation is different from your solution?

    I've noticed that some comments used the benefit of having all keyword data in one report as an excuse but it just doesn't make any sense to me. Why would you compare landing page traffic to the keyword data and vice versa? Your "np-/" with 129 visits can be either 129 various long tale keywords or just 1 short tale (which you can't even find out). What kind of insights can you get from that?
    Finally, bearing in mind that the biggest victims of "not provided" are mostly internet marketing content sites (Econsultancy would be a perfect example)with thousands of posts like this, ranking for millions of long tale keywords. This precise post can easily rank and attract traffic for such a huge range of keywords (analytics hack; not provided keywords, or even completely irrelevant stuff like steal data or Larry Page) but all you would see is a number of visits. How useful is that? Call me old fashioned but when I need landing page data I just go to the landing page reports rather than try to reinvent the wheel.

  20. dan barker dan barker

    E-Business Consultant at Dan Barker

    8:42PM on 24th November 2011

    hi, Kiril, thanks for the comment.

    1. A hack is an inelegant, imperfect, but workable answer to a problem. It often saves time and effort, and is often along the lines of the 80:20 rule. I'd say this fits that quite nicely.
    2. Google took some insight away from me. I implemented this 'hack' to bring some of that insight back. 'Steal back' is shorthand for that. There is limited space in the headline slot, and the first couple of sentences of the post pretty much explain it. I'm sorry if you felt cheated that we're not actively breaking into Google's Datacenters & escaping with illegal goods :)
    3. The way a lot of SEOs I know do their job is to structure things. They target particular keywords with particular pages, or particular types of keywords with particular types of pages, and they have a vague feel for the available search volume for those terms. Therefore, if they see those pages in this report, along with a particular volume of traffic, they can get enough of a gauge to see whether they need to bother investigating further. If you carry out SEO in a structured manner like that, you may find use for this.

    "Your "np-/" with 129 visits can be either 129 various long tale keywords or just 1 short tale"

    If your site is fairly 'scattergun', you are probably right there. In the case of lots of the sites I look at regularly, I know the rough 'keyword'/'collateral' (or brand/non-brand) split of the homepage, and whether it's steady/all over the place. Hence I can use this as a rule of thumb. As I say above, if you don't know the keywords you're targeting with a page, or the kind of keywords it ranks for, you can't do that.

    Finally, the last paragraph may be true for the sites you look at. I have personally seen this affecting a whole range of sites, not just those in an 'internet marketing' bubble. From speaking to friends over the last couple of days, I also know of 2 other sites in the 40k a month '(not provided)' range, and 1 in the 100k a month range. All 3 of those sites have very, very mainstream topics. One is ecommerce, one is content, one is a directory.

    Additionally, you probably know this hasn't fully rolled out in the UK yet. Once it does, I'm sure you will see a much greater effect.

    Anyway - as I say - I've been using the hack to reduce one of my processes from 10 steps to 4. (actually, just to 1, as I automate an email of the report). Your mileage may vary :)

    Thanks!

    dan

  21. Avatar-blank-50x50 David

    9:22AM on 25th November 2011

    Hi,

    Thanks for the filter.

    I tried to set up it but it does not work. I still have the not provided rather than "np - ..."
    Do you think I may miss an information not so developped in your post?

    Thanks a lot

  22. dan barker dan barker

    E-Business Consultant at Dan Barker

    3:54PM on 25th November 2011

    hi, David,

    feel free to send me a screengrab of your filter setup to dj.barker [at] barker.dj and I'll see if I can spot anything.

    I hope all's well!

    dan

  23. Avatar-blank-50x50 Geoff Jackson

    1:32PM on 28th November 2011

    It took a while for the not provided data in our analytics to start causing any major issues but it's picked up in the last 5 days.

    We've implemented this filter so be interesting to see the comparison between our profiles now.

    Thanks for this Dan

  24. Avatar-blank-50x50 Peppe

    1:04PM on 29th November 2011

    I agree with Kiril. You should use a secondary dimension and segmentation to analyze this.

    You are mixing up 2 dimensions in the same report which
    a. is criminal for any experienced BI professional
    b. will not help your end users understanding what they are seeing.

    Landing Pages are not keywords and vice versa. You assume that URL will match keyword searched by users which in many cases isn't true.

    Finally, when analyzing my Top keyword report I want to find out what are my Top performing keyword and their trend overtime. The fact that now roughly 10 % will not showing up anymore (numbers from Matt Cutts- higher numbers have yet to be proven) is very unlikely going to impact my keyword ranking because if that was the case this would mean that users logged in with their Google account have a totally different behaviour than non-logged in users. Quite frankly I doubt it very much.

    You mentioned the 80/20 rule. Even if 20% of data is missing you still have 80% of your data to answer the main question: what are my top performing keyword?

  25. Avatar-blank-50x50 Lee Driggs

    10:56PM on 29th November 2011

    I'm still getting all the normal keywords mixed in with this report. I just double checked the filter and it still seems correct.

    Anything I'm missing? I know I can just 'include' only words that start with "np -"

    Thanks in advance!

  26. dan barker dan barker

    E-Business Consultant at Dan Barker

    8:55AM on 30th November 2011

    hi, Lee, thanks a lot for the comment.

    Yes - you should see 'normal' keywords too. That's the useful part of the filter:

    1. When the data you need *is* already there, it does nothing to hide/break that.
    2. Where Google has hidden the data behind (not provided), it adds in extra info that can help to analyse the intent of the searcher.

    With hindsight, I should have included a screengrab of the data mixed together.

    Does that make sense?

    Thanks again!

    dan

  27. dan barker dan barker

    E-Business Consultant at Dan Barker

    9:36AM on 30th November 2011

    Hi, Peppe, thanks for the comment.

    On the point about mixing 2 dimensions: The first dimension is gone; Google have removed the data we used to get as a matter of course. Sad, but true! The hack does nothing at all when there is useful data there; only where it has been removed.

    The purpose of the hack is to add some useful data back in place of the missing info. I called it a hack in the post, and caveat it quite strongly. I hoped that would indicate that it is imperfect, but can be useful for analysis.

    I noticed you say a couple of times that your key concern is "what are my top performing keywords?" You're right, if that is your only concern, you can still (roughly) ascertain that (and applying this hack would not harm your ability to do that in any way). That question is really a 'reporting' question though. If you ask some 'analysis/optimisation' questions instead, you may find some value in this. For example:

    1. What are my brand visitors doing vs non-brand? (an absolutely, fundamentally crucial question that '(not provided)' really negatively affects.)
    2. What type of content are my visitors looking to consume on the site?
    3. What's the split in volume for my 'head' vs 'mid' vs 'tail' vs 'long tail' traffic?
    4. What's the split in 'buying intent' visits to the site vs 'researchers'?
    etc.

    As I said - the hack is imperfect - but it may assist with some of those questions on some of your sites (it has for me).

    And, as I said in the post, you can get this information in other ways. It can be utterly painful to do so (see my comment further up explaining the steps, and imagine doing it on slow analytics profiles with hundreds of thousands of keywords). So I came up with this to help me. I wrote the blog post to share the info and (hopefully) to expose a few people to some of the interesting possibilities in Google Analytics that don't get talked about much outside of the analytics community. :)

    Thanks again for reading, and especially so for taking the time to comment.

    dan

  28. Avatar-blank-50x50 Stuart

    11:19AM on 6th December 2011

    Hi Dan,

    thanks for the info in this post. I've implemented the filter and will see what it produces.

    Can I ask if anyone's been getting "(not set)" as a result, as opposed to "(not provided)"? That's what I'm getting in our GA implementation and so I've changed the custom filter accordingly, but I've seen references to "(not provided)" everywhere else. Is that something to be concerned about?

    Many thanks,
    Stuart

  29. Avatar-blank-50x50 Amy

    1:16PM on 13th December 2011

    Hi Dan,

    Thank you for a post! I echo @Stuart's comment about (not set). I now see 2 keywords - one is (not set) and another is (not Provided) where (not set) is much higher for my site. What is this new (not set) keyword?

  30. dan barker dan barker

    E-Business Consultant at Dan Barker

    1:18PM on 13th December 2011

    hi, Stuart & Amy,

    '(not set)' may be from your PPC campaigns. Do you run any ads on Google's 'display network'? Or do you run PPC ads that aren't perhaps linked up with your analytics account?

    dan

  31. Avatar-blank-50x50 Saumil

    12:40PM on 21st December 2011

    Thanks, for your helpful tips, i have heard that "not provided" keywords can be shown if you switch your account to Premium Version. Is this true, can you share your opinion on this.

  32. Tim Leighton-Boyce Tim Leighton-Boyce

    Analyst at CxFocus

    1:17PM on 21st December 2011

    I can answer @Saumil. No that's not true. The 'not provided' issue will affect all web analytics tools including GA Premium as well as the normal version of GA.

    Perhaps you are thinking of the keywords for Adwords paid advertising. Those are still available in the relevant reports.

  33. Avatar-blank-50x50 Jeff Bronson

    5:16PM on 3rd January 2012

    Thanks for this Dan, I'll need to apply to a test account first.

  34. Avatar-blank-50x50 Steven Chang

    12:33AM on 4th January 2012

    Cool hack, but won't a stats software like Awstats from your web host pick up the hidden data?

  35. Avatar-blank-50x50 WordPress SEO

    10:06PM on 27th February 2012

    We started out at 19% the first month, then dropped to 14% and have consistently grown and are at 21% now. So, a fifth of our traffic keywords are "Not Provided".

  36. Avatar-blank-50x50 dhanesh mane

    6:54PM on 4th March 2012

    hi I have applied the hack for one of my site and I have tested it from google.com ensuring https:/ but it still show (not provided) in keyword list :(, I have applied all the steps carefully, any clue? did any one facing the same issue?

    Thanks
    Dhanesh Mane

  37. Avatar-blank-50x50 Emily Mace

    2:45PM on 22nd March 2012

    I've referred to this a couple of times since it was posted and I've set this filter up on a couple of sites and it's working really well.

    With the news yesterday that Firefox will be defaulting to this (not provided) Https version of Google then we won't be able to see more of the keyword information in our reports, so this filter is going to be more and more important to us.

    Thanks

  38. Avatar-blank-50x50 Igor B.

    3:59PM on 30th March 2012

    Awesome. Nice hack. Testing out the filter, thanks for helping solve this little kerfuffle...

  39. Avatar-blank-50x50 Valentin

    3:17PM on 31st March 2012

    Awesome, thanks for this Dan

  40. Avatar-blank-50x50 Jonathan

    2:07PM on 1st April 2012

    That's a neat little trick - but I find that I still see a substantial difference in key metrics like conversions and bounce rate between (nor provided) data and other keywords landing on the same pages.

    I have to assume this data differs in some way, but there just is not enough information to take any positive actions.

    My current plan is to look at the poorest performing substantial volume keywords for each page and try to increase these in the hope they are also the ones in Not Provided, but I don't like the 'guess and try' when it if often the single biggest keyword source for a critical landing page.

  41. dan barker dan barker

    E-Business Consultant at Dan Barker

    7:07PM on 1st April 2012

    @Emily - thanks, Emily! Yes, very sad to see the 'not provided' percentage creeping up and up at the moment. I hope life's good at Vertical Leap.

    @Igor, Valentin - thank you very much - glad it's of use!

    @Jonathan - I agree, it's a big shame - that's broadly speaking what I do:

    1. Look at the top performing landing pages.
    2. Look at the 'known' keywords currently bringing traffic to those landing pages, and the keywords I know those pages are actively targeting.
    3. Act on opportunities to increase rankings for those keywords.
    4. Repeat.

  42. Avatar-blank-50x50 Mark

    1:35AM on 6th April 2012

    Hi Dan,

    I'm just trying this and so far it seems to be showing my own site urls instead of the keywords.

    IE np - /library/conversion-rates/slow-sites-will-kill-your-sales

    I've triple checked the filter with what you have above and it seems the same.

    Is this some sort of glitch or have I made a boob somewhere?

    Thanks

  43. Avatar-blank-50x50 Ryan

    12:20AM on 18th April 2012

    First off, thanks Mr. Barker for providing a method that may come in useful for some of us.

    I think there is more than one issue here that keeps the fires lit on this post.

    The Good:
    It's does provide useful information interspersed with accurate data that can be useful, making it really handy to interesting data in aggregated form.
    It doesn't skew actual keywords that are reported, but allows a side-by-side comparitive - think subjectvely people!
    It is useful for sites that target a specific keyword/page relationship.

    The Bad:
    It is not good for truly objective reporting about keyword performance
    It is not good for sites that may intersperse their keywords more naturally, or large sites that have keyword traffic drivers that you may not know are doing well. What if an secondary or ancillary keyphrase you used on any page starts to take off? How would you know it was the traffic driver amidst other phrases on the page?

    The Ugly:
    The term 'hack' was used irresponsibly here. Hack suggests a fix, and/or a not-so-elegant solution. This does not fix or solve the issue. In fact, I'll go so far to say that the reason many people landed here was doing a search variation for a 'not provided' hack for analytics. Obviously, you wanted to rank for this, so I give you props for result but not for ethics. I would also go so far to say that's why people are complaining in your comments - there was enough mis-representation in the title to frustrate your visitors. Lastly, stealing means take what wasn't given. With all due respect, landing page data is provided freely in Analytics.

  44. dan barker dan barker

    E-Business Consultant at Dan Barker

    12:46AM on 18th April 2012

    hi, Ryan, you're very welcome.

    The Good: Thanks!

    The Bad: It's a ready reckoner essentially - the data is not meant to be an accurate replacement. It can be used to aid human comprehension, but it does not spoonfeed perfect data. Hence the words 'hack' and 'some data' peppered throughout. :)

    The Ugly: Thanks for calling me irresponsible, thanks for making assumptions about my motivations, and thanks for the pedantic english lessons. ;)

    In all seriousness - I'm sorry you seem annoyed - I wrote the post to help, and not with any cynical motivations. This isn't my site & - though I very much love econsultancy and all they do - I don't even know how many views this post gets, and I don't benefit from it in any material way.

    I understood when I wrote the post that there would be negative feedback - there always is on posts like this, and I'm probably guilty of posting critical comments myself occasionally. On the other side of that, I've had a lot of contact from people who have used the 'hack' & modified versions of it to help - especially at the moment around estimating year-on-year numbers.

    If you're ever around in London & fancy a coffee I'd love to meet up. I love talking shop & I like to help people when/where I can - those were the real motivations for the post & for most of the writing I do online.

    All the best,

    dan

  45. Avatar-blank-50x50 Stalina Dsouza

    2:20PM on 31st May 2012

    Nice Post I Will Share This Information With My Buddy

  46. Avatar-blank-50x50 Sukumar Jena

    6:16AM on 27th June 2012

    Hi Dan Barker..

    Thanks for your excellent post..

    I had gone trough this and same i did for my analytic, Its done properly ..

    Can you tell me, how much time i want to wait for see my np result

  47. Avatar-blank-50x50 matt

    5:30PM on 3rd July 2012

    Dan,

    Thanks for this article. Google's (not provided) is very frustrating.

    I set this up and it seems to be working.

    That being said, our URLS are very long so the URL method is not ideal. You mentioned one can setup to use Page Titles or something else but I dont see exactly how to do this?
    Page titles would probably work fine. (or finding a way to clean up the URLs to shorter vs long URLs)

    ""There are lots of possible extensions of this too. For example, you could clean up the URLs you see here if you have lengthy ones, use 'Page Titles' instead of URLs if more useful, etc."

    1. How would one substitute page titles over landing page URL? I would imagine it might have to do something with the $B-1 but do you know what the code is for Page Title? Does $B-1 Stand for anything?

    2. How does one clean up the URL?

    Much appreciated
    Matt

  48. dan barker dan barker

    E-Business Consultant at Dan Barker

    6:50PM on 3rd July 2012

    hi, Matt, how are you? Thanks for the great question.

    1. To use the page title instead, simply switch the dropdown to the right of "Field B -> Extract B" to be "Page Title" instead of "Request URI". That will then set the variable '$B1' to be the page title instead of the URI.

    2. That's a bit more difficult & depends on how your URLs themselves are formatted. The principle is to use 'regular expressions' to pick the portion of the URL that you want to use. In practice that depends on what your URLs look like. If you post a few examples of your URLs or drop me an email I'll have a look.

    Hope that helps! Let me know if '1' works for you :)

    dan

  49. Avatar-blank-50x50 Levi | Stat Centric

    3:10PM on 29th July 2012

    I think this works fine unless you have really long urls. Also with a lot of your traffic going to your home page, you end up with a good deal of useless data (but still better than nothing as you have mentioned). I would agree though that using the secondary dimension would be a better option. Our non provided keyword traffic for Stat Centric is at 65% so we are desperately looking for a solution. PPC is so expensive for our market and the apparently the greedy bast##%s at Google are making this the only way to get good conversion stats on this missing data. We've been now refactoring our product to include more SEO tools as keyword data doesn't seem to be the powerhouse it used to be.

  50. Avatar-blank-50x50 Kanna

    7:22AM on 8th November 2012

    Everything is cleared.. But How do I apply the filter??

  51. Emma North Emma North

    Digital Marketing Executive at Koozai

    11:33AM on 29th November 2012

    This is a great blog post and still relevant today, more than a year after release! Also love Avinash Kaushik's blog post on Smarter Data Analysis of (not provided) which I also referenced in my own on the subject recently:
    http://www.koozai.com/blog/analytics/surviving-not-provided-keywords-in-google-analytics/

    Perhaps the only thing that has changed since this blog post was released is the ever increasing proportion of (not provided) keywords we're seeing. How long before everyone uses the https version of Google and all of our organic keywords are not provided? Scary! Good job we're already preparing with posts like this...

  52. Emma North Emma North

    Digital Marketing Executive at Koozai

    4:35PM on 3rd January 2013

    No problem; it's definitely important to be prepared against the ever-increasing impact of (not provided) and make sure we can get the most out of Analytics.

  53. Avatar-blank-50x50 Dessous günstig

    9:29AM on 11th February 2013

    String Tangas sind keineswegs nur Unterwäsche

  54. Avatar-blank-50x50 Jonathan Keller

    12:34AM on 20th March 2013

    Thanks for the tip, I will try it and see how it goes. I haven't read through all the comment yet, but it looks like there might be some gold there, so I will check back later.

    What someone said about copying your profile and making a test zone / having a profile without filters, sent me looking for how to copy profiles and I think this will help me greatly in the future.

    I was disappointed to find that the new profile doesn't have any data, but should still be quite useful.

  55. Avatar-blank-50x50 Nollywood

    7:43AM on 22nd March 2013

    Thanks econsultancy team. You gave us a good knowledge and we found more queries in our keywords data...

  56. Avatar-blank-50x50 Ghana News

    10:38AM on 22nd March 2013

    Thanks for this article dude. very useful for me

  57. Avatar-blank-50x50 Cat5e Plenum

    10:40AM on 22nd March 2013

    Thanks for giving knowledge about How to steal some 'not provided' data back from Google

  58. Avatar-blank-50x50 Kate

    4:28PM on 8th May 2013

    Hi Dan,

    Thanks for this very useful information. I applied the filter and my keyword results table is now showing "not provided" at the top with a large number (but less "large" than before the filter) and then below, the "np - URL" results are mixed in. It looks like this:

    Not Provided 7,000
    np - URL 2,000
    Actual Keyword 1400
    Actual keyword 600
    np - URL2 200
    etc.

    Do the individual "np- URLs" results add up to the "Not Provided" number at the top of the list, or are they separate results? For instance, how do I know, now that the filter is in place, how many searches for "Not Provided" keywords are taking place, regardless of landing page?

    Thanks.

    Kate

  59. dan barker dan barker

    E-Business Consultant at Dan Barker

    11:01PM on 15th August 2013

    hi, Kate, sorry I just spotted your comment - many months late!

    I suspect you've found the answer now, but, just in case not:

    Your data shows both 'not provided' & 'np-' because there will be a mixture of some data *before* the tweak was applied, and some *after*. Shortly, the '(not provided)' line should disappear, entirely replaced by 'np -' data.

    Hope that helps someone, even if too late for Kate! (sorry, Kate!)

    dan

  60. wintro teh wintro teh

    seo at Wintro Tech

    7:43AM on 1st October 2013

    I implement all these above instructions but I have not gotten the result yet.

  61. Nathaniel Bailey Nathaniel Bailey

    SEO Manager at TME Solutions

    11:55AM on 5th November 2013

    Hi, and thanks for a great post Dan!

    Just a little question about this bit at the end of your post

    "Go to https://www.google.com (if it redirects you to Google.co.uk or another international site, you'll need to click the 'Go to Google.com' link at the bottom right of the page)."

    Does this not work on google.co.uk? The reason I ask is because we are UK based (as are our clients) meaning the vast majority of visitors are from the UK and will be using google.co.uk rather than .com

    If it is a case of this only works for .com, do you have any work around to get it working with .co.uk please?

    Many Thanks,
    Nathaniel

  62. Radar Hill Radar Hill

    SEO Analyst / Marketing at Radar Hill Technology

    10:32PM on 23rd January 2014

    Good idea for getting relevant information in the same column, Dan, but on comparing the filter results to Analytics own landing page information, something weird appears to be going on:

    http://i.imgur.com/gFLq2WZ.png

    A lot of (not provided) not being sorted, filtered information not matching the stated landing page, and 0 visits vs. a number of unique visits. Any idea what might be causing this? Something faulty in the Campaign Term regex, maybe, or how Google processes landing pages?

  63. Victor Gonzalez Victor Gonzalez

    SEO Specialist at POP Creative Group

    7:42PM on 3rd February 2014

    Thanks, Dan! Beats relying solely on GWT. Will give this a shot, thanks for sharing.

  64. Jordan Caron Jordan Caron

    SEO Manager at Meaningful Marketing

    10:04AM on 9th February 2014

    This is such a great tip. I have so many of clients wondering what not provided is. Now I can atleast show them what page they landed on to get a better idea of the search term they used.

Log in to post a comment

Insight direct to your inbox

Sign up now

Get the latest Econsultancy blog content, tailored to your specific interests,
via our FREE Daily Pulse newsletter.