[Have a look at all updates below. The phenomenon is way more important than what Google first announced it would be. December 13th: I just keep updating this post with new links because we are seeing worse impacts than what the benevolent G (as in giant Google) said. Still, the peasants do not revolt]
[October 3rd 2013 Update] It’s too late, the Organic Search ship has sunk. The message seems to be “Go f**k yourself, and buy Adwords!”
I have wanted to write this post for some time, but decided that I would wait a little for the conversations to settle a little bit.
As you probably know by now, Google has decided to hide Organic Search terms from Web Analytics reports (all products) when visitors did their searches while logged in their Google account. Here’s the official Google Analytics blog that “explains” why.
It has been highly criticized by many. I think one the most outraged comments came from the very respected SEO expert, Danny Sullivan, whose opinion on this matter is really worth reading (For stronger language from Joost de Valk see here). He goes as far as accusing Google to be hypocritical about privacy, since they are willing to “sell” the search term information as soon as an advertiser buys AdWords. Anyway, this is not the discussion I want to have here today.
Google said that this would have very little impact on analytics results since very few people are logged in when they search. OK, this is probably the case for now. My guess is, however, that the ratio will increase with the coming years. Actually, it seems more and more people are already noticing that “not provided” search (the term used by Google to replace the actual keyphrases) accounts for larger percentages (see here, here, and here). Well, I’m not surprised.
The news was also downplayed by many Web Analysts (not all with obvious vested interests in being Google’s minions), who rightly said that there were many more valuable and useful analyses that can be performed in sophisticated online analytics. I agree.
However…
Organic Search analysis, although of a minor type, is still important (and SEO people would rank it even higher). Depending of what percentage of your traffic it represents, knowing what people had in mind when they searched is not trivial information. Have you ever looked at all those visits who got to your site through super strategic keywords to your business, and left right away, without doing anything of value? Makes you stare in the distance for hours… Definitely something you would want to know.
Also, knowing if people came to your site via brand-related or generic product/service-related terms makes an enormous difference. Is Organic Search in customer acquisition mode, or simply in a navigational one, meaning people are using search engines just to get to your site, since they less and less bookmark or type URLs in the browser (I wrote about it years ago here and here)? That also makes a huge difference.
Add to this that the “navigational” type of search would, in principle, be more prominent toward the end of your customers’ purchase cycle, when they are using search just to get back to your site before converting (if your products/services tend to sell over multiple sessions, which is a lot of cases, since the Web makes competitor comparison so easy), hence attributing campaign conversion to those visits is greatly erroneous. One wonders how well hidden Organic Search terms play with Google Premium new attribution module. If you don’t want to over attribute to Organic Search, you need to be able to discriminate between generic, category-related searches, and brand-related, navigational ones. What kind of fine, accurate attribution analysis will you be able to do when 30% plus of Organic Search visitors are logged in?
So, no, you won’t get your CEO to kiss you on the mouth with Organic Search term analysis. But it is still something important in your analytics toolbox.
And why the heck hide that information if I can’t link it to any kind of user’s PII (Personally Identifiable Information)?
Google doesn’t say.
[November 15th Update]
Fellow Twitterer (!) Dan Barker has brought to my attention this recent article by Econsultancy (November 11th) that shows how more impactful this situation seems to be. Aptly titled The Horror: Google now encrypts up to 33% of search referral data. Worth a read.
Another good read that bring more data to the horror story, by Hubspot Google’S SSL Change Actually Impacts 11% of Search Traffic (thanks to Tudd Bullivant for bringing it to my attention)
[November 17th Update]
This one, on SEOptimise shows showa that “not provided” is now the top term in several instances. Tries to figure out why Google is doing this.
Here another account by LunaMetrics that shows variations in the numbers.
[November 21st Update]
Avinash Kaushik executes somersaults to demonstrate ways to work around the (not provided) situation. I just believe bringing the keywords back is a much simpler one.
[November 27th Update]
eConsultancy also offers some hack to workaround some of the “not provided” terms, whereas the real measure to undergo should be to collectively ask Google to reverse that idiot decision (yes, Google can be an idiot).
[November 28th Update]
Alec Cochrane at Digital Transparency is reporting that SiteCalyst has updated their reports to reflect the new (not provided) phenomenon. He is also reporting important percentages of visits coming from that “term”. I myself have pressed Webtrends to do so for some time already. I have no idea when they will manage that in their reports.
[November 30th Update]
And the workaround dance continues, this time with Yahoo!Web Analytics showing you how to display the (not provided). Wouldn’t it be simpler to tell Google to just bring back the search terms? Are we all that afraid?!
[December 5th Update]
Some more stats from WebProNews showing the not provided impact is already much, much more prevalent than what was first announced by Google.
[December 13th Update]
SEO guru Ian Lurie is getting pissed, really pissed, and here, a few weeks later, he offers some workaround, but it’s ain’t a party.
If you have a big pot of soup and want to adjust the salt, you only need taste to a teaspoon. You don’t need to taste the entire pot.
I don’t need to see all the search phrases. 70% is plenty.
Who the heck cares about private search?
Hello Greg
Sure, you’re right, as long as the sample is big enough, we should be able to get reliable insights on the situation (assuming the times a term is masked and unmasked is evenly distributed). The discussion then becomes a question of how low we can go. Would 10% of declared terms be enough? Maybe yes; I’ve seen smaller samples.
What bugs, though, is why would Google do that? Why not transfer the terms if I can’t even associated them with the actual person? Why is it OK if I paid for an Adword (Sullivan’s point)?
And as a matter of principle, why should I settle for an eventually very small sample of a full data set I used to have?
I wondered the same thing – what does being logged in have to do with privacy? Why would Google think users would have an expectation of privacy around their keywords just because they’re logged into Gmail, or G+, or Google Analytics – etc. (Do users even stop to think or realise that they’re logged in?!) And of course, they’re choosing not to pass keywords, but still passing the fact that users came through Google. If users expected their keywords to remain private, perhaps they’d expect that all the referring details be private, but of course Google wouldn’t possibly do that.
I’m sure Google has a reason for what they did (PPC revenue? Maybe Bing is having a decent hit on it, or at least more than they expected?) I just find it frustrating that they’re not honest about what that reason is. The privacy excuse is a really flimsy argument and just doesn’t add up.
Hi Michele
We can debate for a long time about the real impact on our analysis, but the question of Google’s motivation remains…
Hello, Jacques
I do agree (and all AT Internet team with me) with this interesting analysis of the so well-named “navigational” type of search.
We called it differently in our solutions when we decided (a few years ago) to divide the “search engine source traffic” into Search Engine (understand “real requests”) for one part and “brand awareness” (e.g. “navigational”, and customizable by users) for the other.
But, definitively, I’m very fond of your expression : navigational type of search.
Thanks
Salut Jean-Marie!
Yes, I use that term to express the fact the people use search engines (well, let’s say Google, we all know that’s what they use) to get to sites they already know, thus “navigating” to the site. Search is then definitely NOT in client acquisition mode.
Yes, AT has cleverly made the distinction for years in its solution, and any Web Analyst needs to be aware of those very different types of search engine use.
My regards to the teams in Bordeaux and Paris!
Excellent job of clearing away the smoke!
“What bugs, though, is why would Google do that?
Why not transfer the terms if I can’t even associate them with the actual person?
Why is it OK if I paid for an Adword?”
I haven’t seen a thing from Google that addresses these simple questions.
Hello Chris
A real mystery ;). I wonder how other solutions will treat it. So far, it seems that the “not provided” term is actually the one shown in Google Analytics, and I can’t see it in other solutions.
Regardless of what is going to happen with all web analytics products, it would be nice if Google clearly explained why we need to cope with that new pain in our collective, analytics arse.