I was just about to answer a question to a new client when the implications struck me.  The question was about direct traffic in analytics.  I was busy explaining what it is, where it comes from and was just about to say something about how it might be a good thing to encourage and traffic from a broad range of sources might be seen as a quality signal by Google.  Just as that was coming out I had a moment and stopped in my tracks.

Now, I'll admit that I didn't sleep much over the weekend and my brain is a bit fried. But, can someone idiot check me here:

Google have previously specifically stated that data from Chrome does not affect the organic search algo.
Same for Google Analytics.
Same again for search toolbar if anyone uses that still.

Either:
A.  they are lying toe-rags*
B.  I'm being thick and missing something obvious
C. They don't have a scooby how much direct traffic a site gets
D.  More than one of the above. 

Likewise for other clickstream data (which calls it question some things about 'traffic giving links'). 




*OK, they are lying toe-rags. However I usually work on the assumption that if they explicitly state a fact in answer to question then the answer they give will usually at least be technically true .  It's the gaps and the technicalities that are the killer.

>does not affect the organic search algo.

What about other algos that are used in conjunction with the organic search algo....

>>if they explicitly state a fact in answer to question then the answer they give will usually at least be technically true

I think they've gone too far in their statements on Chrome to back off without litigation. I don't think they would be so explicit if direct traffic phoned home.

But how might they know about direct traffic on some level, albeit not as completely as they know about search traffic

- Google is an ISP - http://www.google.com/tisp/ , https://fiber.google.com/about/

- Google is a CDN - https://developers.google.com/speed/libraries/

- Google has top destinations on the web and knows about a lot of traffic that doesn't come through search - http://youtube.com

- Google has arrangements with ISPs to provide data caching and knows about some traffic even if it never makes it all the way back to Google - http://www.wired.com/wiredenterprise/2013/07/google-internet-traffic/

- Google handles all traffic originating at Starbucks... or will soon - http://forums.appleinsider.com/t/158830/google-to-become-official-starbucks-isp-as-critics-claim-net-neutrality-backpedal

- Google knows about every impression on every page that serves AdSense. I don't know if they can tell whether or not there's a referrer or not, but they certainly know if they user came from search.

So do I think you're crazy to think that Google knows a lot about direct traffic? Yeah. You're crazy ;-)

I'd say that combined data is probably good enough to give a robust representative sample.

Some interesting stuff there. I love breaking out the ol' tinfoil hat from time to time. 

Still not seeing any dataset or combination of that would give them an indication for any site that they choose to look at.  Starbucks would presumably add a lot of breadth to that data, but there are definitely types of site you would and wouldn't access from starbucks.

Their CDN is probably used widely enough to give a good baseline across sectors. Still not enough to apply a metric to all/most sites though. 

brand keyword search may be a good proxy for direct to site traffic

<speculation confirmed = "no" proof = "none">

It's a logical trick - that connection is one step removed so technically they do not take that data directly from Google Analytics. If, however, a site is getting a certain amount of direct traffic it could be construed as brand signal. A brand signal, in its turn, does affect a site's significance in a vertical (undocumented/publicly unannounced Panda-like questions used for internal evaluations of possible consequences of manual and algorithmic penalties: if this site disappears from these SERPs will anyone notice? If the users search for this query do they expect to see this site there? If this site does not rank for this query does it make Google's SERPs look bad? If this site is not on page one what is it replaced with? How likely is the absence of this site in these SERPs to cause any kind of legal action against Google?). So there you go, technically they are not lying but in practice you know what it means. I've always believed if a site is getting good volumes of direct traffic / traffic from sources other than Google it needs to be shown to google, basically stuck in their face.

</speculation>

That said, I guess if you set out to catch discrepancies (mildly put) in their answers you'll have no problem coming up with quite a few of them.

Adding to ergo's list, Google's public DNS is used a lot. 130 billion requests a day according to wiki. Not necessarily direct traffic, but after filtering the ones G knows about... what's left.

um. they are of course lying toerags. But if you put analytics code on a site, they would know what direct traffic the site gets, they don't need to lie about anything to get that data because you tell them. I'm assuming it was Google analytics you were looking at?

and Chrome phones home or I'll eat my tinfoil hat. I think when I read the last google statement re chrome they carefully phrased it to relate to personally identifiable data or something. So you probably don't get adwords in your gmail based on sites you visited, but I bet the behavioural targeting team gets juicy data off of it.

Cutts was questioned on his recently at an event. Can't find the post now though.  The answer given was that they don't use analytics at all in natural search.   

Older one here... Although the last line s open to interpretation a bit http://bit.ly/1dqiq3G

I don't think they'd be stupid enough to issue a straight up denial that was a lie. The fallout if they got caught at that scale would be unreal.