A whole bunch of people at work today asked me if I had seen a recent posting on the Google Blog News Channel: What Search Engines Do Search Engine Companies Use?
That posting summarizes a lot of data from VisitorVille Intelligence, specifically looking at which search companies use which search services.
The thing that everyone seemed to notice was this:
Finally, at Yahoo, 68.9% of employees use Yahoo, but a still-strong 29.8% use Google (compare that to Google’s 100% loyalty). 81.2% use IE (13% some form of Mozilla). Only 4% of Yahoo’s computers run Linux, and that’s only market share stolen from Windows 98, not any of the newer versions.
I see two problems with this.
First off, I'm pretty sure that the data collected is not able to distinguish one employee from the next. I just looked at their data collection page and read this:
There is some data that we do not aggregate. For example, we do not collect information on search terms used. We also do not aggregate information on what specific pages or websites people visit.
Huh? Are search engines not also web sites?
Putting that apparent contradiction aside, there's a big difference between saying "55% of Yahoo employee's searches are on blah" and "55% of Yahoo employees search on blah."
Anyway, the next problem I see is what that 100% Google figure. That seems to imply that Google never bothers to check out the competition. Is that really true? I know for a fact that it's not.
So where are all those hits counted? They aren't I guess. Maybe they're only allowed to browse Yahoo from home? Or they are more careful to mask their activities? (Not hard to do with a proxy.)
Personally, I run most of my searches on at least two search engines (the two that matter) so I can see how they differ. But maybe I'm just weird. I'm not sure why anyone working for a search company would do such a crazy thing.
I guess what I'm trying to say is that I don't have much faith in those numbers.
Do you?
Posted by jzawodn at January 05, 2005 11:03 PM
Well, VisitorVille compiles the data through visitors to the websites who contract it for user data. That means all of the data is based on sites visited through search engines. In turn, that means that only successful searches are counted, not test searches, because test searches rarely result in a click through. This means that Google employees who search for websites all use Google, and they only use other engines for testing, while Yahoo employees do use Google to find stuff.
There are TWO search engines that matter? What's the second, besides Google? I can't think of one.
Only 4% of Yahoo's computers run Linux? Wow, all those servers run Windows 98? Unbelievable.
Heh,
I know at least two properties that have scripts that frequently check google for placement and ranking. I wonder if that study also was smart enough to eliminate LWP user agents? ;)
Yeah I wonder the same thing about the numbers. I'm surprised to see Yahoo report a low usage of Yahoo though (68.9), I find that almost hard to believe.
"There is some data that we do not aggregate. For example, we do not collect information on search terms used. We also do not aggregate information on what specific pages or websites people visit.
Huh? Are search engines not also web sites?"
There seems to be a misunderstanding above. What we say is that we do not make publicly available the search terms that people use ("sex", "drugs", "rock n roll", etc.); nor do we make available the specific pages that people visit (such as "45% of Google people visit sex.html" versus 55% of Yahoo people).
Yes, but that actually does make a difference.
If I do a "whole site search" on http://www.php.net, that uses Google to search their site. In effect, that means I would be doing a google query and using the result even if I had no idea when I created it. Likewise if I were to do a search on CNN (although at least there, it's pretty clear that Yahoo is the back-end search engine). There's a fair number of submarine google searches out there that various sites perform.
All that this data proves is that Google is more widely used. It doesn't reflect user preference nearly as much as one would assume.
According to those stats, 81.2% Yahoo! employees use IE (on MS Windows). Seems like we have way too many managers ;-)
"Anyway, the next problem I see is what that 100% Google figure. That seems to imply that Google never bothers to check out the competition. Is that really true? I know for a fact that it's not.
So where are all those hits counted? They aren't I guess. Maybe they're only allowed to browse Yahoo from home? Or they are more careful to mask their activities? (Not hard to do with a proxy.)"
That's what I said yesterday in the comments of this article. I just don't see how you can draw conclusions about employee loyalty (or anything else) from this data. For example - look at the MS/Apple browser stats. Do you really think that people at Apple are using Windows because they want to, or are they using it because a function of their job (benchmarking, testing, comparison) requires it?
Looks I need to do more clarification.
The data collected in VisitorVille Intelligence is collected from circa 5,000 websites, each one using VisitorVille.com to track their web traffic. This is a statistical sample, folks. It is meant to be representative. It is also the only service in the world that collects or provides this data. The data is for informational purposes, and is not meant to help you invest in Google or any other company. What it reports are trends, based on the statistical sample. It does not purport to definitely answer questions which are not possible to answer (such as what every single person at Google Inc does online at all times). I hope this helps.
To speak for myself, I have no problem with Visitorville's data, the collection process or how it was reported. I think they do a bang-up job, and the interface is cute to boot.
What I do have problems with is trying to use that data to derive anything of the sort as the original article tried to do. It's faulty since the data simply doesn't provide the level of detail neccessary to draw the sort of conclusions that Mr. Weinberg is reporting.
Guys, as someone who works for a MetaSearch property, and knows statistics technologies VERY well, I can tell you that the data and claims made by Visitorville are BS. It's a marketing ploy to get them some "brand" recognition. No one has ever even heard of this company and if you think Google, Yahoo, and others would allow their employees to report their data directly to them then you are kidding yourselves....lol.
Unless the employees of Google, Yahoo, etc... had software which was installed on their desktop machine which the employee could manually report the difference between a "test" search and a "real" search there is NO other way for VisitorVille to distinguish between that search data. A search is a search, it's that simple.
Mr. Weinberg claims that they can distinguish the difference. If so, then they have reinvented the way user/statistics tracking works and better get into behaviorial marketing really fast because they just reinvented the wheel! The ability to possess technology which can determine the intent of a user's search queries is worth billions of dollars right now.
I challenge VisitorVille to find one head of a search engine company who will verify his claim that they allow their empoloyees to report the difference between their "test" or "real" searches to VisitorVille. This is the only way I can see that their claims are viable.
This whole thing is a farce.
68.9% of employees use Yahoo, but a still-strong 29.8% use Google (compare that to Google’s 100% loyalty).
I'm not sure why no one else has mentioned this yet. Maybe it's because I'm wrong, but I'll let you decide on that.
Yahoo! and Google both have popular search engines, but Yahoo! is not ONLY a search engine company. Everything Google does is based around search. Sure, they have gmail, picasa, and some other stuff. It's probably also a requirement to be a Google fanatic to be hired. Yahoo seems to be a broader company with more than just search going on. It's probably not a requirement to be a Yahoo search fan to be a programmer for Yahoo messenger.
VisitorVille's Mayor: "Looks I need to do more clarification."
Looks like you need to justify your research methodology period! How can you pass off any research and sell it no less, without justifying your research methodology?
"The data is for informational purposes...What it reports are trends, based on the statistical sample. It does not purport to definitely answer questions which are not possible to answer"
Your research opens MORE questions than it answers - what good is that research?
If you want to charge for your information, especially on search - you better disclose how you get your information, your methodology. Because SEO-SEM professionals can all find better ways of research than you.
For everyone's information, I find Enquiro.com to be a good research company. They at least have free white papers with research and sound numbers. They DISCLOSE their methodology.
I've been in statistical marketing research VisitorVille's Mayor, and your research is full of holes it's not funny. By the way, from your explanation of how you gather your research - "The data collected in VisitorVille Intelligence is collected from circa 5,000 websites, each one using VisitorVille.com to track their web traffic" - clearly, you can't even remotely pass this information off as a free sidenote. Your info is completely worthless.
I'm not saying VisitorVille is worthless, but I don't really trust those numbers either. Employees should be required to use and compare competing engines. That means the developers and the receptionists. Yahoo! and Google don't need me to say that, because I'm sure someone at both companies has employee generated reports on the topic.
One possible explanation for the difference - Google is secretive about a lot of their research.
According to Visitorville, they track the IP address and the HTTP Referrer of every hit to the sites of their subscribers. So, if a user got to a site using a search engine, they would know which one (from the referrer) and they would be able to tell which one (from reverse DNS on the user's IP). So they would have no idea of actual search engine usage unless the user actually clicks through to one of the results and that result belongs to one of the companies using their service.
