TWCL

[smbhax.com]

Huh! That really surprises me.

Quantcast does it. Here's mine, for instance: http://www.quantcast.com/smbhax.com
_________________

[Clint Wolf]

smbhax.com wrote:

Huh! That really surprises me. Quantcast does it. Here's mine, for instance: http://www.quantcast.com/smbhax.com

I can't for the life of me figure out where they're coming up with those statistics from, since it's not based on accounts with the service. I feel like they're about as accurate as the "People Directories" where you pay to find out about someone (I once tried a trial version of one of those for the hell of it and found out I was still living with my parents, amongst other things that were news to me).

I understand how Facebook pages can have a male/female demographic since it's only measuring internal Facebook traffic and most people's profiles have a chosen gender. But outside of that? Even if they're getting male/female data from ISP accounts, how does that work for something like a household of four (daddy pays, but everyone plays), much less a college campus?

Honestly, I still think the real reason behind Google+ is so that Google can finally get hold of that demographic data like man/woman they'd otherwise have to guess on by having the users supply it. Remember what a tizzy fit they threw over people using Internet names for their accounts?
_________________

[smbhax.com]

I suppose gender is probably inferred from the user's browsing habits--ie, Quantcast has a record of your IP address across the various sites (like mine) that use its tracking cookie, and if you primarily visit certain types of sites, it guesses you're a certain gender.

In which case its predictions are kind of self-fulfilling, but if they can cast a wide enough web--and I think Quantcast probably does--then the predictions can be reasonably accurate, at least for marketing purposes.

I'm almost absolutely certain Google does the same type of thing--and Google has a greater reach than Quantcast--thus my surprise that they don't show it to you in Analytics.

I got into Quantcast when I found that an ad network I wanted to sign up with used Quantcast's report of my site traffic as pretty much the main benchmark for admittance to their program; my traffic was too low for Quantcast to have formed their own estimate, so I caved and installed their tracking cookie on the site (and thus am now "directly quantified"). The ad network then accepted me ("your traffic seems to have grown!" :P) and apparently used the Quantcast demographic data in assigning me to a general ad campaign category ("hardcore gaming" :P), but I've kept using Quantcast's cookie because it does seem to give some pretty useful data, at least as something with which to estimate your own performance over time, or to compare with other sites.

One of the webcomics I respect most uses it (Bad Machinery), for instance, and I've started to notice a lot of other pretty well known sites do as well. Furthermore, I think it's probably fairly reasonable to assume that some prospective advertisers wondering if they should bid directly on your site's Google or whatever (especially the non-Google ones, I suppose, since Google's bidding interface--which I have not seen--probably already provides a fair amount of demographic or at least general market data to prospective bidders) ad boxes will look you up on Quantcast, among other tools, to see if you fit their desired demographic.

The page count and unique visitor numbers I get from it tally very closely with the ones I get (much less conveniently) from my web logs--far closer than Project Wonderful's stats, for instance, although I still glance at those during the day since they update near-real-time, whereas Quantcast's only updates daily.
_________________

[SuitCase]

I don't think Analytics contains any fuzzy numbers at all. Things like gender and general interest categories aren't really counted or represented in the interface at all — Analytics is for straight, by-the-numbers kind of stats like pageviews, region, time on site, etc.
_________________

Bittersweet Candy Bowl, a comic of love, drama and cats.
Updates 3 times a week!

[smbhax.com]

I guess that explains how an outfit like Quantcast stays in business, then!
_________________

[wendyw]

Clint Wolf wrote:

I can't for the life of me figure out where they're coming up with those statistics from, since it's not based on accounts with the service. I feel like they're about as accurate as the "People Directories" where you pay to find out about someone (I once tried a trial version of one of those for the hell of it and found out I was still living with my parents, amongst other things that were news to me).

I was amused to find a Wendy Wood listed on one of those sites with the profession of "guest cartoonist"*. That one was almost certainly me, but the idea that you can draw freebies for other people to put on their website as your job, and even more that I could, made me laugh. Not laugh a lot, just a little bit.

Anyway, I'm rather cynical about things and Quantcast have a lot on their website about how they generate their results, but none of it actually seems to explain any of it. It all seems very vague and I can understand a certain amount of vagueness, because who wants the competition to work out exactly what you're doing and how you're doing it?

The thing is though, I can't seem to find any explanation of the methods. I can find what they're setting out to do and what some of the potential problems are, but not a single piece of information that convinces me that they've actually cracked it.

One thing that would help to convince me personally is some evidence of comparative studies, taking their own results, information gathered via an on site survey or site registrations and maybe more traditional surveys in order to look for correlation. That wouldn't require giving away any secrets or getting too technical.

Another thing that I would personally be interested to know is how they handle regional differences in behaviour. Assuming that they do have a working, reliable system what adjustments have they taken into account for cultural differences?

Of course some of this may be covered and I just can't find it at the moment. If anyone else can, then please point me in the right direction. If not, well, then they're just giving us a blank box and telling us to trust them on exactly what's going on in there.

Like I said, I'm a bit cynical about these things.

(*It might have actually been "guest artist" though. I can't remember for sure, but that was the meaning of the title.)
_________________

Blog