msn adcenter demographic prediction for donationcoder... 67% female?

[mouser]

stumbled on a reference to msn adcenter demographic prediction at aeropause.com
very interesting results for donationcoder.com..... any theories?

try it out on your own urls: http://adlab.microsoft.com/DPUI/DPUI.aspx

You can use adCenter technology to predict a customer’s age,
gender, and other demographic information according to his or her
online behavior—that is, from search queries and webpage views.

http://adlab.microsoft.com/DPUI/DPUI.aspx

[Carol Haynes]

Probably my level of posting ;-)

Actually I think it is totally meaningless ...

My websites come out at 65%/54% male.

Given the 65% male one is mostly pictures of birds, flowers and some photos from a Photoshop course I took (not to mention my rivetting minuet), and the other is mostly family photos what does that say about the sites?

You might argue that they don't get enough traffic (which is true - mainly because they are very boring and haven't been updated in ages from lack of time/enthusiasm after a recent sortie into experimenting with CMSs) but quite how they are calculating these results I can't imagine.

It is quite gratifying that playoutdoors.co.uk comes out in the 23-34 age range ... I must be young at heart (my body is certainly falling apart rapidly  )

It is interesting if you type your name in too ... apparently I am 38% male ... I'm now off to the bathroom to find out which bits ...

[housetier]

On a part of my site its predictions are very close to the truth: 98% male

On another part its predictions are different; but I cannot verify it: I only know a few of the members personally. Time shall tell...

[jgpaiva]

AhAh.. How about this one?

[housetier]

100% female users for google...

It means google managed to appeal to my feminine side 

[Carol Haynes]

The strange thing is that it seems to give a 30-70 to 40-60 split for most sites even when the sites are obviously very specific - and its age range seems very random.

How on earth can it tell from 'site visits' the age of the visitors??

Here are a couple of exmples of very 'female' sites - with an interested male minority (and interesting age range of client):

http://www.needlewoman.co.uk/  Female Oriented 61% 25~34
http://www.menopausematters.co.uk/ Female Oriented 72% 25~34

I can't think of any 24 year old likely to be thinking about menopause !!! And why are 28% of young men interested?

Oh, and by the way Yahoo is 92% female ...

[housetier]

MS has too little sample data, they admit that.

Somehow I am reminded of the (German) saying "don't trust statistics that you didn't forge yourself" (something like that). Anyway, I actually wish they were more accurate...

[jgpaiva]

This is idiot...



How about that site itself?

[app103]

I believe the one for askmen.com...I know a lot of women that like that site. 

The results for my blog:

[Carol Haynes]

Hey - nice to see you appealing to teenage men ... good on yer girl!

[jgpaiva]

Sorry for digging up an old topic.

I just bumped into this, and thought I'd see how things have changed in one year.
Here's the way DonationCoder is rated now:

Just compare it to the way it was one year ago:


Now I think they just calculate random numbers, based on the count of 'c' characters in the page's code 

[edit] Ok, it's confirmed, it just gives a random number, based on the text on the page. I just tried getting the values of my idiotic page, and then changing "text" to "test", and the values changed dramaticly after the one-letter change..  [/edit]

[mouser]

funny.

ps. offtopic:
if you saw a certificate alert pop up when viewing this thread, it was because jgpaiva linked to an image in another thread and used an https link, which triggers the secure protocol for grabbing the file, and causes firefox to warn that the secure certificate for the server is not signed by a known source.  nothing to be concerned about.

[Deozaan]

Ok, it's confirmed, it just gives a random number, based on the text on the page. I just tried getting the values of my idiotic page, and then changing "text" to "test", and the values changed dramaticly after the one-letter change.. 

Sounds like it should be adlib.microsoft.com.