MSN has demographic records of many of its users. MSN's Passport is the mainstay here. If you have a Hotmail account then you have a MSN Password. It is easy to point out that very many Hotmail registrations are completed with entirely fictitious details and MSN is aware of that. When credit cards are involved, however, then personal and demographic details quickly become more accurate. Users who pay for a Hotmail upgrade or other MSN services (potentially even Xbox Live users in the future) are giving Microsoft's search platform better targeting information.
Microsoft's adCenter Labs presents two types of demographic distribution. General Distribution is the breakdown by age of users based on their searches and sites visited over a period of one month. MSN records which sites users of their search engine visits. Predictive Distribution is a prediction of the next month and is based on either a query or a webpage.
Using the popular news site Digg as case study illustrates the strengths and weaknesses of the model.
The news stories which feature on Digg are submitted to the site by the sites users. The full stories are ones which are found elsewhere, on news sites, blogs or anywhere interesting, and only a short summary of the story itself appears on Digg.com.
Microsoft's adCenter Lab (found here) shows a gender demographic split of 63% male and 37% female. That is a higher confidence towards the female demographic than many of Digg's users may expect but MSN's prediction may not be wrong.
It may be the case that those users who sign up and submit stories to Digg are mainly men - a ratio of nearly 3:1. That would suggest to users on the site that most of the visitors are men. MSN's demographic prediction technology does not separate active contributors from those who simply view the site.
Digg's success has made the news. Founder Kevin Rose appeared on the cover of Business Week. This mainstream exposure in turn leads to a rounder distribution curve of audience demographics. Historically, a greater proportion of Business Week readers are women than at the case study Digg.
The third version of Digg launched this month. In this update news stories submitted to the site did not have to technology based - politics, health and general interest stories can now appear and be dugg alongside the traditional male-geek bias of technology. This change made the site more attractive to a wider audience. There is a snowball effect too. As more general interest stories entered Digg then more general interest stories in Digg could be found by MSN's search engine and MSN search engine users. This month more female searchers are more likely to find Digg than last month.
The confidence ratio passes the believability test after this close inspection. The initial 63/37 male/female split seemed high on the female side at first. On inspection it seems likely that Digg has been drawing a larger female audience lately.
MSN's Predictive Age Distribution shows even larger demographic changes at Digg.com. This month the dominant age groups are the 18~24 range (with a 26.8% share of Digg's audience) and 25~34 range (with 27.2% audience share). Next month, according to MSN, over a fifth of the sites users will be less than 18. The younger than 18 audience share leaps from only 9.8% to 20.97%. That's a statically significant leap.
Could this be true? It is certainly hard to argue that a large number of under-18s read Business Week! It /might/ be. Digg has reached that point of "Great Awareness" along with MySpace and YouTube. All those students and school kids on MySpace and YouTube are certainly more likely to pop over to Digg to see if anything interesting is happening than they are to wade through an online newspaper.
The most important caveat is that MSN is still only testing the adCenter software. Our demographic analysis does use a sample of real audience data but is put forward as proof of concept only.
MSN's adCenter is full of promise but these are early days yet. The technology MSN is able to demonstrate proves that demographic targeting of PPC bids could be used.
It will be the arena of PPC bidding combat and ROI victories which will prove whether the technology should be used.
