Demographics in Web Analytics

One thing that's often elusive in doing sites analysis is weeding out the demographics of visitors. There have been natural limitations here as analytics technologies have always had to collect no personally identifiable information, and in connection with this they've also obviously had to keep their privacy policies clear. Vendors like Omniture, Coremetrics, Websidestory and others must adhere to this basic rule, like just about anyone doing business online.

This has meant Web Analysts have been left to things like surveys, offline data sources like call center data, and any associated indirect and/or circumstantial evidence to draw conclusions from. This can get complicated fast, so often marketers opt to keep it simple at least at first. Example:

OK, my top referrer is Myspace, so chances are good most of my visitors are Millennials.

Aside from accuracy considerations, a problem with this though is the results aren't always going to be actionable. The data will come out something that had already been treated as a given...

Well duh, I'm selling ringtones. Tell me something I don't know, like whether we're talking mostly high school or mostly college students...

One fun, free and casual tool that's out now to perhaps help connect the dots is Quantcast. Some of the data it returns on some sites I've worked on looks believable. On low-traffic sites at least though, I'm finding it's way inaccurate, presumably due to limited data: I also checked a well-aged domain held by a popular club here in the city, managed by a few folks I know. Quantcast says it's got roughly 1000 visitors a month, which is of course almost nothing and sounds about right. However they also said an overwhelming majority of visitors to the site are Hispanic, also holding graduate degrees. I do promotional networking at the venue itself on occasion. To my knowledge its average patronage is almost all early-mid 20-somethings (normally a bit young to have finished grad school), moreover pretty Caucasian; perhaps leaning slightly on the WASPy side... So there are some pretty big differences here between who Quantcast says the club's Web audience is, and what I know their meatspace audience to be. It doesn't add up.

I've started connecting a few sites to the tool for quantification, to see how it does with queries regarding others I've background on already. At the very least, on traffic numbers we'll see how it holds up against the reporting sources already in production. They're still aggregating information from their integrated partners, and I'd half-expect them to not come up with demographics at all on some domains even if they weren't still building out their records, but here's crossing fingers and noting a new option for the toolkit in any case.