I had the same reaction as this when reading the New York Times hearting of MySpace’s internal targeting technology group: they’re focusing on the wrong thing.
From the article: “A 100-employee team inside the Fox Interactive Media offices in Beverly Hills, called the “monetization technology group,” has designed computer algorithms to scour MySpace pages. In the first phase of the program, which the company calls “interest-based targeting,” the algorithms assigned members to one of 10 categories that represents their primary interest, like sports, fashion, finance, video games, autos and health.
The algorithms make their judgments partly on certain keywords in the profile.”
And on the cusp of bleeding edge innovation, this is what the company has coming up next:
“For the last two months, Fox Interactive has also experimented with the second phase of its targeting program, called “hyper targeting,” in which it further divides the 10 enthusiast categories into hundreds of subcategories. For example, sports fans are divided into subgroups like basketball, college football and skiing, while film enthusiasts are further classified by their interest in genres like comedies, dramas and independent films, and even particular actors and actresses.”
More categories!
If Google’s contextual targeting can’t work with the largest marketplace of advertisers, MySpace’s internal efforts targeting on the same parameters won’t either. A profile page is too small a sample of digital exhaust to base any interest on, especially for perishable interests like travel or auto (i.e. you travel 3 times a year, buy a car once every three years etc.)
If they said they were targeting based upon search histories that would be different.
Also, an 80% improvement is not as high a benchmark as it may seem. The simple fact of being around for a year under News Corp and signing up more advertisers and advertisers having enough time to measure their campaigns to know that MySpace inventory is worth $0.02 CPM rather than $0.01 CPM would be larger factors than ‘categories’, I would wager.
It just surprises me they are trying to re-invent the contextual wheel on a property that clearly needs different thinking.
