Yahoo RSS Search…and Creative Commons?

Standard

Steve Rubel caught a quick glimpse of an apparent Yahoo blog/RSS-focused search offerings. Kinda interesting. Even more interesting? The text from the screenshot that he took:

This Yahoo! Search service finds lorem ipsum doler itum foo bar blah blah blah mktng spiel. While most stuff you find on the web has a full copyright, this search helps you find content published by authors who want you to share or reuse it, under certain conditions [Emphasis mine.]

Sounds to me like they might be checking for feeds that are Creative Commons licensed? Neat.

Microsoft and Claria

Standard

Fred Wilson posted today on the Microsoft/Claria reports that have been floating around. It is (as I seem to say about everything) an interesting post that you should go read, and also (as I seem to say about almost everything) one that raises some questions for me.

The first is very minor…Fred writes:

First of all, its a match made in heaven. Two of the most hated companies in the technology world getting together. Sounds right to me.

I don’t think it really matters, but that statement started me wondering whether Microsoft is really still one of the top “most hated” companies in tech. I would without question give them the title of “most mistrusted,” but that’s a pretty different thing from hate. Personally, the last time I’d say that I actually hated Microsoft was when I was rebooting my Windows 98 machine two or three times a day. Anyway, that was just an odd little bell that went off in my head — the real question is coming up.

Fred also writes:

But Microsoft can take the good stuff that Claria has; its behavioral targeting technology, its relationships with advertisers who understand that behavior is the best targeting technique for many applications, and possibly Claria’s data.

The question here is whether Claria’s behavioral targeting is really all that good. I’ll preface by saying that I have no firsthand knowledge of their technology…the reason for my skepticism here is really based on textual analysis more than anything else. Some of you may want to stop reading now.

A few months ago, a Claria-sponsored article on behavioral targeting appeared on ClickZ. [Read it, please.] The first postulate for my analysis is that Claria would not have had their name linked to an article with which they substantially disagreed.

The article contends that behavioral targeting on the Web should be “reactive” rather than “predictive.” I consider this statement to be absurd. Behavioral targeting systems are necessarily both reactive and predictive: by analyzing a data set (i.e. “reacting” to past behavior), the system determines what content can most appropriately be displayed for the individual being analyzed (i.e. “predicting” what that individual will respond well to).

The only way that I can see this distinction making any amount of sense is if one is trying to promote the idea that old-fashioned statistical techniques don’t apply to the Web. That is to say: one is promoting the idea that behavioral targeting on the Web should be focused on responding quickly to apparent trends that appear in relatively small amounts of recent clickstream data, rather than modelling, profiling, or any of the other approaches that focus on how an individual’s behavior conforms or departs from that of larger groups.

Why would one want to promote that idea? Well, if “behavioral targeting” was quickly becoming one of the buzzwords of the Web, and my technology were geared towards short-term clickstream analysis, I’d sure be pushing the idea that it was time for some new thinking about how behavioral targeting should work.

It makes sense, I think. Back when Claria was Gator, they were all about quick response: replace ad X with ad Y when the user loads a particular site or ad, pop up ad Z when the user visits a certain type of site, and so on. When Benjamin Edelman (Berkman Ctr. for Internet and Society) analyzed the targeting of Gator ads back in 2003, he noted that:

“Testing to date has shown no signs of profile-based targeting of this sort, and tests with different or invalid user IDs and machine IDs yield the same results as tests with the original IDs as set by Gator’s installation software. Nonetheless, Gator’s marketing materials prominently offer a targeting capability (‘… based on past behavior’). This ‘past behavior’ targeting may reflect profile-based targeting not yet detected by the author. Alternatively, it may refer only to display of an ad as a user leaves a targeted web site (or some number of minutes or clicks after a user leaves that site). Future research may attempt to more fully flesh out Gator’s profiling systems by comparing the ads shown to users with divergent browsing histories.”

I’m sure that Gator (sorry, “Claria”) has been working on profile-based targeting in the last couple of years, but I doubt that it’s been their primary focus — it’s difficult and requires a huge amount of time and data, and until recently I don’t think that Claria would have had enough clients asking for sophisticated targeting to make the investment worthwhile.

Claria’s relationships are definitely valuable, as is their data (to a company that’s not afraid to get a little mud on its reputation, at least), but I’m really not sure about the technology. I have a lot more research to do on the folks working in this area, but if it’s high-end behavioral targeting technology and knowledge that Microsoft is looking for, I think they’d be well served to explore some other options.