I was reminded of the character Cayce Pollard in William Gibson’s novel Pattern Recognition, who had the talent of being able to tell what would be a fashion hit, and what would fail to catch peoples’ eyes. The reality is less dramatic, but this still just focuses upon peoples’ interactions with Yahoo.

Though stuff like a site search can tell you about things that people are looking for, and failing to find. And the other type of information can be used to understand what people are most interested in, what topics are hot, and so on.

Do you stock extra inventory of Wii’s in your stores, when the Wii becomes the most searched for product on your search engine? I would.

Do you take your knowledge of the different interests of different demographics, and build services and content for those? Again, probably. I guess the difficult part is in taking this information, and using it as effectively as possible.

Kidding, this type of info is perfect for dynamic content / ad serviing / email marketing / etc. [soapbox] Marketing on the web is getting so unbelievably laser targeted. [/soapbox]

I do imagine that this is done already on a larger scale than described in the patent. And if it isn’t, Yahoo needs to do more reorganizing than they presently are. Knowing what the hot topics are at any one time can inform choices of what news to show on the front page of the site, what products to display most prominently, what areas to invest more effort towards. That information from the kind of analytics described should be shared internally in a timely and meaningful manner.

SERP clickthroughs and analysis of multiple search sessions, mouseovers, toolbar usage, how and when people use different services, and other ways in which people use Yahoo should be among the tools that many business analysts at Yahoo are compiling and using everyday.

The photo is fairly new – thanks.

The “close with a question” post approach is one that predates Rand, though Rand should get loads of credit and respect for wielding the approach so effectively.

Do I detect a Rand Fiskin close with a question tactic?

And I notice your photo up there now – hi

Back on topic – you would imagine this is already being done in some fashion? It would be fairly trivial given the data that is collected (SERP clickthroughs)?

Rgds