I’m here with Don Dodge, he works here at Microsoft now, but he used to be director of Engineering at AltaVista.
He was also VP of product development at Napster and VP of product development at Groove. He’s a new Microsoftie. Welcome!
Anyway, we were talking about search. He knows a bit about it.
I was telling him that the next big thing in search is contextual search. More on that in a minute.
Don tells me that back in the 90s Natural Language Processing was all the rage. They were hoping customers would put in more words to get better results. In reality, though, that didn’t happen. He said that 85% of all searches are one word.
At AltaVista, he says, they had a thing called Clustering. They’d look at searches around a single word and predict more results. Problem is that took processing time which slowed down the results. Turns out that speed of page mattered — although Don later learned it only matters if response times go higher than a second or two.
Talking with Don you get a sense of the mistakes that AltaVista made too. They were owned by companies that didn’t understand (or care for) search. They thought the big business was going to be in portals, not in building fast, small, services that did one thing well.
He says the secret to Google’s success was two things: its architecture (small PCs that can easily be expanded machine by machine — Altavista ran on big centralized mainframes that couldn’t be scaled out easily) and its linking algorithm.
Anyway, I’ve been reading his blog and he has some really interesting posts for entrepreneurs and Venture Capitalists since he’s been through five startups.
Robert Scoble is the founder of the Scobleizer blog. He works as PodTech.net’s Vice President of Media Development.
Go to Scobleizer …