Google - beyond the interview

I recently interviewed Eric Schmidt, CEO of Google, for a video series sponsored by British Telecom. You can see it either at BT.

I recently interviewed Eric Schmidt, CEO of Google, for a video series sponsored by British Telecom. You can see it either at (registration required) or at interview iTunes. Or you can read on for my unapproved speculations about Google's future. I think it's clear Google needs to move more aggressively into personalization and also into more structured ways of looking at data/metadata. Note that Eric did not necessarily agree...

Personalized search

On personalization: It's inevitable, even though Google is reluctant to get caught up in the privacy issues around user data, especially after AOL's recent fiasco with user data...for which, not surprisingly, it is now getting sued.

Eric Schmidt

Yet it seems clear to me that in order to make its search better, there's little that Google can do other than to understand what each user is looking for. The fat front of search is has been fairly well covered; the long tail needs knowledge about the searcher. Of course, it does no good simply not to collect personally identifiable information. As AOL demonstrated, people give away a lot by their searches. Better, I think, to take the opposite route, and to work with the users: Tell them that you're trying to make their searches better, and ask them to help you. One way to keep management lean is to give managers too much to do, says Schmidt. Make it easy – really easy – for them to turn tracking on and off, and then rely on their common sense to figure out what they want to reveal and what to hide. By trying to be unobtrusive, services end up feeling sneaky. Imagine a service that said: "This is what we know about you, and how." And that then invited you to edit the metadata. In the end, social engineering will be as least as important as technical engineering to do this right. Some early examples include mSpoke and RootMarket, all of which let the user participate.

But once the users want to play, then the system needs to be good at matching behavior and demographics to what users want. On the user side, huge amounts of data, untouched by human hands, can be helpful in matching behavior - which may include all kinds of data other than previous searches, such as – with permission – surfing habits, the contents of mail or even a user's documents. If the user is writing a report on wine production in Italy, you can guess what he might want when he searches on "robust red."

On the searched content side, vendors can use data mining from users' behavior – where Google and other much-used systems will of course have an edge – as well as carefully constructed ontologies or taxonomies of specific domains such as movies and medical care. (Disclosure: I have investments in companies in both these areas, ChoiceStream and Medstory.)

The search service can also ask the user: Do you want personalized search? Do you want personalized ads? Over time, I expect more and more people will say yes – especially once it's clear what they are giving and what they are getting.

So how Google proceeds I don't know. But I'm sure that the company is doing more than it lets on, and that issues such as privacy are probably more of an issue for them than technology. It may be that Google will wait until behavioral targeting becomes more mainstream, and will then launch its own technology into a more relaxed world, where users expect to be followed, but they feel comfortable because they know what's going on.

Index more

The second area that I think Google must address sooner rather than later is indexing of stuff that is currently not on the Web. Here Eric was a little more forthcoming, mentioning that Google already does images and video...but I was thinking beyond that. First, there's indexing (which requires recognizing) all the things in the videos and images. More on that later, when I do the second half of a posting on pattern-recognition.

But there's also indexing all the objects out in the real world, starting with the inventories of every retailer. Here Eric agreed, though without any dates. How will this happen? Well, somehow one retailer will start making data derived from its RFID logistics system available in an effort to increase sales. Other retailers will follow suit in self-defense, and then pretty soon it will be widespread, just as most airlines now let you see what seats are available online. Those who make you wait until after you have booked will eventually lose out to those who let you know beforehand, and the same will happen with respect to product data: Information is added value.

Where does Google fit in? Well, this time it may do something like Froogle right, collecting data feeds from suppliers and making them available conveniently to users. If Google doesn't, someone else will. But this does seem to be something already on Eric Schmidt's mind.

Of course, you can't forget to ask about China. It was one of the hardest decisions the company has had to make, says Schmidt. It seems to me that it's better for Google to be in China than not, even though it's censored. It's not simply that it can still deliver some information. The habit of finding things out is almost as important as what people find least for now. People need to expect to find the answers to their questions. They should get into the habit of wanting to know things. Later on, they'll begin to wonder why there are certain things they cannot know.

Final tip: One way to keep management lean is to give managers too much to do, says Schmidt – with 40 or so direct reports. That way they're too busy handling real problems to interfere where they are not needed.