Sunday, May 15, 2011


I recently ran across a couple of articles about Google. The first was an interview with one of Jaiku's co-founders. The second was an assertion that Google needs to hire people other than engineers. Then there's the common assertion that big companies (such as Google) don't innovate enough, and finally there's the phenomenon that I've always wondered about, which was that Microsoft at its height of influence had the entire valley in fear, while startups in the valley today (and elsewhere) seem to thumb their noses at Google with impunity. One startup I talked to said to me, "Google is incapable of moving fast enough to compete with us even if they wanted to." In this blog post, I'm going to attempt to tie together all of these threads and make them coherent. Feedback, of course is always welcome.

First, it's a myth that big companies do not innovate. One of my favorite books on the topic, How the Mighty Fall, shows that even failing large companies throw more money at innovation, not less. Google's continuing to innovate on multiple fronts in distributed computing, self-driving cars, image processing, and countless other areas that has computer science faculty leaving their tenured jobs to join Google. In fact, if there's one thing that Google is really good at, it's the ability to bring computer science research from academia and make it real in products for millions of people. Google voice actions, for instance, required gobs of data, statistical machine learning, and fast servers to do what it does. The need to do so is now driving Apple to build large data centers despite Apple's notable failure at network and cloud computing. This is an area that Google has a decisive advantage and it must drive Apple nuts. Similarly, Google navigation on the phone requires a huge investment in cars that can crawl the world's streets and send back imagery and image data, coupled with investments in smart routing algorithms, not to mention the ability to stitch together all that data and turn it into maps. I have no doubt that further innovation on the front of real time data processing will enable Google to stay way ahead of the competition.

Then where does Google fail? I think it's not instructive to look at outright failures here, but to look at how Google's approach is completely different from the competition. The most popular feature of Facebook is photos. If you think about Facebook as a photo site with a few other features I think you'll not be far wrong. Why is Facebook photos so popular? It's got crap resolution, not that great a user interface, and is uninteresting. The answer as detailed in The Facebook Effect is tagging. If you look at the act of tagging, there's no real computer science involved: the amount of image processing required is minimal, since the user is the one providing the information about where the faces are. The Google answer here is to spend millions acquiring Neven Vision and then to integrate it into PicasaWeb and Picasa. Not only was this expensive and late (as compared to merely copying Facebook's hacky Face tagging feature), it proved to be nearly useless. Early versions confused people's faces enough that you couldn't trust it to run without a verification (even Google today doesn't let you do this). Further more, the "tagging" didn't copying another important Facebook feature: that of notifying your friends that they were tagged in a photo. Since all that data is locked away in the privacy of one person's account, you couldn't share, improve, and get better. And nobody used the feature. Here's the thing: the guy who did the tagging feature at Facebook probably got lots of recognition for it. Even if some smart engineer decided to simply copy Facebook's feature at Google, it would be very likely that he would be blocked at launch, or that he would simply not be recognized for doing this important work! The concept that a smart hack could be far more important than a computer science breakthrough does not exist at Google!

Once you realize this, several things fall into place. For instance, it explains why PicasaWeb's storage pricing in the early days was insane (it was something like $20 for 6GB per year). While sites like SmugMug, etc., could help defray storage costs by selling photos and revenue sharing with users, copying that feature would not have been an important computer science breakthrough, so Google never did it. While other sites made photographers happy by allowing them to change the background of their photos, Google never did it --- you wouldn't get recognized for doing this. Letting Picasa do something easily useful like stitching together photos automatically wasn't important, because that was a solved problem. This explains why gtags is still a 20% project despite a large number of engineers inside Google depending on it for productivity --- there's insufficient computer science content there for it to get engineers behind it. An alternative project with much more computer science content (and requiring correspondingly much greater resources) was funded and staffed instead.

Orkut, for instance, never got sufficient engineering resources behind the property despite the founders clearly saying that it was an important product for similar reasons. And of course, startups thumb their nose at Google because while most startups do not have the resources to put together GFS, Bigtable, or a major computer science breakthrough, they mostly have no problem coming up with and implementing great applications such as FourSquare (no computer science there), Farmville, or even finding alternate revenue sources for their great photo site. Google, by contrast, isn't hungry enough for that, and at this point, even if Larry Page wanted to change Google's culture to make it capable of recognizing such contributions as being important, there would be too much resistance from the upper ranks of the engineering organization that he probably could not make it happen.

This shows how important corporate culture is to the kind of projects Google should and should not undertake, and my guess is this is why Paul Buchheit made the statement that Google will land on the moon before it beats Facebook. Google certainly has all the engineering and product capability to do social products. Its missing the cultural capability, and that's what matters in this race.
Post a Comment