Archive for April 22nd, 2008
A lot of people constantly harp on the fact that Google is not the market leader in any other market segment than search and contextual advertising and hold it up as proof of the fact that Google is a one-trick-pony. While it is desirable for Google to lead the market in everything it gets into, it is not the only factor that Google is looking for when it kicks off a new product.
To understand why Google does things differently, you need to first understand how goes Google work differently as a company.
At its core, Google is one massive computing infrastructure. What the company excels is in building, and maintaining applications on top of this infrastructure, only parts of which are known to us as Big Table, Google File System and Map Reduce. Almost every application (yes, this is speculation, sue me) is built atop this infrastructure, giving Google the ability to have consistency across storage, classification, categorization of any data that comes into its system. Other companies, like Yahoo! and Microsoft, have years and years of legacy sitting on different frameworks and infrastructure, giving Google amazing leverage over them.
For Google, the only real product is user experience and the value the user derives from using Google’s products. This, in turn, helps further refine and better offerings across the plate for Google, creating an endlessly iterative and self-improving product ecosystem. And the products by themselves are a means to bettering the end-product of user experience.
For instance, not many would have much to say about Google’s “web history”, but not many know that the same is used to do drive recommendations in Google Reader, which also uses geographical data (I was recommended feeds related to Trivandrum after being there for a week) which Google collects in conjunction with ISPs (driven by Google Analytics) to further refine these recommendations.
In a similar manner, Google already tracks the clicks that originate from Gmail and I would not be surprised if they are already tracking and indexing the thousands of billions of messages that flow across Google Talk to better know and predict which link you are likely to click more on Tuesdays and Wednesdays (match data from the messages and your web history), compared to Friday or Sunday. And that is very much in line with their mission statement of being more useful to you, in a manner that borders on the eerie quite a few times.
And that is where the greatest challenge lies for companies that aim to compete with Google. Learning systems that improve itself iteratively with time and usage are hardest to beat, because it improves by using you against yourself (something like going against your best time in a racing game than against a pre-programmed computer run) and since Google has been around for such a long time, the amount of data it has about you is something that the competition can’t match unless a vast majority of Google’s users switch overnight to the competing services.
Which brings us back to the non-search problem. Google really does not need to be the number one in other areas (other than the silly acquisitions like Jaiku). It does not cost Google much to create new products (many Google projects like Reader and News were started as 20% time projects) and it does not cost them anything to run those either (they are written with the same framework that is maintained for their core offerings). So, even if all of them were to fail, it would not make a dent on Google, while the fact is that a lot of them don’t.
Now, add Google App Engine to the mix, which opens up the same infrastructure (leveraging the same Google Accounts identity system) to the wider web. With the App Engine, for the tiny cost of supporting the bootstrap process for free, Google now gets even more focussed and specific data regarding usage(in the hierarchy of usage quality, context is king. Apps would have a context that is locked-down taking out the guesswork for Google and the data that is stored in such contexts would also be in a format that Google natively understands).
It would really be stupid to assume that all these processes and data collection is not already being used to improve the advertising business, which is from where they earn their bread.
p.s: This post has been edited for clarity and a couple of grammatical snafus from its first version.