They Won People Over By A Logical Argument

Via @glinden, I enjoyed this article on why GDrive (an early cloud document/file store) was never launched by Google:

At the time [2008], Google was about to launch a project it had been developing for more than a year, a free cloud-based storage service called GDrive. But Sundar [Pichai] had concluded that it was an artifact of the style of computing that Google was about to usher out the door. He went to Bradley Horowitz, the executive in charge of the project, and said, “I don’t think we need GDrive anymore.” Horowitz asked why not. “Files are so 1990,” said Pichai. “I don’t think we need files anymore.”

Pichai apparently went on to explain in more detail why files are no longer needed.  It has to do with the notion that, in the cloud you just have data and information.  Organizing that information into files is not necessary, especially when you can just start editing that information directly in Google Docs.  I’m going to ignore for a moment the “don’t be evil” ramifications of data portability and lock-in that comes through the dissolution of explicit files — how am I supposed to export my data into the Microsoft Cloud Word or into Open Office or into VisiWord whatever else I’d like to use, if files do not exist?  Instead, I’m going to focus on how this decision was arrived at:

When Pichai first proposed this concept to Google’s top executives at a GPS—no files!—the reaction was, he says, “skeptical.” [Linus] Upson had another characterization: “It was a withering assault.” But eventually they won people over by a logical argument—that it could be done, that it was the cloudlike thing to do, that it was the Google thing to do. That was the end of GDrive: shuttered as a relic of antiquated thinking even before Google released it. The engineers working on it went to the Chrome team.

This is what I find absolutely fascinating.  Here is a company that A/B tests everything in a heavily data driven manner, down which of 41 shades of blue the link anchortext should be.  So you would think that such a momentous decision about killing the whole GDrive project would be data driven.  It was not.  I quote again:

But eventually they won people over by a logical argument—that it could be done, that it was the cloudlike thing to do, that it was the Google thing to do.

Here is an instance where an important decision potentially very large service was made not by the data, but by a HiPPO, the highest-paid person in the room.  Granted, that HiPPO did not just come out and declare his or her omnipotent will.  Reason and logical argumentation were still needed.  But reason and logical argumentation were all that was needed.  Nobody had to go out and “prove the idea with code”, as Silicon Valley loves to say.  Code was written for GDrive, but the code itself did not provide the proof of its own non-release.  And the all-powerful Big Data didn’t even begin to enter into the equation.  What provided the proof was a core logical argument, coupled with a strong vision for the future (“it was the cloudlike thing to do”) with an ounce of emotional appeal (“that it was the Google thing to do”).

This is a very refreshing story and I am heartened and encouraged by it.  The reason this is exciting is that much of the research that I work on, such as iterative relevance feedback and explicit collaboration, is work that does not have an immediate outlet in the consumer search world.  It might take years before the average user is ready to engage with some of these tools and techniques, rather than the typical five-month lifecycle of your average prove-with-code, throw-it-against-the-wall-see-if-it-sticks data-driven feature release.  Furthermore, it takes much longer to develop some of this research, as it is more risky and exploratory, and the market might not be ready for it for a long time.  At the same time, however, if one waits to start developing such technologies until the market is actually ready, then it is already too late.

For example, the common wisdom for over a decade was that users were too lazy or too unwilling to provide explicit relevance judgments on the information or documents with which they are interacting.  So none of these tools were developed.  All of a sudden, the Facebook “Like” button took off, and pretty soon the “+1” button was added.  In complete contradiction and defiance to ten years of “prove it with the data” arguments about users being unwilling to explicitly mark the relevance of their information.

The way around this problem is to be willing to let a HiPPO make a decision — based on logical argument rather than on log data or usage data — thereby clearing the organization to move forward with that decision.  Start working on tools for explicit judgment years ago, and you will be ready with a fantastic solutions once the marketplace catches up.  Are all such HiPPO decisions going to be correct?  Of course not.  But will fewer opportunities be missed, because you are unwilling to use logical argumentation to carve out a bold new vision for the future?  Yes.

Don’t get me wrong; data-driven decision making is very useful.  But it is useful for incremental improvements.  If you want to take big leaps forward, such as the leap Google wanted to take in 2006 with its vision of the cloud, that requires a HiPPO being able to win people over — or being won over — by logical argument.


This entry was posted in General, Information Retrieval Foundations. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *