Big Data: Facebook's next big idea

The Facebook fallacy is that at its core the Social Network is really an advertising company.

If you look at their quarterly report, you see that about 82% of their revenue comes from advertising, so it backs up this assertion. The problem is that "Liking" something does not translate well to "Buying" something.

You may remember that GM recently pulled out as a Facebook advertiser because of this. On the other hand, teaming searches with advertisements so that you can be targeted with advertisements for products you may be interested in buying is a proven approach, and is Google's strength.  

When we think of a traditional web advertising company, the one that comes to mind most is Google. If we compare the Google to Facebook's Price to Earnings (P/E) Ratio, we see the big disconnect.

Quick digression into finance: the P/E ratio is the valuation ratio of a company's current share price compared to its per-share earnings.

Using Google as an example we see it has a ratio, today, of 18.88. This is because Google is currently trading at $636.69 a share and earnings over the last 12 months were $33.72 per share, thus the P/E ratio for Google is  would be 18.88 ($636.69/$33.72).

In June we saw the Google P/E ratio at about 11, while Facebook was trading at 40 times its expected earnings. Thus, the disconnect.

“The daily and stubborn reality for everybody building businesses on the strength of Web advertising is the the value of digital ads decreases every quarter, a consequence of their simultaneous ineffectiveness and efficiency”, according to Michael Wolff.

If Michael Wolff is correct, then Facebook  will need something more than advertising to live up to and overcome the hype. What Facebook needs is a "big idea"'.

Facebook's 955 million users, with a little more than half that many actively using Facebook every day generate a tremendous amount of data. The size of Facebook's data has been estimated at approximately 100 petabytes by Sameet Agarwal, a director of engineering at Facebook quoted in MIT's Technology Review. “Over the last few years we have more than doubled in size every year.”

Much of this data is in a single Hadoop store where, if mined correctly, has the potential to reshape our understanding of human interactions and, ultimately, of how society works. This could be the next big idea for Facebook.

But how to mine it?

With upwards of 100 petabytes of user data, that is growing daily, Facebook has built a data storage system on Hadoop. The Apache Hadoop project was designed to develop open-source software for reliable, scalable, distributed computing.

Hadoop is really a framework for distributing processes of large data sets across many computers. The idea is to be able to scale up from a single server to thousands of distributed servers, delivering a highly scalable and highly available solution to big data problems.

Though the programming model was intended to be simple, it turned out to be quite complex in its evolution, at least for what Facebook needed. If Facebook's 'big idea' was to come from the study of the data, then they would need to adapt Hadoop to the support data science.

Working with folks from Cloudera, Hortonworks and others, Facebook took on the task to build Hive, a data warehouse system for Hadoop. Hive allows the Facebook Data Sciences Team, a group of 12 researchers brought together to apply social science research techniques, to create ad-hoc queries, and complete varied analysis of large datasets stored in Hadoop.

The Hive project provides the Data Science Team with a mechanism to project structure onto this data and query the data using a SQL-like language (HiveQL); thus, allowing them to mine the Facebook data.

“One potential use of Facebook's data storehouse would be to sell insights mined from it. Such information could be the basis for almost any kind of business. Assuming Facebook can do this without upsetting users and regulators, it could be lucrative.”, according to DJ Patil, data scientist in residence at Greylock Partners.

At this juncture, interesting bits of data have come out, but so far, no "big ideas".

What do you think they may be able to find by mining our data?