October 31, 2013
New Web, Mobile Tracking Methods Being Tested By Facebook
redOrbit Staff & Wire Reports - Your Universe Online
Social networking titan Facebook is testing new Web and mobile tracking methods that would drastically expand the breadth of data it collects from members in hopes of better understand user behavior, The Wall Street Journal reported on Wednesday.
Ken Rudin, Facebook’s analytics chief, said the social network may start collecting data about user interaction with various types of content, such as how long members hover their cursor over an advertisement and whether a user is viewing parts of their news feed on a smartphone.
Ultimately, Facebook seeks to enhance its existing set of demographic and behavioral data on members, and would use the additional information to improve its products, Rudin told WSJ during an interview on Tuesday.
Facebook currently collects two kinds of data from its members – demographic and behavioral. The demographic data captures a user’s life outside social media, while the behavioral data, such as a person’s “likes” and “friends,” is collected in real time on the network itself.
The latest tests would greatly expand the behavioral data that is collected, and are part of an ongoing comprehensive technology testing program, although the company should know within months whether it makes sense to incorporate the new data collection into the business, Rudin explained.
“It is a never-ending phase. I can’t promise that it will roll out. We probably will know in a couple of months,” Rudin told the Journal.
As the head of analytics, Mr. Rudin is preparing the company’s infrastructure for a massive increase in the volume of its data.
Facebook uses a modified version of the open-source Hadoop distributed file system, which stores large amounts of data on clusters of inexpensive machines, to manage its data. There are additional software layers on top of Hadoop, which rank the value of data and make sure it is accessible, Rudin said.
The social network also designs its own hardware to store its vast data analytics warehouse, which has grown to 300 petabytes – a 4,000 fold increase during the last four years.
The data in Facebook’s analytics warehouse, which is used in the targeting of advertising, has not been disclosed, and is separate from the company’s user data.
As Facebook moves forward in capturing a growing amount of data to help marketers better target their advertising, the company needs to make sure that data is accessible.
“Instead of a warehouse of data, you can end up with a junkyard of data,” said Mr. Rudin.
He told WSJ that he has led a project to index the data by creating something akin to an internal search engine for the analytics warehouse.
While some Facebook members may be uncomfortable with the idea of the company monitoring their on-screen behavior, the company stressed that it is using the additional information it gathers to improve its products, not for targeted advertising.
“Like most websites, we run numerous tests at any given time to ensure that we're creating the best experience possible for people on Facebook,” a Facebook spokesperson said on Wednesday in an interview with Cnet.
“These experiments look at aggregate trends of how people interact with the site to inform future product decisions. We do not share this information with anyone outside of Facebook and we are not using this information to target ads," said the spokesperson.