Facebook isn't the first software company to tackle so-called inbox clutter. Microsoft and Google have each introduced software that helps employees prioritize, or de-emphasize messages in their inbox. But Campos said that such relevancy ranking is a core competency Facebook has cultivated in its consumer product for more than a decade. That will ultimately enable Facebook to make messaging and collaboration much more crisp, and less noisy. That's the idea, anyway. "Employees are just inundated with information, but they also have a fear of missing something that they need to see."
Collecting, organizing and normalizing the data
Harvesting and analyzing this productivity information requires a hefty infrastructure stack with the capability to churn through hundreds of petabytes of data. Facebook uses a number of external and internal open source software tools to collect, organize and normalize, or weed out inconsistencies from the data. These include the Hadoop analytics software, as well as technologies that Facebook built and released to open source, such as its Hive data warehouse software and Presto real-time analytics tool. Campos says one challenge is figuring out which tools to use; some data must be accessed in real-time, while other data is historical, outlining trends occurring over several months. Presto is designed for rapid analysis, while Hadoop can crunch the trend data.
Such infrastructure forms the bedrock of analyses for other business processes. For example, to help connect employees occupying offices in dozens of regions in 27 countries, the IT team incorporated in each conference room sensors that track trends such as bookings and occupancy rates. In these rooms, employees use 1,300 video conferencing systems to meet "face to face" with colleagues around the world. Campos and his team monitor room occupancy and video conferencing consumption via a software dashboard to see if and where they can improve planning for room availability.
Campos, for example, knows that conference rooms are almost always occupied in New York, while rooms in India are busy in the morning and late afternoon. This sort of information helps Facebook determine whether or not to add more rooms. "The more that we understands these patterns, the better solutions we can provide to help everyone communicate," he says. For example, he split large conference rooms that were being used by a small number of employees into several small conference rooms.
Facebook's focus on using data to drive efficiency extends to its headquarters' parking lots, where each spot is fitted with a sensor that relays information about its occupancy. As motorists drive into the parking lot, they will see green lights over each row directing them toward available parking spots. Rows without occupancies are marked by red lights. "We know exactly how much available parking we have at any given time, when it fills up, and how it's trending," he says. This is important because parking comes at a premium; there are only parking spots for 3,600 employees at its Menlo Park, Calif., headquarters.
Sign up for MIS Asia eNewsletters.