SAN DIEGO—More than one billion people log into Facebook each month and they generate a lot of data. Users comment and hit the “like” button 3.2 billion times a day, comprise a network of 125 billion friendships worldwide and upload 300 million photos per day. This is what Namit Raisurana, business intelligence engineer at Facebook, told a standing-room-only crowd at the Tableau Software Customer Conference in San Diego on Nov. 7. “We have a lot of interesting data at Facebook, a huge amount of data,” he said. “So much data that no conventional database can handle it.”
Data has been one of the drivers of Facebook’s success. Its Hadoop cluster sees more than 100 terabytes of compressed data each day. But the company faced a dilemma—its data was largely useless if it couldn’t be analyzed. Historically the company used business intelligence applications, a homegrown reporting tool and the reporting platform MicroStrategy. “Our reporting tool worked great in connecting with internal databases, but to do anything advanced with it, it wasn’t the best solution,” said Raisurana. “We saw an opportunity to look for a data discovery tool to help us understand our data and build analytics on top of it much more quickly.” Facebook also wanted to end reliance on developers to serve up reports and analysis and instead put analytics in the hands of every user.
Enter Tableau. In the short time it’s been deployed at Facebook, there’s been enthusiastic user adoption, said Raisurana. The company quickly built a Tableau Facebook users group. The group members answer questions for each other, building an internal database, said Raisurana. Users can scroll up and down and see questions that have already been asked and answered by other users. That means there is “never a dependency on any of us to answer these questions,” he said.
Facebook also held training sessions. Although Raisurana called Tableau “very intuitive,” users were moving from other tools into the new software, so the sessions were essential. The company also began embedding Tableau’s dashboard into its internal applications. Anywhere you go, said Raisurana, “any visualization you see, you see the Tableau dashboard.”
Yet Facebook faced some challenges with integration. The biggest pain point, said Raisurana, was scalability. Facebook’s data tables can be enormous. Tableau, he said, works “great with 100-200 million rows, but if you start pushing it, you start facing a bunch of issues. We are talking to the Tableau engineering teams to see how we can optimize the data sets we’ve got.” Tableau also lacks a Mac OS X client, a problem for Facebook, as the majority of those using Tableau are also using an Apple computer. “Having to install Windows has been a bit of a pain for us,” said Raisurana. He said they would also like to see Web authoring tools, so it’s not necessary to install the Tableau application on a desktop.
Perhaps because this is Facebook however, with its intensely data-driven culture, there’s been some creative hacking too. “That just means we’ve tried to push the Tableau tools to their limits and if that doesn’t work for us, we build things around it to make it work for us,” said Raisurana. For instance, users built a Tableau metadata hack to enable a richer analysis. “We wanted to track information over several months and now we can do that, and track distinct users, desktop usage, etc. so we can plan our architecture in a much better way, seeing how people are using Tableau and what user groups are doing.”
Facebook also built its own integrated reporting portal. Although they have several reporting applications in use, the idea was to make it more user-friendly. “Instead of trying to figure out which tool they needed to get in to which kind of data, we built this and it brings up MicroStrategy or Tableau so they don’t have to worry about it,” explained Raisurana.
Although Tableau usage is growing and users are happy with it, Facebook continues to be concerned with the scalability issue. Information Week reported in late September that Facebook was looking for next-generation big data tools. It quoted Jay Parikh, vice president of infrastructure engineering at the company saying big data was evolving and new analytics, business intelligence and data visualization tools are needed. “Tableau Software has been wildly successful,” Parikh was quoted as saying, “but it was built before big data tools were even around.”
A Tableau spokeswoman noted that the company offers an iPad app and support for mobile business intelligence applications.
Eilene Zimmerman is a journalist whose work appears in The New York Times, CNNMoney.com, The Christian Science Monitor, Crain’s NY Business, Salon.com and other publications. She lives in San Diego.
Correction, Nov. 9, 2012: In the original version of this story, Facebook’s Namit Raisurana said that he wanted to see an iPad app version of Tableau Software. A Tableau spokeswoman said the company does offer an iPad application. The story has been updated.