Seeking Yodas for data analysis

Would Jedi Master Yoda be able to help federal agencies do big data better?

The notion of the popular Star Wars character saying "Do big data or do not, there is no try," or "Fail fast, young Jedi," seems comical. Yet the take-home message from Carahsoft's April 4 Government Big Data forum in Washington D.C. was that it will be individuals and their insights – at least as much as technology – that shape how useful big data is in the future.

It was keynote speaker Kirit Amin, deputy chief information officer of the U.S. Department of Commerce, who invoked Yoda as a metaphor for the master of an art.

"Why can't we hire 12 to 15 Yodas who understand the big picture (of big data) and tell them, 'your mandate is to make this big thing happen?'" Amin asked.

Amin's impromptu proposition and pop-culture reference came in response to an audience question about the need for the right personnel to analyze the mountains of data that grow larger by the day. The question also referenced federal policies that discourage data-sharing between agencies, hindering feds' ability to combine data from different sources to glean larger insights.

In his response, Amin said it would make sense to spread those highly trained, data-focused individuals across the government – be they visionaries or strategists – with end to end systems expertise along with business acumen.

"My proposition is to find these people – they are out there – and their mandate is to set vision and break barriers," Amin said, suggesting that such mandates could come from Capitol Hill, the Office of Management and Budget or even the Oval office.

"Groom them to be the next leaders in the federal government," Amin added.

Paul Brown, a former technical fellow at the National Security Agency and current CEO of the enterprise software company Koverse, piggy-backed on Amin's statements.

The government may have been a late adopter of cloud compared to industry, but Brown argued that it is already leading the way in big data -- and warned that technical innovation alone will not be enough to maximize its potential.

It is still early in the big-data era – it might be 10 to 20 years before we see where it goes, Brown said – but by then youngsters who doodle on iPads and have never seen a rotary phone will be running the big-data show.

A leader's "goals should be to create the right environment so that the kid who is nine years old and doesn't know a world without iPads, when they come to government, it's right for them to innovate," Brown said.

How not to do big data

Aside from personnel, the panel discussions at the forum dwelt at length on questions of how not to do big data, with a variety of experts explaining common mistakes agencies tend to make in their pilot programs. "Big data is no excuse for a (pilot program) taking two to three years," said Chris Biow, federal chief technology officer and vice president of MarkLogic.

Biow said the longer agencies play a waiting game to see if a big data pilot works, the more money they are wasting. A better bet is an agile approach where failures can be quickly detected and adjustments made immediately by agencies -- in other words, it is better to fail fast than fail big.

"What's done wrong is we commit ourselves to two- to three-year processes because we think big data is hard so it probably takes that long," Biow said. "When you do that, you set yourself up for the most expensive software bug that exists, which is your requirements."

Other mistakes agencies tend to make include not catering their big data pilots to specific problems and not outlining returns on investment.

Agencies, Brown said, should go in thinking, "I'm going to solve this specific problem, it's going to have this specific return on investment."

"Have a very specific mission in mind," he added.

Jim Campbell, corporate systems engineer at HP, said potential big-data users in the public sector face scope problems, too.

"In the government sector, you get told you're doing cloud, you're doing big data, you're doing Hadoop,' and you're having this technology chasing questions or problems," Campbell said. "But before you do that, you have to define the scope of what you're trying to define. All other questions fall in place once you've done that."