Fast failure could lead to big-data success
- By Frank Konkel
- Jan 30, 2013
Failing fast in big data initiatives is a good way for agencies to avoid failing big, according to Chris Biow, Federal CTO and vice president of MarkLogic Corporation.
Biow, speaking Jan. 30 at a big data symposium hosted by the Defense Strategies Institute, said that despite big data’s increasing popularity as a buzzword, it remains a field of high-risk investments where big agency missteps can bleed IT budgets dry.
Instead, he advised, agencies should try a series of smaller big data investments – with some of those investments doomed to fail – in an effort to find solutions that do work, and then expand upon them.
"These are still high-risk investments, we should expect these projects to fail," Biow said. "It’s about spending your bets. You don’t want to spend them all in one place. Split it out, and let 1,000 flowers bloom."
In the quest for agencies looking to return value from big data projects, Biow said agencies should evaluate progress rapidly, within "weeks or months," and decide quickly whether a project is working or not.
For example, if an intelligence agency cannot pull information that might be useful to an intelligence analyst from a big data prototype after a few months, they are unlikely to get much out of it in the long haul, Biow said.
"If they aren’t returning mission value within weeks to short months, they should be shutting it down," Biow said.
As budgets shrink, Bio said the "fail fast" logic makes more sense for agencies that do not have as much money to throw at IT initiatives as they used to. It also enhances agility, with agencies less tied to projects that might not work.
Biow's argument has parallels in the book "Little Bets," by Peter Sims. In a 2011 interview with FCW, Sims outlined his case.
"Making little bets is the antidote to risk aversion. That’s the beauty of the concept. No one likes to lose anything," Sims said then. "People are twice as likely to want to avoid a loss as they are to want to get a gain. A little bet is a low-risk, affordable way of learning something. So long as the bets are little, affordable, essentially risk free, that gives people the ability to take what the researchers would call an affordable loss."
Frank Konkel is a staff writer covering big data, mobile, open government and a range of science/technology issues. Connect with him on Twitter at @Frank_Konkel.