1 of 1
Dan Ryan likes to break things. “My mantra here is, ‘I’ll help you fix everything that’s broken and break everything that’s fixed,’” says Ryan. Last year he used that knack for creative destruction to help make the Obama For America website a campaign fundraising powerhouse, and this summer he’s using helping the companies in the Gig Tank start-up accelerator program.
After working as a web developer in Chattanooga for 13 years, Ryan led the front-end development team for the Obama web site, creating the web software that ran on browsers, as opposed to the server software. As Gig Tank technologist in residence, he advises all the companies participating this summer—as well as CO.LAB, which runs the Gig Tank—on infrastructure, scaling and user interface.
“It’s better that we break it and figure out where the failure points are, and correct them before you launch, than to launch it and have your service go down and you lose all your customer base because the one day you have buzz you can’t handle it,” he says.
While a business incubator wants to help companies come into being, he says, “An accelerator wants to get you born or to get you to die quick. Can we get you through this process of launching a business in 90 days, so if it doesn’t work you haven’t committed two years of your life and a million dollars? A lot of that is failing fast.”
On the Obama site, “big data” ruled the roost. Decisions about web design issues, such as how to ask for donations, had immediate, tangible consequences: increasing or decreasing receipts. Those decisions were based on data from extensive testing—which pulls more donations: A color or black-and-white photo of the President? A photo of the President or the First Lady? (Answers: color, and the First Lady)—and on crunching massive amounts of data, rather than extrapolating based on a small sample.
It turns out that testing a theory against data is a lot like failing often. Ryan says his team came to the campaign with a lot of ideas about effective web design. The campaign web site had so many visitors and transactions that for the first time in most of their careers, they had enough data to determine if their ideas were correct.
“We were right about half the time,” he says. “That was a big learning thing, too, coming to grips with the fact that your gut is tossing a coin. There were times I would have bet everything I own in a test that one side was going to lose and it won. After about the third time it happened, I just stopped trying to figure out what they would do.”
One result was that there were very few judgment calls, a rarity in web design. “It’s actually kind of nice,” says Ryan. “The pressure is not on you any more. The pressure is on coming up with good tests to run.”
Now he’s bringing the lessons of big data to Gig Tank companies. For example, some are building services to analyze data for other businesses. He posed an unexpected question to them: Are they using big data on themselves, to test their own assumptions?
“That was a fun conversation,” says Ryan. “I watched their gears spinning. They need to know if they’re getting wins for people.”
This kind of testing means questioning a company’s most fundamental assumptions about itself and its product, which is where some of Silicon Valley’s biggest successes have come from. Twitter began as an internal communications tool for a company whose original product failed. Flickr was initially a minor feature that turned out to be better than the original product. Both only became huge successes because a company was willing to question its assumptions and rethink itself.