First, admit you have a problem
Greg Godbout's advice to agencies confronted with data-based evidence of a project's failure? Admit you messed up and quickly change direction.
(Image: Robert Adrian Hillman / Shutterstock)
Sometimes the bad news can be more important than the good news – if only the message is heeded.
Greg Godbout, the EPA’s CTO, founder of 18F and owner of the Arlington Cinema and Drafthouse, called on his fellow government honchos to show “adaptive leadership” when it comes to data and decision-making.
He used his own failure to demonstrate data’s power.
Godbout told the audience at AFCEA Bethesda’s June 30 Data Symposium how his cinema suffered two negative quarters at the end of 2014, and he was caught off-guard.
“Our customers changed their buying habits, and I missed it,” Godbout said. “The data proved, without a doubt, that I suck.”
By pouring through the data, Godbout was eventually able to see that changing Arlington, Va., neighborhoods had brought in customers who cared less about price than they did quality and entertainment, and so, swallowing his pride, Godbout slashed budget offerings and focused on improving the quality of the entertainment he booked for his venue (which hosts musicians and other acts in addition to movies).
“We fundamentally changed the direction of the whole business based on that one day [of data analysis],” Godbout said.
It pained him to admit mistakes, but the moves helped make the first quarter of 2015 profitable for his business.
“I wonder,” Godbout said, “is the federal government ready for leadership that will say, upon seeing signs of things going wrong, ‘OK, I suck, this was a bad initiative, let’s go in a different direction?’”
He stressed the importance of user-centered design and empirical evidence showing exactly how a project is succeeding or failing.
“User-centered design is actually data-driven metrics,” Godbout said, while stakeholder-centered design boils down to the “hunches” of bigwigs.
“We follow metrics that are silly,” Godbout said of the government, recalling how one agency he worked with based performance reviews on hiring data rather than performance metrics.
“HealthCare.gov was I think a green light on Performance.gov the day it released and failed,” he noted.
“That dashboard, to me, loses credibility.”
The next few years will be characterized by a flood of data scientists coming into government and helping improve processes, Godbout predicted.
They’ll also help validate – or invalidate – their bosses’ hunches.
“It’s great to have a hunch. Hunches are important,” he said. “But we need to then have evidence to back up those hunches.”