5 ways to turn big data into better government

She was there to help mitigate the disaster that was Healthcare.gov. Now Mina Hsiang is trying to help government “extend democracy” with data — and she says the “easy stuff” can bring in the lion’s share of improvement.

Hsiang, the health data advisor at the U.S. Digital Service, offered a quintet of tips to a standing-room-only crowd at FedScoop’s Data Innovation Summit 2015 on May 7.

1. Deploy analytics to watch how users actually use your sites.

“In the early days of Healthcare.gov, the ‘user funnel’ was an exciting, useful way to see where the pipes were clogged,” Hsiang said.

By using a user funnel to essentially follow users as they branch and navigate through your site, you can see exactly how customers are interacting with your product – whether or not it matches with your plans or users’ expressed desires.

“Design like you’re right, listen like you’re wrong,” Hsiang advised, saying the important thing with users is to “see what I do, not what I say I want.”

According to conference organizers, the crowd was 400-strong and roughly 70 percent of the attendees were from government. It was a stark demonstration of room for adaptation, then, when Hsiang asked how many people used back-end data analytics on customer-facing websites. Only “half a dozen hands” signaled that they were using back-end analytics.

2. Target, target, target.

Whenever they’re rolling out new initiatives or merely executing existing missions, agencies need to ask themselves up front, “Who is eligible?” and, “Who is most likely to respond?” Hsiang said.

Too often, programs or surveys are mired down with responses from citizens who were never eligible in the first place, and outreach dollars spent attracting non-eligibles are wasted.

“You cannot overestimate how much further your money will go” if outreach dollars are targeted precisely at the right populations, she added.

3. Use data and research to design public policy.

Hsiang touted the profound impact big data could have informing public policy proposals.

In a later presentation, the Commerce Department’s Deputy Chief Data Officer Lynn Overmann echoed Hsiang’s advice, noting how crucial the accurate demographic and police incident data from Baltimore and other communities are to informing the debate around – and potential policy proposals for addressing – concerns about policing.

4. Make the complex easy for users.

Deploy “collaborative filters” in the way that private companies do so that users going through government procedures -- say, a name change -- can benefit from the experience of the crowd and see a list of “suggested forms” pop up in the same way that Amazon shows “Customers Who Bought This Item Also Bought” boxes under each product, Hsiang urged.

5. Make people’s lives easier.

Agencies must “toe a fine line between creepy and incredibly convenient,” Hsiang said.

It’s a thought the Federal Communications Commission CIO David Bray echoed later in the day. Government and the public need to have a conversation centered around the question, “At what level do we want to retain checks and balances and at what level do we want to make your lives easier?,” Bray said, citing concerns about government tracking and the fact that, had the Founding Fathers been around to see the growth of big data, they’d probably be skeptical of too much data collection, no matter the benefit.

Hsiang proposed a fix from the private sector to alleviate privacy concerns: Openly display the methodology used to offer users suggestions, in the same way LinkedIn has begun showing users the reasons it suggests certain people as possible connections.

Overall, Hsiang said, big data offers simple solutions to the vast majority of customer complaints.

While highly complex machine learning tools can add even more sophistication to the user experience, Hsiang said: “Eighty percent of the gains [to customer satisfaction] come from the easy stuff.”