Chief data officer, City of Chicago
Brett has moved on from this role since the interview.
Chicago’s Chief Data Officer Brett Goldstein is a pioneer in more ways than one. He’s the nation’s first big city CDO and a force for transparency in what has long been a secretive environment. Under the direction of Mayor Rahm Emanuel, Goldstein helped establish the city’s data portal, data.cityofchicago.org, which includes more than 200 datasets. In prior roles, he worked at the Chicago Police Department and as director of IT at OpenTable, a restaurant reservation service, during its early days. He took time to share some lessons about creating a culture of openness and encourage constituencies and colleagues to share and use data in both the public and private sectors.
What does a chief data officer do?
Well, I’m the first CDO of a major municipal government, so I think we’re finding out. One of the things that came up in the mayor’s campaign was the importance of sharing data with the public. My job is to make that an operational reality.
How do you share data with the public?
We have this portal that we pour a ton of data into—everything from employee salaries to crime to all the 311 calls. Over the last two months, it has really taken off. It speaks to a variety of constituencies. It speaks to my mom, who can find out how much I make or pull up a map and know whether there’s a crime near my house. It also speaks to developers. For example, they pull our street sweeper data to build an app that sends out a notification before the sweeper comes to your street. We’re also helping the media with Freedom of Information Act (FOIA) requests.
And researchers are definitely one of our constituencies, too. A couple of months ago we published the biggest incident-level crime database out there. If you’re in the social sciences, this is a treasure trove. When we released it, someone tweeted, “You’ve finally let me complete my dissertation.”
Are you the voice of data when it comes time to make big decisions?
I go to policy meetings. If we’re discussing a problem, whether it’s vacant buildings or food deserts or things like that, I’m definitely part of that discussion. I also work with various departments throughout the city. They come to me and ask, ‘How can we do things better?’ I provide a framework for how can we get at answers.
What are the challenges in getting answers from the data?
You can twist data to say anything you want. I help make sure we’re using it correctly—but also to show what it’s capable of. Another part of my job is to introduce advanced analytics, prediction, data-mining and modeling to government. I spent three years at the Chicago Police Department focused on shootings and homicides. Now I’m taking these ideas to the city. How do we predict problems? For some people, that sounds really radical. But for a company like IBM or any financial services company, this is standard business. It’s applying mathematics, computer science and statistics to large data sets and asking what are the best decisions we can make.
What types of problems can you predict?
There’s a problem in one part of the city where, whenever the alley lights go out, the garbage cans disappear. You can imagine the model we’ve built. When the lights go out, we’re going to replace them ASAP, because we save $54 for every garbage can that doesn’t disappear. The city is an ecosystem. We have copious amounts of data—permits, financial information, 311 calls, crime—and a lot of it can point you to early warning signs. You just need to understand your data. And that really comes down to pushing computer science into the ecosystem, which is the area where I get most excited.
How does your experience as a CDO compare to what you did at OpenTable?
From a technology perspective, it’s quite different. At OpenTable, I had autonomy and freedom to try new technologies. You inherit architectures coming into large bureaucracies and that’s a challenge. Government has not typically approached IT holistically, so you end up with a hodgepodge. Also, hard disk space was still expensive when I was at OpenTable and things like the cloud didn’t exist. But doing intelligence on the data was also a theme there. I was worried about similar things: how do we make sure every restaurant is connected? We need to sure make the data is secure.
So, data-mining works similarly in the private and public sectors?
Data is data. When you get into data science, it’s not domain specific. I’ve had the same principles about data for years. I adhere to the concept that nothing is random. You just need to know how to listen for it. The core of what I’m doing with spatial analytics has roots in epidemiology. You can take the same principles whether you’re talking about public safety or marketing.
Was it difficult in the beginning getting restaurants to share data?
In 1999 and 2000, we were really pushing a cultural change. We were going into restaurants that had used some sort of big, old-school reservation book for decades and saying, ‘Hey, use a computer instead and enter your entire inventory of seats there and this will help you do business better. Better yet, we’re going to expose your inventory to the Internet.’ But a few years later, we experienced the network effect. At that point, if you weren’t online, you were losing reservations. Fast-forward more than a decade and now I’m talking about predictive analytics, data mining and data modeling in sectors of government where we might be really proud if we push data into a spreadsheet. It’s another case of cultural change.
Do people usually feel threatened by new systems?
If people have been using X for 10 years, and you introduce Y, no matter how much better it is, there’s going to be resistance. But I never argued that we should consider data and nothing else. It’s about making a decision informed by data. One of the techniques we use is to take what we know from data intelligence and find a way to marry that with human intelligence. In other words, what do we know from boots on the ground? When I was with the Chicago Police Department, my field training officer, Rod Gardner, is one of the best cops I had ever met. Human intelligence like his—you just can’t beat it. But police officers move around. They go to specialized units and different beats. If I had ten thousand Rods, maybe I wouldn’t need a model. But I don’t. Rod is going to retire. So, the question becomes, ‘How do you harness the algorithm in his head?’
Any tips about introducing a new system like this?
One of the lessons I’ve learned is that you need to invest the time up front. You need to engage the users and help them understand the benefit. You always find one or two people in a department who have the aptitude. But for 99.9999 percent, this is the most boring thing in the world. I live in a Unix environment. But if I try to hand off a red Unix screen, they’re not going to buy in. It’s about bringing together a system in a visual, friendly way that uses high-performance computing built on top of an architecture of advanced analytics. What we’re doing is as data-driven as you’re going to get in municipal government. It’s cutting edge. But it’s coming down in a way that’s completely usable.
Chicago is lucky to have you—but what happens when you leave?
I do worry about the scalability. This has to be a sustainable system. I spend a good deal of time mentoring. Is that a guarantee? No. But if you go to the portal and see all that data—we’re not updating that by hand every day. All these things are automatic. I can go on vacation tomorrow or go away and the architecture is sustainable. That’s hugely important. This can’t be about me. Everything has to be documented. One of the worst things that people do is to undergo a big project and not document anything.