One of the great promises of open government this decade has been that it can serve as a catalyst for a new civic-centered “innovation ecosystem.” This ecosystem, replete with successful startups and data-driven advancements in government operations, could not only enhance transparency, but spur replication and generate economic value for cities. In 2013, McKinsey and Company concluded that open data, in all its forms, had the potential to contribute $3 trillion a year of value across the global economy. This and other reports set off a wave of excitement, encouraging The Economist to declare that “the open data movement has finally come of age.”
Yet in the nearly three years since, progress at the local level has been incremental. Cities like Chicago have achieved many operational data-driven advancements, yet the process of replication has been slower than many in 2013 may have hoped. Consider that for almost a year now, Chicago’s food inspection program has been guided by predictive analytics.
The original predictive algorithm, developed by the city’s Department of Innovation and Technology (DoIT), relied entirely on open data to identify which restaurants were most likely to be in critical violation of city health codes. Working collaboratively with the Department of Public Health (CDPH) and pro-bono data scientists at Allstate Insurance, the algorithm was piloted in late 2014 to considerable success: inspectors were able to discover critical violations 25% faster, on average, then if they had used the traditional inspection method. By 2015, the algorithm was integrated into regular food inspection operations.
The algorithm is open source, of course, and free to be used by anyone. Yet despite its efficacy at a low cost for Chicago, there has not been a rush from other cities to replicate the program (at least not yet). Following food inspections coverage from this website, the program’s success—and its replication challenges, for that matter — were even covered by The Atlantic and PBS NewsHour, helping increase awareness.
So why is this lack of replication the case? In short, it isn’t always easy for cities to do it – even when the algorithm itself is free. Some governments don’t have the internal staff or capacity to take on such projects. Others may be leery that the effort it takes to adopt an analytics program, whether perceived or actual, is not worth the cost. If cities without internal capacity wish to seek partners or vendors, they either may not know which are best to reach out to, or have their efforts complicated or undermined by complex procurement structures.
While these barriers are only part of the story for why the innovation ecosystem is perhaps not as robust as it could be, a good starting point to overcoming them is to enhance open government beyond raw information on a portal. It’s a point not lost on Tom Schenk, CDO of Chicago, who acknowledges the challenges of spurring replication. “The specifics do change between cities,” Schenk said to The Atlantic. “To even pick up code and adapt it to your specific business practice still takes work.”
To address these challenges, Schenk has expanded Chicago’s open data program during his tenure to include a growing GitHub repository of algorithms and codes, and even OpenGrid, an open-source, map-based platform to visualize city data. OpenGrid in particular has been made available to other cities as a one-click download on Amazon Web Services’ marketplace, easing barriers to access in the process.
Yet for adopting something more advanced, like Chicago’s food inspection program, sometimes the basic principles of hard work and a nimble operation can get the job done. Since the food inspection program was launched last year, one locality has taken advantage of the algorithm: Montgomery County, Maryland, located just outside of Washington, D.C.
Municipal Analytics, the Startup Way
Montgomery County did not learn of Chicago’s food inspections algorithm project on its own, nor did it attempt to replicate it through the efforts of its own internal staff. Rather, it began the process by working with Open Data Nation (ODN), a D.C.-based civic startup that had taken notice of Chicago first and reached out to the County accordingly with the hopes of replicating it. ODN, founded less than two years ago by former Brookings and Urban Institute analyst Carey Anne Nadeau, uses open data to integrate predictive analytics processes into city operations.“From what I’ve seen, a lot of cities and local governments want to take on these big analytics projects themselves — they get that data-driven decision making is the future,” says Nadeau. “Doing that has worked in some big places, like New York, Chicago, and Boston, but is that model really applicable everywhere?
It may not be in Montgomery County, which did not have the full capacity to take on a food inspections analytics program on its own. ODN was able to develop an advanced algorithm that used the county’s available open data to prioritize inspectors’ work order task lists, so inspectors go to the highest-risk locations first. Building upon Chicago’s original model, ODN also added in additional variables that Chicago did not originally include — such as new construction permits and data from Yelp.
ODN then launched a pilot to compare Montgomery County inspection results from analytics-enhanced operations to results from business-as-usual operations. Like Chicago’s, ODN’s pilot was a success — and even yielded similar results: when using the predictive algorithm to guide food inspections, Montgomery County was able to identify 27% more violations, on average, and do so 3 days sooner than traditional methods. That increased efficiency was estimated to recapture an estimated $2 million for Montgomery County in its first year alone.
These Montgomery County results are significant: it shows that such algorithms can not only be effective, but effective in highly varied settings. Whereas Chicago is a heavily urban area, Montgomery County is a combination of urban, suburban, and rural areas, containing roughly one third of Chicago’s population in nearly double the amount of geographic space.
ODN’s work with food inspection algorithms has not been an ad hoc project, either, but rather the start of a business model. Following its successful work with Montgomery County, ODN has taken its process and turned it into a program called FIVAR (Food Inspection Violations, Anticipating Risk). FIVAR combines ODN’s algorithm with a web application so that cities may use their open data in real-time to enhance their food inspections processes. The application, currently undergoing beta testing, is already in use in Montgomery County.
In its use of analytics, FIVAR’s end goals are ultimately the same as those of Chicago’s DoIT analytics team: to use analytics as a tool to keep costs down by enhancing government efficiency and effectiveness.
Startups and the Civic Innovation Ecosystem
ODN’s work shows that startups can play an important role in helping fill the innovation replication gap in local government. It’s a role that Nadeau believes can make advanced analytics accessible to all cities and governments, not just large ones.“In our consultations with Montgomery County — and in interviewing health inspectors in more than 30 cities —we’ve realized that it often takes a good amount of resources, energy, and effort to transform government from within,” says Nadeau. “But whether or not they’re wary of analytics, they all have one key value in common: doing the best with what they have. And that’s where startups can come in — to generate, test, and operationalize effective algorithms, they can augment cities’ capacity for data science in a way that can’t be done in-house.”
ODN’s work echoes the economic value and cross-jurisdiction replication activity that many have wanted to see in a growing civic innovation ecosystem. Projects like FIVAR also illustrate the mutually beneficial, give-and-take processes that startups have been generating with cities. “Since Chicago has made their algorithms open source from the start, they’ve helped open up opportunities for startups like us—but in return, they’re helping themselves, too,” Nadeau notes.
Indeed, by adding variables that were not in Chicago’s original algorithm, ODN’s food inspection model has provided Chicago with lessons on how they can enhance their own model as well. ODN has also been able to review and add changes, or “commits,” to Chicago’s GitHub code page, strengthening it in the process. “It’s not just Open Data Nation that’s doing this type of work, either,” says Nadeau. “There have been others who have helped improve what they have on GitHub as well.”
From Chicago’s perspective, having organizations like ODN take advantage of its open government resources helps justify the program in the first place. “We’re always happy to talk to people who are interested in working with our code,” says Schenk. “At the end of the day, we’re all working towards the same mission. The more we can work together to make data and analytics easily accessible to everyone, the better off we’ll be.”
