Connecting state and local government leaders

Using Predictive Analytics, Chicago Is a Trailblazer for Health Code Inspections

Chicago has more than 15,000 restaurants to inspect. The power of open data and predictive analytics will help the city prioritize its efforts.

Chicago has more than 15,000 restaurants to inspect. The power of open data and predictive analytics will help the city prioritize its efforts. EQRoy /


Connecting state and local government leaders

Harnessing municipal data, it’s easier to prioritize where to concentrate inspection efforts. And GitHub makes it easier for other cities to use the city’s model.

Chicago is one of America's biggest restaurant cities. Destination restaurants such as Alinea and Frontera Grill draw customers from around the world, and the city limits are home to more than 15,000 restaurants, grocery stores and other dining establishments. The city's Department of Public Health is now turning to a new tool to make sure those dining establishments don't make diners sick: Predictive analytics using massive open data sets.

Under the new initiative, Chicago will determine which restaurants are priorities for inspections using a predictive model and data sets which are posted on open source site GitHub.

Restaurants whose data points show anomalies when analyzed will be slated for priority inspection; the city of Chicago hopes that making their data and models available on GitHub will allow them to both leverage talent outside their organization to improve the lives of city residents—and to have other cities look to Chicago as a trailblazer.

Essentially, Chicago posted raw data on the city's filthiest restaurants to GitHub and asked for the public's help to both sanitize (pun not intended) the data they uploaded and to help improve the health department's statistical models. It's an unconventional approach designed to take hobbyist help from college and graduate students, data scientists and counterparts from other city agencies and public health departments and apply them to a very specific real world case.

The finished product is the result of a team led by Tom Schenk, Chicago's chief data officer.

Schenk, who is responsible for implementing data for use by city agencies and making sure it is accessible to the public and other stakeholders, says the move "[a]ctually impacts day-to-day lives."

In the process of building the models and creating an analytics program for the Public Health Department's internal use, which took about a year and a half, his team found outside input from other government and public health data wonks to be invaluable.

"We had a few conversations with other cities," Schenk told Route Fifty. "I suspect for us it was helpful in the process. Everything we used to create these conditions is online." He credits open sourcing and work by contributors inside and outside his organization to a successful test where the new model successfully increased spotting of critical violations by 25 percent.

Critical violations, according to Schenk, are the ones most likely to cause illnesses in diners.

Schenk's team were working on the initiative thanks to a $1 million grant given to Chicago in 2013 by the Bloomberg Philanthropies' Mayors Challenge. The grant money, which partially funded the project, was designed to enable data-driven decision making in city government.

“The use of open data will result in a more streamlined approach to overseeing food safety, targeting our resources at higher-risk establishments without compromising safety oversight at any food business across the city,” Mayor Rahm Emanuel explained in a statement. “This open data model is allowing Chicago to execute restaurant inspections with 21st century efficiency and effectiveness, and enhancing our ability to provide the best services to our residents.”

Chicago's initiative depended on deciding on which data points to use for the model, and finding the optimal way to analyze them. Data points include a mixture of restaurant/grocery store inspection data collected by the Public Health Department's inspectors, phone calls to 311, established risk categories, sanitation complaints located nearby an establishment, and city permit information. Schenk's team then combined these data points to build a model which showed which businesses had the greatest chance of critical violations in the near future.

Researchers from Allstate Insurance and the Civic Consulting Alliance helped in the modeling process.

Under Emanuel, Chicago has been a national leader in making government data open to the public. The city's Data Portal contains reams of data sets related to the minutae of urban life and made available by Schenk and his department.

For Chicago, making this information available free and online does more than make community groups happy; it also essentially outsources some civic improvement efforts to enthusiastic volunteers, freeing up government resources for other projects. More than 600 data sets are offered on the portal.

The project is the Chicago Department of Public Health's second effort to leverage crowdsourcing in order to prevent health code violations. In 2013, the department launched an ongoing project called FoodBorne, which mines Twitter for tweets about food poisoning in Chicago, and then reached out to those Twitter users to file an online complaint form that could trigger a possible inspection. According to the public health department, FoodBorne led to 150 additional inspections in the project's first year.

Chicago's restaurants and grocery stores are subject to the new open source data inspection project.

(Photo by EQRoy /

Editor's Note: This article has been updated to clarify that the new open source data inspection project only looks at restaurants and grocery stores.

Neal Ungerleider is a journalist based in Los Angeles and writes for Fast Company and consults on the tech industry.

NEXT STORY: Mapping America's Renters