IE 11 Not Supported

For optimal browsing, we recommend Chrome, Firefox or Safari browsers.

Data Analytics and the Soup that Made You Sick

How can 32 Chicago inspectors monitor 1,500 restaurants? Figuring out which ones aren't likely to pass inspection is a good start.

For public leaders, it's a familiar conundrum: Many governments are perpetually cash-strapped, but demand for their services never wanes. The rise of predictive analytics gives the public sector a tool to deal with the problem by providing data that allows agencies to deploy their resources more efficiently.

It's been nearly two decades since New York City deployed CompStat to identify where crimes were most likely to occur and inform then-Police Commissioner William Bratton's decisions about where to deploy officers. Predictive analytics have come a long way during that time.

You don't have to have been a math major to know how hard it is for 32 inspectors to monitor Chicago's more than 1,500 restaurants. A pilot program at the city's Department of Public Health (DPH) holds the promise of making that ratio a little more manageable. By analyzing a decade's worth of publicly available data, the department determined that having past violations and being located near a construction site are among the variables that correlate with the likelihood of a restaurant not passing inspection.

What might surprise you, though, is that the single best predictor of restaurants that are likely to fail inspection is when the weather is such that ingredients are more likely to go bad. Utilizing data like these to determine when and where to conduct inspections has led to a 4 percent increase in the critical health-code violations discovered by those 32 inspectors.

Social media is providing data that can further focus government agencies' efforts, since many restaurant patrons are more likely to report food poisoning online than to the authorities. Chicago's DPH mines Twitter for tweets that include words linked to it and uses the data to help determine inspection targets. Similarly, New York City's Department of Health and Mental Hygiene scans posts on review websites such as Yelp and flags mentions of food poisoning to help guide the deployment of its inspectors.

Policing and restaurant inspections are only the beginning of predictive analytics' potential applications. When the New York City fire department studied data on buildings that had been illegally subdivided, it found four variables that make some city buildings 40 times more likely to catch fire: buildings whose owners are experiencing financial difficulty, those with a history of illegal-conversion complaints, those built before the advent of modernized building codes in 1938, and structures in low-income, high-immigrant neighborhoods.

Scarce resources are likely to remain a fact of life in the public sector. There is no silver bullet that can address all the challenges scarcity creates, but technologies that allow agencies to deploy their resources more efficiently are one promising strategy for maintaining the level of public services that citizens demand.

This story was originally published by Governing