September 20, 2013 By Stephen Goldsmith
In times of emergency, good government means fast government, able to react nimbly and purposefully to new conditions as they arise. Having speedy government tomorrow, though, depends on preparation and prepositioning of critical resources today with the understanding that seemingly random events often fit into actionable patterns. By understanding these patterns now through clever combinations of data and new modeling techniques, governments can improve their responses and become more effective.
In large cities, for example, must police officers simply patrol hoping they will see a crime, or might they use data on offenders, past crimes, neighborhood conditions and time of day to improve surveillance? Hot-spot crime analysis has grown in popularity for just this reason, giving patrols a way to focus their energies on specific areas that have a higher propensity for crime. Recent advances continue to refine these statistical methods, such as a new system developed for and in use by Seattle, Los Angeles and Santa Cruz that feeds past and current crime trends into the model used to predict earthquake aftershocks, adapting that model to predict crime as well.
Building a fully integrated system -- making different datasets speak to each other and instilling a cooperative strategy across departments -- is a lot of hard work. But being able to look at different phenomena at once can pay off both during emergencies and in the day-to-day functioning of a city. Chicago, for example, has been developing WindyGrid, a predictive-analytics platform that has begun to reveal relationships such as spikes in stolen trash bins when a block's streetlights go out. Those little extra costs add up, and now the city knows what kind of actions need to be taken while it works on a streetlight repair.
WindyGrid's planned capacity includes being able to preemptively react to a range of emergency situations -- knowing when a water main is likely to break, for instance, or being able to respond more quickly during a massive snowstorm. The goal is a preemptive platform comprehensive enough to solve issues in areas ranging from infrastructure to public safety while accessible enough that a city employee could simply query the database when he or she has a hunch that might result in better service and big savings.
For decades, cities have worked to optimize their ambulance response times by having drivers park in locations with a high incidence of emergencies rather than wait in firehouses. In New York City, after an initial brainstorming process where various theories about placement were introduced, the data team was able to granularly measure 911 responses from dial to arrival, which enabled systemic improvements to the entirety of the response. More important, it demonstrates the iterative nature of this work, from common-sense hypotheses to data-driven enhancements allowing in the end a deep and comprehensive understanding of the entire 911 transaction.