Your 311, Streamgraphed

Six years ago, Wired published an in-depth article about using 311 data as a tool for engaging with and measuring the needs of New York City’s constituents. The captivating image, shown here, presented 7 days of data represented as a single 24-hour view and left me thinking deeply about 311 calls. Sewer maintenance problems were being called in at 3am more often than I would expect. Clearly 6:30am was (and probably still is) the best time to contact 311, as the fewest number of people call around then.

A streamgraph visualization of NYC 311 service request data

This type of data visualization is called a streamgraph. It’s basically a trippy, gravity-defying stacked bar chart. The one above didn’t even bother with numbers. It’s easy to tell that noise complaints made up the vast majority of calls at midnight, but not how many calls were received overall. I have wanted to make a visualization like it for a while. Two cities have recently published their service request data using the new Open311 GeoReport bulk data standard we helped create, so it seemed fitting to reward them in a similar psychedelic-for-geeks fashion.

Louisville 311 Service Requests Streamgraph

Louisville, Kentucky, released 23 years of MetroCall 311 data. Above you can see the top ten complaints during the last four years in full streamgraph glory. Because there’s no color key, I’ll just tell you that the big blue spike in March of 2015 was for pothole complaints. Aside from having the heaviest snowfall since 1968, Louisville embarked on a #502pothole campaign, encouraging people to report them online rather than by phone. The central green band that grows every summer is, well, reports of unkempt grass and weeds that grow every summer. Below is the same graphic, but this time gravity and vertical axis labels have been turned on. March 2015 was indeed a busy month, with more than 7,500 calls, and that doesn’t include the calls for other types of services beyond the top ten.

Louisville 311 Requests Streamgraph

I’m not the only one to visualize Louisville’s 311 data. Check out these Sankey and circle-packing diagrams, as well as this mosquito map by Matthew Goth-Olsen, who works for the City.

About 760,000 people live in Louisville/Jefferson County, but Bloomington, Indiana, has just more than 1/10th that population. Therefore, it’s no surprise that Bloomington’s volume of requests for service is lower. Based on Bloomington’s uReport data, below is a peek at the City’s top ten requests since November 2015. (If you’re looking for inspiration for your own analysis, try doing a comparison of both cities’ annual requests per capita.)

Bloomington 311 Requests Streamgraph

Bloomington received about 100 reports a month related to recycling issues, indicated by the yellow band that was consistent year-round. The red band on top that started in March and slowly declined into October represents concerns about yard waste pickups. Finally, the orange band that appeared in April and dropped off around October were reports of overgrown yards.

People in both Louisville and Bloomington care about the appearance of their neighborhoods. As you can see from the graphs above, the time of year clearly influences the volume of requests. Since both of these cities released their service request data in the same format (with some minor issues), it was easy to use the same tool to process their respective datasets and create these visualizations.

In 2012, Mark Headd wrote about Open311 serving as a foundation for municipal collaboration. These visualizations are just the tip of the iceberg. If the more than 25 cities that already publish their service requests inconsistently joined the first two innovators, some of the tools developed by Code for America fellows and brigades could be quickly adapted. GovEx is also working on a tool to test compatibility with the Open311 GeoReport bulk data standard. We strongly support open data standards that help reusable tools and data scale across the globe. Stay tuned! We have even more magic up our sleeves that we plan to reveal soon.

And if you’re interested, below is the R script to generate an interactive version of the visualizations illustrated above.