From Accident to Insight: How the Data Flows


Finally New York City! Last week, our Harvard Kennedy School team of five students took the train down to The Big Apple to meet with representatives from the New York City Department of Buildings (DoB) and the Mayor’s Office of Data Analytics (MODA).

Our main goal was to understand what happens within the DoB once an incident occurs. Specifically, how exactly is incident data collected, stored, and analyzed? Also, how is this data used—and by who?—to carry out the DoB’s mandate to prevent injuries and fatalities on construction sites within the five boroughs?

A Lengthy Process from Construction Site to Incident Database

One key finding from our visit is that there is a multi-step process before anyone can use incident data. The process from accident to usable data includes:

  • First, the DoB has an Emergency Operations Center (EOC) that receives complaint calls from those reporting construction site concerns in NYC.

  • Next, the incident is categorized by the EOC officer into an Oracle database that automatically pushes out an email to the respective borough office and unit responsible to follow up on the issue.

  • The head of the respective department then assigns an enforcement Inspection officer to visit the site and check it for building safety or environmental control board violations.

  • Next, the inspector emails his findings from the site to their respective supervisor, who then enters the findings into a separate MS Access database built to record incidents.

  • Finally, an analyst copies the data, and once a month, publishes the raw data on the website for public viewing as a PDF.

Three Critical Issues in Data Management

Our team has identified three main problems areas at the DoB:

  1. There is data input, but little output: While a lot of people are involved in collecting construction incident data at the DoB, it is not clear to us that this data is being analyzed or looked at again by someone within the organization.

  2. What is collected is mainly inspection data, not safety data: The recent implementation of an enterprise IT solution provides inspectors in the field with tablets to receive assignments and easily enter inspection findings. However, it does not give them access to the construction incident database, which makes it harder for them or others to input safety data remotely.

  3. Data by itself do not lead to action: The Department of Buildings provided us with a very clear goal: “dive deeper into incident and fatality data to develop an actionable, predictive tool…to avoid future accidents.” But before we start exploring predictive analytics, the question for our team remains: who should be seeing and using incident data? And, more importantly, what can they do with it to improve safety?

Next Steps

The next research phase for our team is to identify and interview a variety of people inside and outside of the DoB, including possibly the general public, front-line construction workers, DoB managers, and DoB leadership. Our goal is to understand the complex process of data collection, storage, analysis and dissemination in order to identify ways in which we can improve this process. We hope that our efforts will ultimately help the Department of Buildings make data-driven risk mitigation decisions, which would result in fewer construction related injuries and fatalities.

Anthony Arendt, Dan Bacon, Howaida Kamel, Kirsten Rulf, and Daniel Wagner