Impact On AFRD and Atlanta

Figure 9.4: Interactive map of fires and inspections. The colored circles on the map represent fire incidents, currently inspected properties, and potentially inspectable properties in red, green, and blue, respectively. Inspectors can filter the displayed properties based on property usage type, date of fire or inspection, and fire risk score. Callout: activating the Neighborhood Planning Unit overlay allows an inspector to mouse-over a political subdivision of the city to view its aggregate and percentage of the fires, inspections, and potential inspections.

that fit their criteria for inspection. This included a shorter list of 6,096 new inspectable properties which are of the top 100 currently inspected property usage types (e.g., restaurants, motor vehicle repair facilities, etc), and a longer list of all commercial properties (19,397) that fit any property usage type that had been previously inspected. This dataset included the associated building- and parcel-level information for those properties in the form of a CSV file, with a subset of those properties (5,022) assigned a fire risk score. We then provided AFRD with an interactive map-based visualization tool, as part of the Firebird framework, for the fire inspectors and AFRD executive staff to use to augment their policy and decision-making processes. The map in Figure 9.4 was made using the open source map-decision-making tools Mapbox and Leaflet to create the base map layer. Then, using the Javascript visualization library D3.js, we displayed differently colored circles on the map to represent fire incidents, currently inspected properties, and potentially inspectable properties in red, green, and blue, respectively, using their longitude and latitude coordinates.

We also built a user interface for the Firebird map developed through discussions with the AFRD Assessment and Planning Section, and refined by incorporating feedback from fire

inspectors and AFRD executive staff. The map includes an information panel for displaying property information when hovering over a property on the map, such as its business name, address, occupancy usage type, date since fire incident or inspection, and fire risk score, if available. The map also includes a user interface panel with the ability to filter the fire incidents, the currently inspected, and the potentially inspectable properties according to their property usage type, the date of fire incident or inspection, and their risk score. Finally, we incorporated a set of regional overlays requested by the AFRD executive staff, including the AFRD battalions, and the Atlanta Neighborhood Planning Units (NPU) and Council Districts, which are both political subdivisions of the city. We included dynamically updated counts and percentages for the displayed fire incidents, current inspections, and potential inspections for each regional overlay (Figure 4), so the AFRD executive staff could make decisions at a battalion, NPU, or Council District level.

This map, and the Firebird framework in general, could be used as a powerful tool for supporting data-driven conversations about personnel and resource allocation and inspection decisions, and may even be used to inform decisions regarding community education programs for fire safety and prevention. To ensure AFRD could update the risk model and property visualization as new fire incident and inspection data becomes available, we shared our source code and process with AFRD’s Assessment and Planning Section, and made it publicly available on Github.⁴

9.6.3 Impact on AFRD Processes

After receiving the dataset of properties needing inspection, prioritized according to their fire risk score, AFRD has begun integrating the results of the analytics into their fire inspection process. Increasing the number of annual inspections by 6,096 (237%) overnight was not feasible without significant changes in organizational processes, local ordinances, or increased staffing. As an initial effort, AFRD assigned the 69 high-risk properties to the

4https://github.com/DSSG-Firebird

inspectors covering those respective areas. Of those, 27 had current or out-of-date fire safety permits that required re-inspection, 13 properties required new permits, and 15 properties recently went out of business. The remaining properties were found to not require a fire safety permit. Most significantly, the inspectors assigned to review these properties found a total of 48 violations that needed to be addressed to meet the Fire Code. As AFRD continues working through the list of potentially inspectable properties in descending risk order, the sheer number of additional inspections (increased workload) and potential violations identified and mitigated (positive outcomes) has already had a transformative impact on the daily operations of AFRD’s fire inspection process.

In addition to the immediate impact on the daily property inspections, the results of this work have stimulated important conversations within the executive leadership of AFRD and the Assessment and Planning Section about 1) how to more effectively allocate inspection personnel; 2) how to update and utilize the model to provide dynamic risk data in real time (e.g., on a monthly basis when new inspection assignments are given to the inspectors); 3) how to motivate increased data sharing between various government departments such as the Office of Buildings and AFRD; 4) how to give teams of firefighters access to fire safety permit and violation information when they respond to a fire emergency at that commercial property; and 5) how to extend the risk prioritization to residential properties using more behavioral data such as noise or sanitation ordinance violations, and consumer data from companies like Experian or ESRI.

Though there are many more inspectable properties than AFRD currently has the person-nel capacity to handle, AFRD has already begun to take steps toward a more efficient use of their existing personnel, by discussing how to assign inspectors to regions with a higher proportion of properties requiring inspection, rather than by the geographical assignment to fire battalions currently in use. In addition, they have begun to discuss altering properties’

inspection frequencies to reflect their fire risk levels. By prioritizing future inspections and more efficiently allocating inspection personnel to target the commercial properties most at

risk of fire, we hope this work will lead to a reduction in the frequency and severity of fire incidents in Atlanta. We also hope that this framework can be instructive for other municipal fire departments to improve their fire inspection processes.