Boda-bodas are a vital part of Kampala's transport network. Improving safety requires more than drawing a heatmap: the underlying records must be reliable, the analysis must account for exposure, and proposed interventions must be evaluated after implementation.

Motorcycles parked along a city street
Location alone misses part of the story. Add time patterns and road context before choosing a safety intervention.

The Uganda Police Force Annual Crime Report 2025 records 5,383 road deaths nationally and 2,525 motorcyclist deaths. Those are national figures, not a Kampala-only count and not proof of any particular junction-level pattern. They provide a reason to investigate road-user risk carefully, not permission to invent local clusters.

Start With a Defensible Dataset

A useful analysis would combine police crash records with verified coordinates, time, road-user type, injury severity, road geometry, lighting, traffic volume, and reporting coverage. Before modelling, document missing locations, duplicate reports, changes in reporting practice, and whether near-misses are captured. Protect personal identifiers and publish only aggregated locations.

Use Spatial Analysis as an Investigation Tool

Density-based clustering such as DBSCAN can help identify groups of recorded incidents when the number of clusters is not known in advance. Results depend on coordinate accuracy, distance thresholds, time windows, and the completeness of the source data. A cluster is a signal for field inspection, not proof that road design caused the crashes.

Compare locations across time periods and normalise where possible for traffic or motorcycle exposure. Separate fatal and serious-injury incidents from minor collisions. Validate candidate sites with road-safety engineers, traffic officers, local communities, and direct observation before recommending changes.

Turn Evidence Into Testable Interventions

Potential interventions include clearer junction markings, speed management, safer pedestrian crossings, improved lighting, visibility improvements, and better-managed stopping or waiting areas. The correct measure depends on the observed crash mechanism. Do not prescribe a breathalyser checkpoint, speed bump, or lane redesign from a map alone.

For each pilot, define a baseline period, an implementation record, comparison locations where feasible, and post-intervention measures. Track injury severity and exposure as well as raw incident counts; a fall in reports may reflect changed reporting rather than safer roads.

Make the Method Reproducible

Publish the data dictionary, cleaning rules, coordinate precision, model parameters, limitations, and the date of the source extract. Keep sensitive records private. A reproducible analysis that exposes uncertainty is more useful to city planners than a dramatic map with unsupported percentages.

The engineering opportunity is real: make trustworthy evidence easier to inspect and act on. The safety outcome must still be demonstrated in the field.

Sources and further reading