What passengers reveal: Measuring safety and transit quality in informal transport with crowdsourced feedback

Informal transit serves the majority of urban travelers in many low- and middle-income countries, yet systematic data on safety and transit quality remain limited. Big data sources like telematics provide granular information on a limited set of indicators, while traditional passenger surveys offer comprehensive insights but are expensive and administered infrequently. We test lightweight, phone-based surveys to elicit volunteered passenger feedback across 77 Kenyan matatus (minibuses), combined with telematics data from 25 vehicles. Passengers rated driving speed, perceived safety, and provided open-ended comments. We assess whether such feedback provides information beyond existing sources like telematics and how it should be interpreted given its volunteered nature and subjective passenger perceptions. Passenger reports of high speeds align closely with telematics-measured speed, indicating crowdsourced feedback accurately captures observable driving behavior despite self-selection. In contrast, perceived safety shows weak association with telematics indicators; joint analysis of ratings and comments highlights the subjective nature of safety perceptions. Open-ended comments add value by revealing context telematics miss: distracted driving, alcohol use, and unsafe pickup practices. Though comments primarily reflect the survey’s safety emphasis, some address COVID-19 compliance, service quality, and vehicle condition, suggesting broader informational value. Large language models accurately classify comments in zero-shot settings, enabling scalable analysis. While challenges remain in consistently eliciting feedback, results demonstrate that short, low-burden phone-based surveys can generate actionable information on transit safety and quality.