ETAPrediction

Bus and shipment ETAs predicted using historical, real-time, weather, and crowd-sourced data for accuracy

Business impact

• ETA prediction accuracy improved by 15% to 70% depending on deployment — More accurate ETAs reduce delays, missed connections, and improve customer satisfaction
• Reduced miles driven and fuel consumption (e.g., UPS saved 100M miles annually) — Optimized routes and ETAs lower operational costs and environmental impact
• Decreased idle time at ports and delivery wait times — Better ETA predictions enable efficient scheduling and resource allocation

Data requirements

• Historical transit and trip data (Structured) — Used to train models on typical travel times and patterns
• Real-time GPS and traffic data (Numeric) — Provides current conditions to adjust ETA predictions dynamically
• Weather data (Numeric) — Incorporated to account for environmental impacts on travel times
• Crowd-sourced user data (Text) — Augments real-time inputs with user-reported delays and conditions

AI methods and techniques

• Predictive AI — Models forecast ETAs by learning from historical and real-time data patterns

AI models and model families

DeepETA, Gradient Boosted Trees, Neural Networks, Transformer-based models, Llama, GPT-4o

Industries

Real-world evidence

6 documented case studies on record.

Companies using this: DHL, Kpler, Transit, UPS, Uber, Vale.