Last-Mile Delivery Platform
ML-powered route optimisation and Flutter driver app for 3,000 daily deliveries.
All case studiesA last-mile delivery company operating in 5 Indian metropolitan areas — Delhi NCR, Mumbai, Bengaluru, Hyderabad, and Chennai — engaged Cydez Technologies to build an ML-powered route optimisation and fleet analytics platform. The company managed 3,000 daily deliveries with a fleet of 400 delivery riders, serving e-commerce clients, D2C brands, and quick-commerce operators.
Drivers used Google Maps for navigation with no multi-stop route optimisation — each driver manually decided their delivery sequence. Failed first-attempt deliveries ran at 22%, costing the company approximately INR 180 per failed attempt in re-delivery logistics. Fuel costs consumed 28% of revenue, and there was no analytics capability to identify patterns in delivery failures, driver performance variations, or route efficiency across different city zones.
Cydez built an ML-powered route optimisation engine using OR-Tools and custom heuristics that generated optimal multi-stop routes considering real-time traffic (via Google Maps API), delivery time windows, vehicle capacity constraints, and historical delivery success rates by address and time slot. A Flutter-based driver app provided turn-by-turn navigation, electronic proof-of-delivery with photo and OTP verification, and real-time communication with the dispatch centre. The React operations dashboard delivered fleet analytics covering delivery success rates, driver performance scorecards, fuel consumption, zone-level profitability, and customer satisfaction metrics.
- ML route optimisation for multi-stop deliveries
- Flutter driver app with navigation and ePOD
- React operations and fleet analytics dashboard
- Delivery failure prediction and prevention
- Driver performance management system
- Zone-level profitability analysis
A last-mile delivery company operating in 5 Indian metros was losing margins to inefficient routing, high fuel costs, and failed delivery attempts. Drivers used Google Maps for navigation with no route optimisation across multiple stops. Failed first-attempt deliveries ran at 22%, and there was no analytics to identify patterns in delivery failures or driver performance.
Cydez built an ML-powered route optimisation engine that generated optimal multi-stop routes considering traffic patterns, delivery time windows, vehicle capacity, and historical delivery success rates. A Flutter-based driver app provided turn-by-turn navigation, proof-of-delivery capture, and real-time communication with the dispatch centre. A React operations dashboard delivered fleet analytics covering delivery success rates, driver performance, fuel consumption, and customer satisfaction.
How we delivered this project
Discovery
Analysed 6 months of delivery data across 5 metros. Mapped failed delivery patterns by address, time slot, and customer segment. Benchmarked driver performance variations. Documented existing dispatch and routing workflows. Conducted ride-alongs with 20 drivers to understand field challenges.
Design
Designed the route optimisation engine using OR-Tools with custom constraints for Indian urban conditions. Created the Flutter driver app UX with large-button interface for use while riding. Designed the operations dashboard with zone-level drill-downs. Specified the delivery failure prediction model.
Development
Built the route optimisation engine with Google Maps API integration for real-time traffic. Developed the Flutter driver app with offline capability, ePOD capture, and turn-by-turn navigation. Built the React operations dashboard with real-time fleet tracking. Trained the delivery failure prediction model on historical address-level data.
Launch
Deployed city-by-city over 8 weeks starting with Bengaluru. Each city required traffic pattern calibration and driver training. Conducted A/B tests comparing optimised routes against driver-chosen routes for 30 days. Achieved target metrics within 60 days of full deployment across all 5 metros.
What we built
ML Route Optimisation
Multi-stop route optimisation considering real-time traffic, delivery windows, vehicle capacity, and historical success rates. Average 28% fuel reduction vs. driver-selected routes. Re-optimisation triggered by real-time traffic changes.
Driver Mobile App
Flutter app with turn-by-turn navigation, electronic proof of delivery (photo + OTP), customer communication, and break time management. Offline capability for areas with poor connectivity. In-app training modules for new drivers.
Delivery Failure Prediction
ML model predicting delivery failure probability at address + time slot level. High-risk deliveries flagged for pre-delivery customer confirmation call. Suggested optimal delivery windows for repeat addresses.
Fleet Analytics Dashboard
Real-time fleet tracking, driver performance scorecards, fuel consumption analysis, and delivery success rate trends. Zone-level profitability analysis enabling data-driven pricing decisions.
Driver Performance Management
Individual driver scorecards covering deliveries per hour, success rate, fuel efficiency, customer ratings, and safety compliance. Automated incentive calculation based on performance metrics.
Customer Experience
Real-time delivery tracking shared with end customers via SMS/WhatsApp links. Delivery time window predictions with 88% accuracy. Post-delivery feedback collection and service recovery workflow.
Measurable outcomes
- 28% reduction in fuel costs through route optimisation
- Failed first-attempt deliveries reduced from 22% to 9%
- Average deliveries per driver per day increased from 35 to 48
- Driver onboarding time reduced from 3 days to 4 hours with in-app training
Built with
Want similar results?
Tell us about your project and we'll explore how Cydez can deliver comparable outcomes for your organisation.
Start a Conversation