Why Route Optimisation Is Different in India
Route optimisation algorithms designed for Western markets often underperform in India because they don't account for:
- •Road quality variation: A 10 km route on a state highway may be faster than a 7 km route through a city with potholes and speed breakers
- •Traffic patterns: Indian city traffic is highly unpredictable - peak hours vary by city and season
- •Toll costs: Multiple toll plazas on national highways can make a longer route cheaper overall
- •Delivery time windows: Many Indian customers (especially B2B) have strict unloading windows
- •Vehicle restrictions: Many city centres have time-based restrictions for commercial vehicles
A good route optimisation system for India must incorporate all of these factors.
The Vehicle Routing Problem (VRP)
At its core, route optimisation solves the Vehicle Routing Problem: given N delivery locations and M vehicles, find the set of routes that minimises total cost (distance, time, or fuel) while satisfying constraints (vehicle capacity, time windows, driver hours).
VRP is NP-hard - finding the mathematically optimal solution for large fleets is computationally infeasible. In practice, we use heuristic algorithms that find near-optimal solutions in seconds.
Common Algorithms
| Algorithm | Best For | Typical Quality |
|---|---|---|
| Nearest Neighbour | Quick baseline, < 20 stops | 15–25% above optimal |
| Clarke-Wright Savings | Medium fleets, 20–100 stops | 10–15% above optimal |
| Genetic Algorithm | Large fleets, complex constraints | 5–10% above optimal |
| Google OR-Tools | Any size, open source | 3–8% above optimal |
For most Indian SMB fleets (10–100 stops per vehicle), Clarke-Wright Savings or OR-Tools provides excellent results.
Static Route Planning
Step 1: Geocode All Delivery Locations
Convert addresses to GPS coordinates. Indian addresses are notoriously inconsistent - use a geocoding service with India-specific training (Google Maps, MapmyIndia, or HERE Maps).
Common issues:
- •Incomplete addresses (no pin code, no landmark)
- •Multiple locations with the same name
- •New developments not yet in mapping databases
Solution: Maintain a verified location database. When a new delivery address is added, require the driver to confirm the GPS pin on first delivery.
Step 2: Define Constraints
For each vehicle:
- •Maximum payload (kg or volume)
- •Maximum driving hours per day (legal limit: 10 hours in India)
- •Start and end depot location
- •Permitted vehicle types for each route (some roads restrict heavy vehicles)
For each delivery:
- •Time window (earliest and latest acceptable delivery time)
- •Service time (how long the driver spends at each stop)
- •Special requirements (refrigerated vehicle, tail-lift, etc.)
Step 3: Run Optimisation
Input all locations and constraints into the routing engine. The output is an ordered sequence of stops for each vehicle with estimated arrival times.
Step 4: Driver Briefing
Share the optimised route with drivers via the mobile app. Include:
- •Turn-by-turn navigation
- •Customer contact details
- •Delivery instructions
- •Expected arrival times to share with customers
Dynamic Re-Routing
Static routes break down when reality diverges from the plan: a customer is not available, traffic is worse than expected, or a vehicle breaks down.
Triggers for Re-Routing
- •Customer cancellation or reschedule
- •Traffic delay > 30 minutes on current route
- •Vehicle breakdown (redistribute remaining stops to other vehicles)
- •New urgent delivery added mid-day
Re-Routing Architecture
The system monitors live GPS positions against planned routes. When a vehicle deviates significantly or falls behind schedule, it triggers a re-optimisation for the remaining stops on that vehicle's route.
Re-optimisation runs in < 5 seconds for routes with < 50 remaining stops, making real-time re-routing practical.
Multi-Stop Sequencing for Last-Mile Delivery
For last-mile delivery (many stops, small packages), the key optimisation is stop sequencing - the order in which stops are visited.
The Milk Run Pattern
For regular delivery routes (same customers, different quantities each day), use a fixed sequence with dynamic quantity adjustment. This reduces driver cognitive load and allows customers to predict delivery times accurately.
Time Window Optimisation
When customers have strict time windows, the algorithm must balance:
- •Minimising total distance
- •Ensuring all time windows are met
- •Minimising the number of vehicles required
A common approach: cluster stops by time window first, then optimise sequence within each cluster.
ROI Calculation
Baseline Measurement (Before Optimisation)
- •Average km per delivery: ___
- •Average fuel cost per km: ₹___
- •Average deliveries per vehicle per day: ___
- •Number of vehicles: ___
Expected Improvements
- •Distance reduction: 15–25% (typical for unoptimised routes)
- •Fuel saving: proportional to distance reduction
- •Additional deliveries per vehicle: 10–20% (from time saved)
Example Calculation
| Metric | Before | After (20% improvement) |
|---|---|---|
| Daily km per vehicle | 200 km | 160 km |
| Fuel cost per vehicle/day | ₹2,400 | ₹1,920 |
| Saving per vehicle/day | - | ₹480 |
| Fleet of 20 vehicles | - | ₹9,600/day |
| Annual saving | - | ₹28.8 lakh |
Implementation Roadmap
Month 1: Data Foundation
- •Geocode and verify all delivery locations
- •Install GPS trackers on all vehicles
- •Establish baseline metrics (current km/delivery, fuel cost)
Month 2: Static Optimisation
- •Implement route planning for next-day deliveries
- •Train dispatchers on the routing tool
- •Measure improvement vs. baseline
Month 3: Dynamic Optimisation
- •Enable real-time re-routing
- •Integrate customer notification (ETA updates via WhatsApp)
- •Review and refine constraint settings based on real-world performance