E-Commerce & Last-Mile Logistics
Reduction in cost per delivery drop
On-time delivery rate achieved
About
A group of e-commerce retailers and their delivery couriers were all fighting the same battle against soaring last-mile costs. Static routes ignored rush-hour traffic, couriers idled 22% of each shift, and SLA fines reached Rs. 11 lakh a month. We built a real-time vehicle-routing engine that recalculates every five minutes and pushes turn-by-turn instructions to a rider app.
Industry
E-Commerce & Last-Mile Logistics
Company size
1,000 – 5,000 employees
Founded
2015
The Company
Retailers and couriers fighting the same cost pressure
A group of e-commerce retailers and the courier networks delivering for them shared a common problem: last-mile delivery cost was rising faster than order volume, and each side had limited visibility into what the other was actually experiencing on the road.
Delivery routes were built once at the start of the day and dispatched to riders as a fixed plan, with no mechanism to adjust as real conditions — traffic, new orders, cancellations — diverged from what the morning plan had assumed.
The challenge
A static morning plan against a day that never stayed still
Static routes took no account of rush-hour traffic once the day was underway, and couriers ended up idling roughly 22% of each shift waiting between poorly sequenced stops. SLA fines from missed delivery windows reached Rs. 11 lakh a month across the network.
Every new order placed after the morning route was built had to be manually slotted in by a dispatcher, a process that scaled poorly as order volume grew and left little room to actually optimise the resulting sequence.
The Solution
A vehicle-routing engine that never stops recalculating
We built a real-time vehicle routing problem (VRP) engine that recalculates arc costs and route sequencing every five minutes rather than once at the start of the day, pushing updated turn-by-turn instructions directly to a Flutter rider app as conditions change.
Every KPI — cost per drop, idle time, on-time rate — logs to Redshift, giving the operations team a continuous improvement loop rather than a once-a-day retrospective on how the previous day's static plan had performed.
The Results
20% lower cost per drop, on-time delivery at 97%
Cost per delivery drop fell 20% as continuous route recalculation eliminated the idle time and inefficient sequencing that had been built into the static morning plan. On-time delivery reached 97%, up from a baseline well short of the 96% target the network had been struggling to hit.
The results were strong enough, and the underlying engine general enough, to open the door to scaling the same routing approach into new geographies without a fundamentally different build each time.
KEEP READING
Explore more studies on AI, analytics, and enterprise intelligence.
