Algorithmic Order Batching: The Mathematics of 35% Faster Hyperlocal Deliveries

In on-demand delivery, fulfilling orders sequentially (Store A → Customer A, return to base, Store B → Customer B) results in high rider idle time, excessive fuel burn, and unsustainable delivery fees.

The Geospatial Clustering Architecture

FirstMartt's routing engine continuously computes dynamic spatial clusters across active orders:

• **Voronoi Polygon Partitioning:** Divides the city into dynamic micro-zones based on active merchant order density and rider availability.
• **Dynamic Insertion Heuristics:** When a new order arrives while a rider is en route to pick up an order 200 meters away, the algorithm evaluates whether inserting the new pickup adds less than 3 minutes to the total route.
• **Multi-Merchant Bundling:** Allows a customer to order fresh milk from Kirana A and croissants from Bakery B, which a single rider picks up in one trip and delivers together.

Concrete Efficiency Gains

Algorithmic batching increases rider hourly drops from 1.4 to 3.2, cutting the cost per delivered drop by over 45% while boosting rider hourly earnings.