Dynamic Pricing, Route Planning, and Order Assignment at Dada‑JD Daojia: Algorithms and Practices

On October 22, at the PyCon China conference in Shanghai, Liao Ruiqi from Dada Logistics’ algorithm team shared the company’s practical experiences in dynamic pricing, route planning, and order assignment for the Dada‑JD Daojia platform.

Background

Dada is a leading instant logistics information service platform that, using only crowdsourced couriers, delivers orders from JD Daojia’s fresh‑food supermarkets, restaurant delivery platforms, JD Mall, and long‑distance same‑city express services.

Many challenges in this process require artificial‑intelligence techniques, which have significantly improved platform efficiency and generated substantial benefits.

Dynamic Pricing

The platform connects diverse demand sources with millions of crowdsourced couriers. Efficient operation depends on balancing supply and demand: excess couriers cause attrition, while insufficient couriers degrade service quality.

Imbalance can appear across three dimensions:

Spatial imbalance : heat‑maps of each region are used to evaluate supply‑demand gaps.

Temporal imbalance : weather, holidays, and peak hours cause large fluctuations in order demand and courier availability.

Order‑level imbalance : each order has multi‑dimensional characteristics that affect delivery difficulty and supply‑demand status.

To address these imbalances, Dada implements real‑time, per‑order dynamic pricing ("thousand‑order‑thousand‑price") to balance courier earnings, user experience, and platform profit.

Route Planning

Couriers often handle multiple orders simultaneously; the platform must plan routes that minimize travel distance and maximize delivery efficiency.

The single‑courier routing problem is a classic Traveling Salesman Problem (TSP):

Because TSP is NP‑hard, Dada employs a heuristic genetic algorithm to quickly generate near‑optimal routes. The algorithm produces short routes and total distances within a few milliseconds even when planning for more than ten points, while maintaining high accuracy.

With this foundation, the system can evaluate whether orders are on the same route and whether a specific courier is suitable for a given order.

Order Assignment

Dada uses a hybrid of order‑grabbing and dispatch mechanisms to allocate orders fairly and efficiently.

Assignment considers route compatibility, delivery efficiency, courier preferences, capacity, activity level, and fairness, making it a constrained multi‑objective optimization problem that balances platform performance with courier satisfaction and operational stability.

By combining dispatch and grabbing, most orders are allocated within one minute of being issued.

Conclusion

The algorithmic challenges Dada faces are novel, with few existing best practices. Unlike traditional internet problems with clear inputs and goals, offline‑online logistics involve complex, multi‑objective optimization.

Problems such as multi‑order route compatibility are more intricate than ride‑hailing (which typically considers only two orders). Real‑time logistics demand stricter timeliness and higher variability.

These challenges also present opportunities: algorithmic optimization can create huge value. With rapid AI advances and the rise of new retail, the logistics sector is undergoing massive transformation, and Dada‑JD Daojia plays a pivotal role. The team invites interested students to join and explore this exciting algorithmic frontier.

Author Bio : Liao Ruiqi, head of Dada’s delivery algorithm team, has extensive experience in machine learning, logistics algorithms, and computational advertising. He joined Dada in early 2016, built the algorithm team from scratch, and now leads work on order assignment, dynamic pricing, and route planning.