Using ML and Optimization to Solve DoorDash’s Dispatch Problem
A behind-the-scenes look at how we match Dashers with orders
At DoorDash, we constantly look for ways to improve the customer experience. One of the most critical aspects of the customer experience is the speed of delivery. We want to ensure our customers get their food as quickly as possible.
One of the biggest challenges to delivering food quickly is matching Dashers with orders. Dashers are independent contractors who use their cars to provide food. Many factors go into matching a Dasher with an order, such as the restaurant’s location, the customer’s location, and the time of day.
In the past, we used a manual process to match Dashers with orders. This process was slow and inefficient. We often had to make compromises, such as assigning a Dasher to an order that was too far away or giving an order to a Dasher who was already busy.
Solution
We decided to use machine learning and optimization to solve the dispatch problem. We built a DeepRed system that uses data from past deliveries to predict the best way to match Dashers with orders. DeepRed considers factors such as the restaurant’s location, the customer’s location, the time of day, and the availability of Dashers.
Results
Since implementing DeepRed, we have seen a significant improvement in the speed of delivery. We have also seen a decrease in the number of complaints from customers about the wait time.
In addition to improving the speed of delivery, DeepRed has also helped us improve our operations’ efficiency. We have been able to reduce the number of Dashers who are idle, which has saved us money.
DeepRed is a powerful tool that has helped us to improve the customer experience and the efficiency of our operations.
Additional information
DeepRed is a complex system, and there are a lot of factors that go into its decision-making process. However, here are some of the most critical factors:
- The location of the restaurant
- The location of the customer
- The time of day
- The availability of Dashers
- The distance between the restaurant and the customer
- The estimated time of arrival (ETA)
- The customer’s preferred delivery method
DeepRed uses these factors to calculate a score for each Dasher. The Dasher with the highest score is assigned the order.
DeepRed is constantly learning and improving. As we collect more data, DeepRed becomes better at predicting how to match Dashers with orders.
The original article is “Using ML and Optimization to Solve DoorDash’s Dispatch Problem.“