DoorDash has been granted a patent for a system that generates dynamic estimated time of arrival (ETA) predictive updates for the delivery of perishable goods. The system uses trained predictive models that consider factors such as historical restaurant data and courier performance to generate ETA predictions for different delivery events. The models are continuously trained and updated based on received timestamps. The system also includes a neural network that automatically generates ETA predictions based on user input. GlobalData’s report on DoorDash gives a 360-degree view of the company including its patenting strategy. Buy the report here.
According to GlobalData’s company profile on DoorDash, supply chain management systems was a key innovation area identified from patents. DoorDash's grant share as of September 2023 was 77%. Grant share is based on the ratio of number of grants to total number of patents.
Dynamic estimated time of arrival predictive updates for delivery
A recently granted patent (Publication Number: US11755906B2) describes a server that utilizes a neural network to generate estimated time of arrival (ETA) predictions for a series of events in an order. The server includes a processor and memory, with the processor being configured to train the neural network using a training dataset and receive confirmation messages from a user device. The neural network consists of multiple subnetworks, each corresponding to the duration between two successive events in the order.
The training dataset used to train the neural network includes a series of training events with known time durations between them. The processor inputs different combinations of training events and compares the generated ETA predictions with the known time durations to iteratively adjust the weighted factors in the neural network. This allows the neural network to dynamically output accurate ETA predictive updates.
The server can receive multiple confirmation messages from the user device, each containing information corresponding to a different event in the order. The processor then inputs this information into the neural network to update the ETA predictions. The weighted factors used in the neural network include various parameters such as time, date, weather, number of dishes in a restaurant order, sub-total of a restaurant order, number of pending orders at an associated merchant, historical restaurant data, historical courier performance, and size of markets.
The user device can be a customer device, a courier device, or a merchant device. The ETA predictions generated by the server can be provided to the customer device, allowing them to track the progress of their order. Additionally, the ETA predictions can be used to pair delivery opportunities with a courier corresponding to the courier device.
The series of events in the order includes the placement of the order by a customer, the confirmation of completion of the order by the merchant, and the pickup of the order from the merchant by a courier or the delivery of the order to the customer. The neural network includes subnetworks that correspond to the duration between these events.
The patent also covers a non-transitory computer-readable medium storing programs configured to execute the same functionalities as the server. Furthermore, a method is described that involves training the neural network, receiving confirmation messages, and generating ETA predictions based on the information provided.
Overall, this patent presents a server system that utilizes a neural network to generate accurate ETA predictions for a series of events in an order, considering various factors that may affect the delivery time.