Delivery Time Range Prediction Across Online Order Stages
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current online concierge systems provide inaccurate and disconnected expected delivery range predictions across different stages of the order process, leading to increased cancellation rates and higher-than-expected delivery costs, and lack coherent delivery presentation logic.
Innovation Solution
Implementing computer models to predict delivery times at various stages of the order process, optimizing for metrics such as conversion, revenue, and cost, and updating delivery time ranges accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If machine learning based expected delivery ranges are applied only at checkout, then the system complexity is reduced, but the user experience becomes disconnected and inaccurate across different order stages
Solution Approach 1:
The patent divides the order process into multiple distinct stages (home page, store page, cart, checkout) and applies delivery prediction logic independently at each stage. This segmentation allows the system to provide stage-appropriate delivery information without overwhelming complexity, resolving the contradiction between simplicity and consistency.
Solution Approach 2:
The patent performs delivery time predictions in advance at each order stage rather than waiting until checkout. By calculating and displaying expected delivery ranges preliminarily at home page, store page, and cart stages, the system ensures consistent user experience across all interactions while maintaining manageable complexity through modular prediction logic.
2Loss of information
If early/late based expected delivery ranges are used, then user expectations are managed, but cancellation rates increase and delivery costs become higher than expected
Solution Approach 1:
The patent changes the prediction parameters from simple early/late indicators to multi-metric predictions including conversion probability, revenue impact, and cost estimation. By optimizing delivery time ranges based on these refined parameters, the system maintains user expectation management while improving conversion rates and controlling delivery costs through data-driven adjustments.
Solution Approach 2:
The patent implements feedback loops where predicted delivery ranges are continuously refined based on actual order outcomes, cancellation patterns, and cost data. This feedback mechanism allows the system to learn from past performance and adjust predictions to balance user expectation management with conversion optimization, reducing the harmful effects of overly conservative estimates.
3Reliability
If accurate delivery time predictions are provided across all order stages, then user experience improves, but computational resources and system complexity increase
Solution Approach 1:
The patent implements dynamic prediction logic that adapts computational effort based on the order stage and user context. At early stages like home page, simpler predictions are used, while more sophisticated multi-metric predictions are applied at later stages like checkout where accuracy is most critical. This dynamic approach maintains high prediction accuracy where needed while managing overall computational complexity.
Data Source
AI summary
One or more trained computer models are used to determine, at different stages of an order, an estimated time range for delivery of the order at an online system. The online system retrieves a set of candidate ranges of delivery times for the order. The online system applies the one or more computer models trained to predict a value of a metric for each candidate range in the set of candidate ranges, based on one or more features associated with the order. The online system selects a range of delivery times for the order from the set of candidate ranges, based on the predicted value of the metric for each candidate range. The online system causes a device of the user to display a user interface with the selected range of delivery times for the order.