Real-Time Machine Learning Model for Re-Order Item Recommendations
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Users face inefficiencies in re-ordering items online due to time-consuming browsing across multiple pages and the likelihood of forgetting preferred items, especially since more than half of items in online orders are re-ordered items.
Innovation Solution
A system and method utilizing machine learning to analyze historical interaction data, identify shopping journeys, and build real-time models to recommend re-order items that satisfy price thresholds, re-ranking items based on attributes and transmitting these recommendations through a graphical user interface during the user session.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Loss of time
If users manually browse through multiple pages to locate and add re-order items, then they can find and add items to their order, but the process becomes time-consuming and users may forget items
Solution Approach 1:
The system performs preliminary actions by automatically analyzing historical interaction data before the user needs to re-order. It pre-identifies items that the user is likely to re-order based on past behavior patterns, so when the user wants to re-order, the items are already prepared and presented to them, eliminating the need to manually browse through multiple pages.
Solution Approach 2:
The system enables self-service by automatically generating re-order recommendations based on the user's historical interaction data without requiring manual input. The system serves itself by using its own stored data about the user's preferences and purchase history to autonomously identify and present relevant re-order items.
2Measurement precision
If the system builds real-time machine learning models to recommend items, then recommendation accuracy improves, but computational resources and processing time increase
Solution Approach 1:
The system performs preliminary actions by pre-processing and storing historical interaction data in a structured format before real-time recommendation is needed. This preprocessing includes organizing user interactions, item attributes, and purchase histories in advance, so that when real-time recommendations are required, the machine learning model can quickly access pre-organized data without extensive computational overhead.
Solution Approach 2:
The system applies local quality by building specialized machine learning models tailored to different user segments, shopping journeys, and basket types rather than using a single generic model. Each model is optimized for specific local conditions and user characteristics, improving recommendation accuracy for particular contexts while allowing the system to efficiently handle diverse recommendation scenarios without requiring one overly complex universal model.
Data Source
AI summary
Systems and methods including one or more processors and one or more non-transitory computer readable media storing computing instructions that, when executed on the one or more processors, cause the one or more processors to perform: receiving historical interaction information corresponding to a user in a marketplace; identifying a shopping journey and a basket type for the user based on the cart context and items in a cart for the user for a current user session; identifying a price threshold for the cart for the user; building a machine learning model for the current user session in real-time based on the historical interaction information, the cart context, the basket type and the price threshold to determine a ranking of new items to display to the user to add to the cart for the current user session, wherein the new items satisfy the price threshold; re-ranking the ranking of the new items to display to the user in the current user session based on item attributes of the new items; and transmitting the re-ranked ranking of the new items to the user via a graphical user interface (GUI) during the current user session. Other embodiments are disclosed.