Multi-Item Retargeting Model for Online Advertising Bids
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing online advertising systems often fail to effectively utilize retargeting strategies by focusing on single items, leading to lower conversion rates, as they lack the ability to dynamically select and combine multiple relevant items in real-time for personalized advertising.
Innovation Solution
A system that employs a machine-learned retargeting model to select and combine multiple items for advertisements based on consumer behavior data, such as previous interactions and item values, to generate bids for online advertising slots, optimizing the selection process for higher conversion rates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If a single item is used in retargeting advertisements, then the advertising system is simple to operate, but the conversion rate is lower
Solution Approach 1:
The patent combines multiple items into a single retargeting advertisement presentation, merging several product recommendations into one cohesive ad unit. This allows the system to maintain operational simplicity while improving conversion rates by presenting multiple relevant items to the consumer simultaneously, thereby resolving the contradiction between ease of operation and conversion rate effectiveness
Solution Approach 2:
The patent segments the item selection process into distinct functional components: a retargeting model that identifies relevant items based on consumer behavior, a bidding component that determines bid amounts for each item, and an advertisement generation component that assembles the final presentation. This segmentation enables complex multi-item retargeting while maintaining ease of operation through automated processing
2Reliability
If multiple items are selected and combined dynamically, then the conversion rate increases, but the system complexity increases
Solution Approach 1:
The patent implements preliminary action by pre-training the retargeting model offline using historical consumer behavior data before real-time advertising operations. This pre-computed model is then reused during live bidding, eliminating the need for complex real-time calculations and reducing system complexity while maintaining high conversion rates through accurate item selection
Solution Approach 2:
The patent introduces an intermediary retargeting model that acts as a bridge between raw consumer behavior data and advertisement generation. This model processes and transforms complex behavioral patterns into simplified item recommendations, reducing the computational burden on the bidding system while improving conversion rate through data-driven item selection
3Reliability
If real-time bid generation is performed, then the advertising is personalized and effective, but the processing time increases
Solution Approach 1:
The patent performs preliminary training of the retargeting model offline using historical data, creating a pre-computed model that can be rapidly applied during real-time bidding. This separates the computationally intensive model training from time-critical bid generation, maintaining advertising effectiveness while reducing processing time to meet real-time requirements
Solution Approach 2:
The patent implements a dynamic system where the retargeting model is trained offline but applied dynamically during real-time bidding operations. The system adapts to changing consumer behavior through periodic model retraining while maintaining rapid response times during live advertising auctions, balancing personalization effectiveness with processing speed requirements
Data Source
AI summary
This disclosure describes systems, methods, and computer-readable media related to retargeting online advertisement campaign recommendations for advertisements with multiple items or services. Bids may be based on a combined advertisement creative comprising two or more items or services. Dynamically selecting multiple items at bid time using a retargeting model to determine a potential revenue generation amount associated with an event may increase the probability of a conversion event based on the creative that includes the selected items. In some embodiments, a machine-learned retargeting model may be used to select multiple items to be displayed in an advertisement. The retargeting model may be applied to items that were previously viewed by the consumer and may determine a value for each of the items using factors. A bid may be calculated for each of the selected items using the values determined by the retargeting model.