Explore/Exploit Ad Layer for Dynamic Creative Combination Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing recommender systems for online content delivery, such as ad serving, lack an effective mechanism to maximize revenue through dynamic creative optimization by dynamically selecting ad combinations based on user characteristics and display environments.
Innovation Solution
Implementing an explore/exploit layer (EEL) at the frontend ad serving engine to store and dynamically draw combinations of ad assets optimized for predicted performance using machine learning models, ensuring each ad combination is tailored to specific user segments and display contexts.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional ad selection models are used at the backend server, then ad selection speed is maintained, but ad revenue and click-through rates are suboptimal due to lack of dynamic creative optimization
Solution Approach 1:
The system segments the ad serving functionality by introducing a separate explore/exploit layer at the frontend that handles dynamic creative optimization independently from the backend ad selection process. This allows the backend to maintain simple ad selection logic while the frontend layer handles complex combination optimization, resolving the contradiction between improved revenue and system complexity.
Solution Approach 2:
The explore/exploit layer acts as an intermediary between the backend ad selection model and the final ad rendering. It receives ad selections from the backend, dynamically optimizes asset combinations based on user characteristics and display environments, and serves the optimized combinations to users, thereby improving revenue without requiring complex changes at the backend.
2Productivity
If ad combinations are dynamically optimized based on user characteristics and display environments, then click-through rates improve, but real-time processing requirements and system complexity increase
Solution Approach 1:
The system pre-computes and stores optimal asset combinations in the explore/exploit layer during periods when ads are not being served. When an ad selection is made at the backend, the frontend layer can immediately retrieve and apply pre-optimized combinations based on user characteristics, avoiding real-time computation delays while maintaining high click-through rates.
Solution Approach 2:
The explore/exploit layer implements local optimization by maintaining separate combination distributions for different user segments and display environments. Each ad selection can be locally optimized based on specific user characteristics without requiring global real-time computation, reducing processing time while improving click-through rates for each segment.
3Speed
If combination distributions are stored at the frontend ad serving engine, then ad combination optimization speed improves, but memory requirements increase
Solution Approach 1:
The combination distributions are segmented by ad selection identifiers and user characteristics, allowing the frontend to store only relevant combinations for currently active ads and user segments. This reduces memory requirements while maintaining fast retrieval speeds for the specific combinations that are actually needed.
Solution Approach 2:
The system stores a partial set of combination distributions at the frontend - specifically, only those combinations relevant to currently active ad selections and user segments. Less frequently used combinations can be computed on-demand or retrieved from the backend, balancing memory usage with selection speed requirements.
Data Source
AI summary
The present teaching relates to displaying ads. An explore/exploit layer (EEL) is provided at frontend ad serving engine for storing combination distributions with respect to multiple ads. Each ad has multiple attributes. Each attribute can be instantiated using one of multiple assets. The frontend ad serving engine requests a recommended ad for bidding an ad display opportunity in a slot of a webpage viewed by a user on a user device. The recommended ad is one of the multiple ads. When the auction is successful, a combination of assets for the ad is drawn from the combination distributions in EEL and each of the assets instantiates a corresponding attribute of the ad. The combination is transmitted to the user device to render the ad.