Ad Distribution System Using Bid Unit Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for selecting online ad requests struggle with data sparsity, variance, and volume issues, leading to inefficient ad placement and budget mismanagement in online advertising systems.
Innovation Solution
The system uses performance metric data from test ads to aggregate and rank bid units, addressing data sparsity by grouping units with insufficient data, stabilizing against outliers, and optimizing budget allocation based on predicted spend, ensuring effective ad distribution.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If contextual advertising approaches are used to identify online ad requests by analyzing webpage text, then ad relevance is improved, but processing time and computational resources increase significantly
Solution Approach 1:
The system performs preliminary classification of bid units into segments (high, medium, low performance) using historical data before real-time ad requests arrive. This pre-processing allows the system to quickly retrieve and use pre-analyzed performance metrics during actual ad serving, avoiding the need to perform complex webpage text analysis in real-time.
Solution Approach 2:
The patent introduces bid units as an intermediary layer between webpage context and ad selection. Instead of directly analyzing webpage text for each ad request, the system uses bid units that pre-encode performance characteristics, serving as a mediator that simplifies the matching process between ad requests and suitable ads.
2Adaptability or versatility
If behavioral advertising approaches are used to identify user characteristics, then ad personalization is improved, but data sparsity reduces the reliability of user interest deduction
Solution Approach 1:
The system merges multiple bid units with similar characteristics into broader segments to accumulate sufficient data. By combining data from multiple sources and dimensions, the system overcomes individual data sparsity issues and achieves statistically reliable user interest profiles for personalized advertising.
Solution Approach 2:
The patent introduces additional dimensions for analyzing ad request performance beyond simple user behavior, including bid unit characteristics, segmentation metrics, and multi-dimensional performance data. This dimensional expansion allows the system to deduce user interests more reliably by cross-validating across multiple data dimensions.
3Productivity
If existing techniques select online ad requests based on simple metrics, then processing speed is maintained, but they cannot account for data variance, sparsity, and volume issues
Solution Approach 1:
The system segments bid units into distinct performance categories (high, medium, low) based on historical data analysis. This segmentation allows the system to apply different processing strategies to different segments, maintaining high processing speed for well-understood segments while investing more computational resources in analyzing less certain segments.
Solution Approach 2:
The patent dynamically adjusts performance metric calculations by changing parameters such as data weighting, segmentation thresholds, and confidence levels based on data availability and quality. This allows the system to maintain processing speed while improving metric accuracy by adapting its calculation methodology to the specific characteristics of available data.
4Area of stationary object
If marketers bid on all available ad requests, then ad coverage is maximized, but budget is exhausted too quickly on high-volume low-yield traffic
Solution Approach 1:
The system applies different bidding strategies to different bid unit segments based on their specific characteristics. High-performing segments receive aggressive bidding to maximize coverage, while low-performing segments receive reduced or no bidding. This localized quality approach ensures budget is allocated efficiently across different traffic sources and ad requests.
Solution Approach 2:
The patent implements partial action by selectively bidding on only a portion of available ad requests - specifically those in high and medium performance segments - rather than bidding on all requests. This partial participation strategy prevents budget exhaustion on low-yield traffic while maintaining adequate coverage on high-value opportunities.
Data Source
AI summary
Techniques for distributing online ads by targeting online ad requests using test data to predict performance. The techniques can target ad requests in automated online advertising systems in which ad requests are generated by an ad exchange server and bids are placed by marketer devices in real time. The techniques aggregate bid units and compare bid unit characteristics to select bid units to target in ways that address data sparsity, variance, and volume issues. Data sparsity issues are addressed by aggregating bid units to avoid using bid units having insufficient data. Data variance issues are addressed by computing stability metrics for bid units that enable discounting the effect of outliers. Data volume and processing efficiency issues are addressed by grouping similar bid units based on similar metrics (e.g., normalized ROI) and/or similar stability scores, and then ranking the bid units and selecting the top ranked bid units to target.