Variant-Level Search Ranking Using User Activity Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing search engine technologies are deficient in accurately ranking search results for items with multiple variants due to inconsistencies in user activity data, leading to inaccurate rankings and increased computer resource consumption.
Innovation Solution
Generate user activity data for each variant of an item and rank listings based on individual variant data, rather than the entire listing, to improve accuracy and reduce computational overhead.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If user activity data is aggregated at the listing level for ranking, then computational operations are simplified, but ranking accuracy deteriorates due to inconsistencies among variant data
Solution Approach 1:
The patent segments the aggregated listing-level user activity data into individual variant-level data. Instead of treating all variants uniformly, the system divides the data by variant identifiers, allowing separate analysis and ranking of each variant based on its specific user activity patterns. This segmentation resolves the contradiction by enabling accurate variant-specific rankings while maintaining manageable computational complexity through structured data organization.
Solution Approach 2:
The patent applies local quality by treating each variant differently based on its specific characteristics. Rather than applying a uniform ranking approach to all variants, the system customizes the ranking process for each variant based on its individual user activity data, ensuring that variants with different performance metrics are ranked appropriately according to their own quality indicators.
2Reliability
If user activity data from all variants is aggregated for ranking, then data completeness is improved, but ranking accuracy deteriorates due to inconsistent variant performance
Solution Approach 1:
The system segments the complete user activity data by variant identifiers, preserving all data points while enabling differentiated analysis. This allows the system to maintain data completeness across all variants while simultaneously achieving ranking accuracy by evaluating each variant's specific performance metrics separately rather than aggregating them into a misleading overall score.
Solution Approach 2:
Instead of aggregating variant data upward to listing level and then attempting to derive variant rankings, the patent inverts the approach by directly analyzing individual variant data from the ground up. This inversion allows the system to maintain complete data usage while achieving accurate variant-specific rankings by starting with individual variant performance rather than attempting to decompose aggregated data.
3Measurement precision
If ranking is performed based on individual variant data, then ranking accuracy is improved, but computational resource consumption increases
Solution Approach 1:
The patent segments the computational process into efficient, modular operations that work on individual variant data. By organizing computations around variant-level segments rather than requiring complex aggregated calculations, the system achieves accurate variant rankings while optimizing resource usage through structured, scalable computational approaches that can be efficiently implemented.
Solution Approach 2:
The system changes the parameters of computation by working directly with variant-level user activity metrics rather than performing complex transformations from aggregated listing-level data. This parameter change enables accurate variant ranking while reducing computational overhead by using the natural granular structure of the data as the basis for ranking calculations.
Data Source
AI summary
A first variant of a listing and a second variant of the listing are received. The listing describes an item for sale in an electronic marketplace. The first variant describes a different iteration of the item relative to the second variant. First user activity data of the first variant is generated and second user activity data of the second variant is generated. The first user activity data corresponds to user input metrics associated with the first variant. The second user activity data corresponds to user input metrics associated with the second variant. A search engine receives a first query. Based at least in part on the first user activity data relative to the second user activity data, a first search result associated with the first variant is ranked higher than a second search result associated with the second variant.