Facet Ranking Normalization for Positional Bias
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The placement of facets on a website affects their selection frequency, leading to erroneous conclusions about the importance of product attributes, as facets higher in the list are more frequently selected, regardless of their actual importance.
Innovation Solution
A system that determines facet rankings by analyzing selection data across different positions and using rules to estimate selection rates if facets were in a different position, normalizing facet scores to account for positional bias, and displaying facets in an order based on their calculated importance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If facets are displayed in a list with varying positions, then users can easily browse and select product attributes, but facets higher in the list are selected more frequently leading to erroneous conclusions about attribute importance
Solution Approach 1:
The system applies preliminary anti-action by implementing a normalization process that counteracts the positional bias before measurement. The normalization factor is calculated based on the position of each facet, and this factor is used to adjust the selection counts to compensate for the inherent advantage that higher-positioned facets have in being selected. This preliminary correction prevents the erroneous conclusion about attribute importance.
Solution Approach 2:
The system changes the parameter of facet selection measurement by introducing a normalization factor that transforms the raw selection counts into position-adjusted selection counts. This parameter change allows the system to maintain the ease of faceted browsing while obtaining accurate measurements of facet importance that are not distorted by positional effects.
2Loss of information
If facet selection frequency is used to determine attribute importance, then customer preferences can be identified, but positional bias causes less important attributes to appear more important
Solution Approach 1:
The system applies preliminary anti-action by calculating a normalization factor based on facet position and applying it to the selection counts before determining attribute importance. This counteracts the positional bias that would otherwise cause less important attributes to appear more important, ensuring that the final ranking reliably reflects true customer preferences rather than placement effects.
3Productivity
If facets are reordered based on raw selection data, then popular attributes can be prioritized, but the ordering is distorted by placement effects rather than true importance
Solution Approach 1:
The system changes the measurement parameter by transforming raw selection counts into normalized selection counts using a position-based normalization factor. This allows the system to prioritize product attributes based on their true importance to customers rather than their initial placement, improving the accuracy of attribute prioritization while maintaining productivity.
Data Source
AI summary
Systems and methods including one or more processing modules and one or more non-transitory storage modules storing computing instructions configured to run on the one or more processing modules and perform acts of displaying a plurality of facets of a product type on a website of an online retailer in a plurality of facet orders, determining a first individual number of times a facet was selected when in a first position, determining a second individual number of times the facet was selected when in a second position, estimating a first estimated number of times the facet would have been selected if the facet had been in the first position rather than the second position, determining a ranking of the plurality of facets, and coordinating displaying at least a portion of the plurality of facets on the website of the online retailer in an order of the ranking.