Geolocation Analytics Segmentation for Bias Reduction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional geolocation analytics platforms are not well-suited for performing complex analyses on large data sets, leading to biases and misleading results due to simplifying assumptions, particularly in stochastic analyses of user behavior on networks, where group-to-group variation can overwhelm treatment effects and selection biases are difficult to manage.
Innovation Solution
A process involving obtaining device identifiers and geolocations, assigning them to treatment or control collections based on hash values, and determining visitation rates to places of interest, which helps distinguish content effects from targeting effects and mitigates selection biases, enabling analysis of complex multi-group populations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional geolocation analytics platforms perform complex analyses on large data sets, then measurement precision improves, but device complexity and computational resources required increase significantly
Solution Approach 1:
The patent segments the population into multiple groups based on shared attributes and performs separate analyses for each group rather than treating the entire population as a single homogeneous unit. This segmentation allows for more precise measurement within each group while managing computational complexity through structured organization of the analysis process.
Solution Approach 2:
The patent performs preliminary actions by pre-defining groups based on shared attributes and pre-organizing the population structure before conducting the actual treatment effect analysis. This preliminary organization enables more efficient processing during the analysis phase and reduces computational burden.
2Ease of operation
If simplifying assumptions are made to make analysis tractable, then ease of operation improves, but measurement precision deteriorates due to biases and misleading results
Solution Approach 1:
By segmenting the population into homogeneous groups based on shared attributes, the patent reduces the need for simplifying assumptions within each group. The segmentation ensures that individuals within the same group are more similar, reducing variability and allowing for more accurate treatment effect estimation without requiring aggressive simplifying assumptions.
Solution Approach 2:
The patent applies local quality by making analysis assumptions specific to each group rather than applying uniform assumptions across the entire population. Each group can have tailored analysis approaches that are appropriate for its specific characteristics, improving measurement precision while maintaining ease of operation through standardized group-level procedures.
3Adaptability or versatility
If group-to-group variation is present in the population, then adaptability improves, but measurement precision worsens as group variation overwhelms treatment effects
Solution Approach 1:
The patent directly addresses group-to-group variation by segmenting the population into distinct groups based on shared attributes. By performing separate analyses for each group, the patent isolates treatment effects within homogeneous sub-populations, preventing group variation from overwhelming the treatment effects. This segmentation allows the system to adapt to population diversity while maintaining measurement precision within each group.
4Ease of operation
If selection biases are present in the data sample, then ease of operation improves by using available data, but measurement precision deteriorates due to skewed results
Solution Approach 1:
The patent segments the population into groups based on shared attributes that can be identified from available data, even when selection biases are present. By analyzing treatment effects within each group separately, the patent can identify and account for selection biases that may affect different groups differently. This segmentation allows for more robust analysis that acknowledges data availability constraints while improving measurement precision through group-specific adjustments.
Data Source
AI summary
Provided is a process including: obtaining device identifiers of a population of user computing devices; obtaining groups of the users computing devices obtaining one or more places of interest; assigning user computing devices to either a treatment collection xor a control collection based on hash values of the device identifiers; directing application of the treatment according to the assignment; obtaining geolocations visited by the user computing devices; assigning the geolocations to either the treatment collection xor the control collection based on hash values of device identifiers associated with the geolocations; assigning the geolocations to one or more of the groups based on the device identifiers associated with the geolocations; and for each group, determining a respective amount of visits to at least some of the one or more places of interest attributable to the treatment based on the geolocation assignments.