Probabilistic Ad Event Matching via Bayesian Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current digital advertising technologies cannot accurately connect and verify the association between Demand Side Platform (DSP) log events and Ad Server log events with a single real-world impression, hindering the ability to associate ad serve events with individual users or cohorts without violating privacy.
Innovation Solution
A system and method that creates independent geographic closeness and sole rightful heir factors from event log data, applying probability and combinatoric game theoretical analysis to match DSP and Ad Server log events, using deterministic and probabilistic record matching to segregate events into unit groups and define a time-difference window to filter and classify pairs as matches or unmatches.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deterministic and probabilistic record matching is applied to connect DSP and Ad Server log events, then the accuracy of associating ad serve events with individual users is improved, but the system complexity increases due to multiple matching algorithms and probability calculations
Solution Approach 1:
The patent segments the matching process into distinct phases: deterministic matching first filters candidate pairs using exact field matches, then probabilistic matching applies to remaining candidates. This segmentation allows each method to operate on appropriately sized data sets, improving accuracy while managing complexity through structured progression.
Solution Approach 2:
The patent introduces an intermediary probabilistic scoring mechanism that bridges deterministic matching results and final event association. The probabilistic model acts as a mediator that evaluates candidate pairs from deterministic matching, applying weighted factors to determine likelihood of true matches, thus resolving the contradiction between precision and complexity.
2Reliability
If probabilistic matching with multiple factors is used to verify event pairs, then the reliability of event association is improved, but the computational time and processing resources increase
Solution Approach 1:
The patent applies preliminary deterministic filtering before probabilistic analysis, pre-processing the data by eliminating obviously non-matching pairs through exact field comparisons. This preliminary action reduces the data set size before applying computationally intensive probabilistic calculations, thereby maintaining reliability while reducing processing time.
Solution Approach 2:
The patent implements a two-tiered matching approach where deterministic matching provides a sufficient filter for obvious cases, and probabilistic matching is applied only partially to remaining candidates. This partial application of the more complex method achieves adequate reliability without the full computational cost of applying probabilistic matching to all possible pairs.
3Quantity of substance
If event logs from multiple sources are connected to single real-world impressions, then the quantity of available data for analysis is improved, but the difficulty of detecting and measuring accurate matches increases
Solution Approach 1:
The patent implements feedback mechanisms through probabilistic scoring that evaluates the quality of matches. The system calculates likelihood scores for candidate pairs and uses these feedback signals to refine matching decisions. This feedback approach enables the system to handle increased data quantity from multiple sources while maintaining verification accuracy through quantitative evaluation.
Solution Approach 2:
The patent transforms the verification problem by changing parameters from binary match/non-match to probabilistic scoring with multiple weighted factors. This parameter transformation allows the system to accommodate data from multiple sources with different characteristics, measuring match quality through configurable weights rather than rigid criteria, thus reducing verification difficulty.
4Productivity
If time-difference window filtering is applied to reduce search space, then the productivity of matching process is improved, but the measurement precision may be affected by arbitrary window selection
Solution Approach 1:
The patent applies dynamic time-difference windowing where the window parameters can be adjusted based on observed data characteristics and matching results. Rather than using fixed arbitrary windows, the system can adapt window sizes and offsets to match actual event timing patterns, maintaining productivity benefits while reducing precision loss from arbitrary parameter selection.
Data Source
AI summary
A system and method for accurately matching corresponding DSP event data and Ad-Server event data with associated with a single real-world ad serve event by (a) pairing DSP event data and Ad-Server event data into data pairs, (b) comparing various field data in associated source fields from each of the DSP event data and Ad-Server event data to determine if the field data is a match or unmatch, and (c) based on the likelihood that a match of field data in a particular source field indicates an overall event match, which is determined using a Bayesian analysis, determining the probability that the DSP event data and Ad-Server event data in the data pair truly corresponding to the same single real-world ad serve event.