Online Advertising Impression Estimation via Panel Data Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for estimating impression characteristics in online tracking technologies fail to accurately distinguish between human and machine-generated browsing activities, leading to inaccurate impression data and irrelevant advertising.
Innovation Solution
A system and method that collects and filters raw impression data from browser extensions to eliminate spurious machine activity, using a reference tag and statistical projection to estimate global impression volumes and characteristics, ensuring data quality and relevance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Quantity of substance
If machine-generated browsing activity is included in impression data collection, then data volume increases, but measurement precision deteriorates due to irrelevant machine activity contaminating human browsing data
Solution Approach 1:
The patent extracts and removes machine-generated browsing activity from the collected impression data through filtering mechanisms. Browser extensions identify and eliminate impressions generated by crawlers, bots, and automated systems, retaining only human-generated browsing data for accurate measurement.
Solution Approach 2:
The patent introduces browser extensions as intermediary components that mediate between the tracking ecosystem and data collection. These extensions act as filters that distinguish human from machine activity and selectively pass relevant impressions to the measurement system.
2Adaptability or versatility
If tracking technologies load on all browsing activity, then data collection coverage increases, but reliability deteriorates due to inclusion of non-human browsing activity
Solution Approach 1:
The patent extracts machine-generated impressions from the total impression data and removes them from the dataset. This extraction process ensures that only human browsing activity contributes to impression characteristics measurements, maintaining reliability while preserving broad coverage of human user behavior.
Solution Approach 2:
The patent implements feedback mechanisms where browser extensions continuously monitor and identify machine-generated activity patterns. The system uses feedback from tracking technologies to distinguish between human and machine browsing, dynamically adjusting data inclusion based on the source of each impression.
3Object-affected harmful factors
If privacy settings block certain tracking technologies, then user privacy is protected, but measurement precision deteriorates due to incomplete impression data from panel users
Solution Approach 1:
The patent introduces browser extensions as intermediaries that operate within privacy constraints. These extensions collect impression data while respecting user privacy settings, acting as mediators between privacy protection requirements and data collection needs. The extensions capture necessary measurement data without violating user privacy preferences.
Solution Approach 2:
The patent uses panel user data as a representative copy or sample of the broader internet population. By collecting precise impression data from panel users with browser extensions, the system creates a measurable subset that statistically represents overall browsing behavior, enabling accurate measurements without requiring complete data from all internet users.
Data Source
AI summary
A system for accurately estimating global impression volumes at tag and host levels, and for computing global page views and reach estimates, collects impression volumes for various tracking technologies and for several hosts from a panel of users. The collected panel data is normalized, e.g., to minimize spurious and/or non-human activity data. A scaling factor is computed using measured global impression volume of a reference tag and impression volume of the reference tag with respect to the panel. The required global estimates are obtained by scaling the normalized panel data using the computed scaling factor.