Ad Conversion Attribution via Machine Learning Models
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional deterministic attribution methods in advertising measurement face challenges when persistent user identities, such as cookies, are not available, leading to difficulties in associating advertisement views with conversion events.
Innovation Solution
The development of systems and methods that create impression-level attribution models without requiring user identifiers, using deep learning models to predict conversion events and provide reports on advertisement effectiveness, even in the absence of user identifiers, by accounting for data distribution shifts through techniques like covariate shift modeling and importance sampling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If deterministic attribution methods using persistent user identities are used, then attribution accuracy is improved, but the system becomes unreliable when user identifiers are unavailable
Solution Approach 1:
The patent introduces machine learning models as an intermediary between advertisement impressions and conversion events. Instead of directly linking them through user identifiers, the model learns patterns and relationships from training data to predict conversions from impressions, enabling attribution without requiring persistent user identities.
Solution Approach 2:
The patent replaces the mechanical deterministic linking system (based on user identifiers and direct matching) with a statistical machine learning system. The model uses learned patterns from historical data to probabilistically associate impressions with conversions, substituting direct mechanical linkage with statistical inference.
2Ease of operation
If traditional deterministic attribution is used, then user-level tracking is achieved, but the system cannot handle cases without user identifiers
Solution Approach 1:
The patent creates a universal attribution system that handles both identified and unidentified users through the same machine learning model. The model processes advertisement impression data regardless of whether user identifiers are present, making the system multi-functional and adaptable to different data scenarios.
Solution Approach 2:
The patent changes the fundamental parameter of attribution from deterministic user-level matching to probabilistic impression-level prediction. By shifting from requiring user identifiers to using model-based probability estimates, the system adapts to work with or without user identification data.
3Measurement precision
If machine learning models are trained on data with user identifiers, then model accuracy is improved, but the model must be adapted to work with data without user identifiers
Solution Approach 1:
The patent performs preliminary training of the machine learning model using data with user identifiers to establish accurate prediction patterns. The model learns from this labeled training data and then applies the learned patterns to predict conversions from impressions in scenarios where user identifiers may not be available, eliminating the need for separate model adaptations.
Data Source
AI summary
Systems, methods, and computer-readable media are disclosed for modeled advertisement conversion attributions. An example method may include receiving first input data comprising first advertisement impression data and first advertisement conversion data, wherein the first input data includes one or more user identifiers associated with both the advertisement impression data and advertisement conversion data. The example method may also include training one or more machine learning models using the first input data. The example method may also include receiving second input data comprising second advertisement impression data, wherein user identifiers are unavailable for the second input data. The example method may also include determining, using the one or more machine learning models, second predicted conversion data associated with the second input data.