Ad Influence Model for Dynamic Frequency Cap Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current online advertising technologies fail to effectively differentiate between entities that can be positively influenced by advertisements and those that may be negatively influenced, leading to over-exposure and inefficient ad spending.
Innovation Solution
A system that creates an influence model by comparing features of entities exposed to advertisements that converted with those that did not, allowing for individual frequency caps, targeting, and bid pricing adjustments to optimize ad delivery based on the likelihood of influence.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If frequency caps are specified to prevent over-exposure, then advertising exposure is limited, but the likelihood of influence for individual customers is not taken into account
Solution Approach 1:
The patent changes the parameter from a fixed frequency cap to a dynamic frequency cap that is adjusted based on the predicted likelihood of influence for each entity. The system calculates an influence score using machine learning models that consider entity features, advertisement characteristics, and historical data, then sets individualized frequency caps proportional to this score. This resolves the contradiction by making the frequency cap adaptive rather than static, allowing precise control over exposure while accounting for individual influence likelihood.
2Quantity of substance
If advertisements are sent to all potential customers, then coverage is maximized, but spending on unresponsive individuals increases
Solution Approach 1:
The patent applies local quality by treating each entity differently based on its specific characteristics and predicted responsiveness. Instead of uniform advertising coverage, the system calculates an influence score for each entity using machine learning models that consider individual features, and then tailors the advertising strategy (frequency cap, bid price, targeting) to each entity's predicted likelihood of being influenced. This ensures coverage is optimized for each local case rather than applying a blanket approach.
Solution Approach 2:
The patent implements partial action by selectively applying advertising efforts only to entities with high predicted influence scores. The system uses the machine learning model to identify a subset of entities most likely to be positively influenced, and concentrates advertising resources on this partial group rather than attempting to reach all potential customers. This resolves the contradiction by making advertising spending efficient through selective partial coverage.
3Ease of operation
If behavioral models are used to assess conversion likelihood, then bidding decisions are informed, but the distinction between conversion likelihood and influence likelihood is not made
Solution Approach 1:
The patent segments the assessment into two distinct machine learning models: one that predicts conversion likelihood and another that predicts influence likelihood. The conversion model assesses the probability of purchase based on behavioral data, while the influence model specifically predicts whether an entity will be positively influenced by advertising exposure. This segmentation resolves the contradiction by preserving the distinction between these two different probabilities while still providing comprehensive information for bidding decisions.
Solution Approach 2:
The patent introduces an intermediary influence score that bridges the gap between conversion likelihood and actual advertising effectiveness. The influence model acts as an intermediary assessment layer that evaluates whether an entity is likely to be positively influenced by ad exposure, separate from but complementary to the conversion model. This intermediary measurement prevents loss of information by explicitly capturing the influence dimension that the conversion model alone cannot provide.
Data Source
AI summary
An influence system for predicting advertisement impact for advertising response selection. A treatment group of entities is selected which have received an advertising treatment of a campaign. A control group of entities is selected which excludes the treatment group entities is selected which have not received the advertising treatment of the campaign. An influence model is created for each campaign by comparing features of the treatment group converters to features of the control group converters. A campaign is selected for an opportunity to expose a specified entity to advertising based on the result of applying each respective campaign's influence model to features of the specified entity.