Dynamic Frequency Cap for Ad Influence Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for selecting audiences for online advertising fail to differentiate between entities likely to be influenced by advertisements and those who may be negatively influenced, leading to inefficient ad exposure and potential over-exposure.
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 blacklists/whitelists, and optimized bid responses to focus ad spending on likely influencers.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If frequency caps are used to limit ad exposures, then over-exposure is prevented, but the likelihood that an advertisement will influence an individual customer is not accounted for
Solution Approach 1:
The patent changes the parameter from a fixed frequency cap to a dynamic individual frequency cap based on predicted influence likelihood. Entities with higher predicted influence likelihood receive higher frequency caps, while those with lower predicted influence likelihood receive lower frequency caps or are excluded entirely. This resolves the contradiction by making the frequency cap adaptive to individual characteristics rather than uniform.
Solution Approach 2:
The patent segments the audience into different groups based on predicted influence likelihood using behavioral models. Entities are classified into segments such as high influence likelihood, low influence likelihood, and neutral segments. This segmentation allows different frequency cap strategies to be applied to different segments, preventing over-exposure for some while maintaining exposure opportunities for others.
2Productivity
If advertisements are sent to customers who already intend to purchase, then conversion is achieved, but over-exposure may deter the customer from making a purchase
Solution Approach 1:
The patent performs preliminary assessment of entity characteristics using behavioral models before ad delivery to predict influence likelihood. This preliminary action allows the system to identify entities that are likely to be positively influenced by ads versus those who may be deterred by over-exposure. By acting in advance with this prediction, the system can adjust frequency caps or exclude entities before harmful over-exposure occurs.
Solution Approach 2:
The patent uses feedback from conversion data and entity behavior to continuously refine the behavioral models and improve prediction accuracy. Conversion data from treated and control groups provides feedback that updates the influence likelihood predictions, enabling better differentiation between entities that will convert and those that may be deterred by over-exposure.
3Productivity
If behavioral models are used to assess suitability for advertising, then conversion likelihood is inferred, but the distinction between entities likely to be influenced and those likely to convert is not made
Solution Approach 1:
The patent segments the population into distinct groups based on predicted influence likelihood using behavioral models that analyze entity characteristics. This segmentation creates separate segments for entities likely to be positively influenced by ads versus those unlikely to be influenced, enabling differentiated ad delivery strategies that account for both conversion potential and influence likelihood.
Solution Approach 2:
The patent applies local quality by tailoring frequency caps and ad delivery strategies to individual entities or small groups of entities based on their specific predicted influence likelihood. Rather than applying a uniform strategy to all entities, the system adjusts parameters locally for each entity or segment, optimizing both conversion and influence considerations for each local group.
Data Source
AI summary
An influence system for predicting advertisement impact for audience selection. An advertising probe campaign is operated by sending an advertisement to each entity in a treatment group of entities. A control group of entities which excludes the treatment group entities is selected and no campaign advertising content is sent to the treatment group entities. An influence model is created by comparing features of the treatment group converters to features of the control group converters. An individual frequency cap is selected for each entity that is a candidate for the advertising campaign based on a result of applying the influence model to the features of the candidate entity. The entity may be selected to receive an advertisement based on the individual frequency cap. Some embodiments are integrated with a real time bidding (RTB) exchange and a bid response may be configured based on the results of applying the influence model.