Marketing Content Attribution Using AI Element Decomposition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current marketing analytics methods, such as MMM and MTA, lack granularity in attributing performance to specific elements of marketing content like CTAs and headlines, and insights are not centrally stored, leading to loss and accessibility issues.
Innovation Solution
A system utilizing computer vision and generative AI models to decompose and analyze marketing assets, generating structured data sets for training prediction models to provide insights and predict performance of new content.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If Marketing Mix Modeling (MMM) or Multi-Touch Attribution (MTA) is used to analyze marketing campaigns, then overall campaign effectiveness can be measured, but granular attribution to specific content elements (CTA, headline, objects, colors) is lost
Solution Approach 1:
The patent segments marketing content into discrete visual elements (CTA, headline, objects, colors, layout) that can be individually analyzed. Computer vision models decompose ad images into these constituent parts, allowing granular attribution of performance to each element rather than treating the entire ad as a single unit.
Solution Approach 2:
The patent introduces computer vision models and generative AI as intermediary tools between the raw marketing content and the attribution analysis. These intermediaries automatically extract and structure visual elements, bridging the gap between complex ad creatives and measurable performance attributes.
2Loss of information
If marketing insights are manually captured in PowerPoint presentations, then they can be presented in meetings, but insights are not accessible to non-attendees and can be lost over time
Solution Approach 1:
The patent creates digital copies of marketing insights in the form of structured datasets and trained AI models. These digital replicas can be stored, accessed, and queried indefinitely without degradation, replacing the ephemeral nature of verbal presentations while maintaining the core analytical value.
Solution Approach 2:
The patent transforms marketing insights from a static presentation format into dynamic, queryable parameters within a database. This allows users to retrieve and analyze specific aspects of the insights on demand, changing the delivery mechanism from passive presentation to active exploration.
3Loss of information
If traditional marketing analytics methods are used, then overall performance metrics can be obtained, but actionable insights at the element level are unavailable
Solution Approach 1:
The patent segments marketing content into discrete visual elements (CTA, headline, objects, colors, layout) that can be individually analyzed. Computer vision models decompose ad images into these constituent parts, allowing granular attribution of performance to each element rather than treating the entire ad as a single unit.
Solution Approach 2:
The patent applies partial action by focusing analysis only on the key constituent elements of marketing content that are most likely to drive performance. Rather than analyzing every possible attribute of each ad, the system selectively extracts and analyzes the most impactful elements (CTA, headline, primary objects, dominant colors).
Data Source
AI summary
Systems and methods are provided to generate marketing content insights. In one embodiment, a disclosed method includes receiving input data including one or more advertising assets and one or more metrics corresponding to the one or more advertising assets; using a computer vision model, identifying one or more constituent elements of the one or more advertising assets; appending the one or more constituent elements and/or descriptions of the one or more constituent elements to the input data to generate a structured data set; training a generative AI model using the structured data set; training a prediction model using the structured data set; based on the trained generative AI model and a prompt, generating an additional advertising asset not included in the one or more advertising assets; and based on the trained prediction model, generating at least one predicted metric of the additional advertising asset.