Video Marketing Analysis

"AI-driven tool using Google Gemini to analyze marketing videos, identifying engagement hotspots and comparing ads for effectiveness—helping marketers optimize campaigns with actionable insights."

Comment

Video Analysis for Marketing Campaign Effectiveness

Category: Marketing & Video Analysis

Idea:
I developed a marketing analytics solution that leverages the Gemini API to analyze marketing campaign videos. The goal is to assess customer engagement—such as attention hotspots—and use AI to suggest improvements for future campaigns. I applied an AI Ethics Framework to ensure that all collected data respects user privacy.

Technical Implementation:
I used tools from video_analysis.ipynb to create a pipeline for extracting insights from marketing videos. By combining these insights with explainable AI, I generated comprehensive reports on engagement. The solution also uses function-calling to initiate different types of analyses, like sentiment and engagement tracking.

Implementation Steps:

  1. Project Setup:
    Defined project information and initialized Vertex AI to interact with Google Cloud services for efficient video data processing.

  2. Video Analysis Pipeline:
    Analyzed video content using video_analysis.ipynb tools to extract metrics like attention hotspots and engagement duration. Videos were loaded from Google Cloud Storage (GCS), and GCS URIs were converted to accessible URLs for visualization.

  3. Using Gemini for Content Analysis:
    Configured a generative model (Gemini 1.5 Pro) to analyze video content, focusing on facial expressions, gestures, and visual elements. Safety settings ensured adherence to ethical guidelines.

  4. Crafting the Video Analysis Prompt:
    Created prompts to guide the analysis, asking the model to assess viewer engagement by identifying key visual and audio elements and categorizing their impact as positive, neutral, or negative.

  5. Generating Analysis:
    Used the Gemini model to provide detailed insights into customer engagement by identifying key moments in the video where engagement changed and categorizing them.

  6. Explainable AI and Ethical Considerations:
    Utilized explainable AI for transparency in recommendations and adhered to an AI Ethics Framework to protect user privacy and handle data responsibly.

Enhanced Feature: Multiple Video Analysis for Comparing Ads

  1. Multiple Video Analysis:
    Added functionality for analyzing multiple marketing videos simultaneously to determine which ads were more effective in capturing audience engagement.

  2. Parallel Analysis:
    Used an asynchronous approach to analyze multiple videos concurrently, saving time and resources.

  3. Aggregating and Displaying Results:
    Consolidated results from all videos into a single summary for comparative analysis, ranking ads based on engagement metrics. Used a DataFrame to organize and display the data.

Challenges and Learnings:
Handling quota limits when interacting with the Gemini model was a significant challenge, requiring careful scheduling of requests. I also learned about balancing AI-driven analytics with ethical considerations, ensuring privacy and avoiding manipulative practices.

Summary:
The solution analyzes individual ads and compares multiple ads simultaneously, offering a ranking of the best-performing campaigns. This provides insights into which creative elements work best, helping marketers make informed decisions for future campaigns.

Built With

Share this project:

Updates