Headgum: Automating Ad-Read Evaluations
Revolutionizing the podcast industry through automated ad-read analysis and insights.
Automating Ad-Read Evaluations
Project Location:
Los Angeles, California
Industry:
Podcast Network
Use Case:
Automating the evaluation of ad-read transcripts using AI
Website:
Headgum.com
Headgum sought to leverage the power of Google Cloud's Large Language Models to automate the process of evaluating ad-read transcripts against advertisers' objective and subjective requirements from a script.
About Headgum
Headgum is an LA-based podcast network that produces a diverse array of popular shows across various genres, including comedy, true crime, music, and storytelling. With a roster of hit podcasts such as Doughboys, Dead Eyes, Punch Up The Jam, If I Were You, and Not Another D&D Podcast, and many others, Headgum has established itself as a leading creator of engaging and entertaining audio content. The network is dedicated to delivering high-quality, original programming that resonates with audiences and attracts a loyal following.
Project Challenges
Headgum wanted to perform a Proof of Concept (PoC) to understand the value of Generative AI in automating the ad-read evaluation process while ensuring the effort would not be considered a science project.
Scope of Work
Zencore assisted Headgum in targeting the right use case, deploying a Generative AI Sandbox, providing enablement sessions, implementing Ad-Read Evaluation models, and using Terraform for project provisioning.
Key Achievements
- Successfully demonstrated the value of Gen-AI to Headgum's executive team
- Developed a new Headgum offering that significantly benefits advertisers
- Implemented language models (PaLM and Model Garden LLM) that improve language understanding and generation capabilities
Solution
Zencore worked closely with Headgum to deploy a Generative AI Sandbox, provide enablement sessions, implement Ad Read Evaluation models, and use Terraform for project provisioning. The resulting Streamlit application is fully functional and user-friendly, enabling Headgum to effectively interact with ad reads and obtain meaningful responses from the language models. The language models (PaLM and Model Garden LLM) show improvement in language understanding and generation capabilities, meeting Headgum's requirements for ad read evaluation.
Business Goal
Headgum's primary goal was to automate the process of evaluating ad-read transcripts against advertisers' objective and subjective requirements from a script, leveraging Google Cloud's Large Language Models to improve efficiency and provide valuable insights to advertisers.
Business Value
By automating the ad-read evaluation process, Headgum can eliminate a significant amount of manual effort required to perform the evaluations and provide advertisers with valuable data on the quality of their ads.
“The project was a huge success. Zencore assisted in targeting the right use case in which our executive team could understand the value of Gen-AI, while also making sure the effort wouldn't be considered a science project," said Andrew Pile, Headgum Co-Founder and CTO. "This new Headgum offering will significantly move the needle for our advertisers and Zencore was instrumental in this project's success.”