AWS Only
Boosting Ad Relevance for LiveIntent with AI-Powered Solutions
LiveIntent faced the challenge of delivering contextually relevant ad copies in newsletters, as their existing system inserted random ads, reducing clickthrough rates. To address this, Firemind developed a solution using Large Language Models (LLMs) to generate ad copies tailored to the newsletter's content, ensuring relevance and improving user engagement. This AI-powered approach significantly enhanced ad performance and user satisfaction.
AWS ONLY ·
Boosting Ad Relevance for LiveIntent with AI-Powered Solutions
LiveIntent faced the challenge of delivering contextually relevant ad copies in newsletters, as their existing system inserted random ads, reducing clickthrough rates. To address this, Firemind developed a solution using Large Language Models (LLMs) to generate ad copies tailored to the newsletter's content, ensuring relevance and improving user engagement. This AI-powered approach significantly enhanced ad performance and user satisfaction.
At a glance
LiveIntent provides a platform for news publishers and advertisers to deliver targeted ads and marketing materials via email. By leveraging email-based data, they enable advertisers to reach consumers across multiple devices without relying on cookies.
Challenge
LiveIntent's main challenge was delivering contextually relevant ad copies in newsletters, as their existing system inserted random ads that didn't align with the content, leading to lower clickthrough rates and reduced user engagement.
Solution
Firemind developed an AI-powered solution using Large Language Models (LLMs) to generate contextually relevant ad copies by analysing newsletter content and advertiser information. This approach ensured ads were aligned with the newsletter's tone and content, improving relevance and engagement.
Services Used
Amazon Bedrock
AWS Lambda
Amazon S3
Amazon API Gateway
Outcomes
Increased clickthrough rates with contextually relevant ad copies significantly improving user engagement and clickthrough rates.
Optimised advertisement copy quality matching the tone of the newsletters and focusing on advertiser content.
Business challenges
Improving advertisement relevance in newsletters with contextual accuracy
LiveIntent faced a significant challenge in delivering ad copies that were contextually relevant to the content of newsletters. Their existing system inserted generic ads into newsletter slots, which often resulted in ads that did not align with the newsletter’s subject matter or tone. This lack of relevance diminished user engagement, leading to lower clickthrough rates and reduced overall effectiveness of their advertising efforts.
Additionally, the inability to tailor ads to the content of the newsletters created a disconnect between the advertiser’s message and the readers’ interests. This mismatch not only affected user satisfaction but also hindered the advertisers’ return on investment. To address these issues, LiveIntent needed a solution that could generate ad copies aligned with both the content of the newsletters and the promotional goals of the advertisers.
“It was about finding an AWS Partner that understood our ethos and values. It’s been really refreshing talking to Firemind about the the project with clear communication and without the usual jargon and abbreviations.”
Pete Kilbane, Commercial Director — MRC
Solution
AI-driven contextual advertisement copy generation for enhanced relevance
To address LiveIntent’s challenge of delivering contextually relevant ad copies, Firemind developed a solution utilising advanced AI and AWS services. The solution involved integrating Large Language Models (LLMs) through Amazon Bedrock, specifically testing models like Claude Instant and Claude 3 Haiku, to generate ad copies aligned with newsletter content. This approach ensured ads were contextually relevant and engaging for readers.
The process began with scraping and preprocessing data from newsletters and advertiser URLs using AWS Lambda functions. These functions extracted plain text from HTML content, which was then summarised and analysed by Amazon Bedrock. The LLMs used this summarised data to generate ad copies tailored to the newsletter’s tone and content, addressing the issue of generic, irrelevant ads.
A frontend interface was created using ReactJS, allowing users to input URLs, select LLM models, and view the generated ad copies. The results were stored in Amazon S3 buckets, with metadata managed in Amazon DynamoDB. This setup ensured easy access to ad copies and allowed for secure API interactions through Amazon API Gateway.
Overall, this AI-driven solution effectively resolved the problem of irrelevant ads by generating contextually accurate and engaging ad copies. The use of AWS services and LLMs led to improved clickthrough rates and increased advertiser satisfaction, demonstrating the value of tailored ad content in enhancing user experience.
Enhanced advertisement relevance
Firemind’s solution successfully integrated Large Language Models (LLMs) through Amazon Bedrock to generate ad copies that were contextually aligned with the newsletter content. By analysing both the newsletter and advertiser data, the AI-generated ads were tailored to fit seamlessly with the existing content, significantly improving relevance and engagement.
Improved clickthrough rates
The introduction of contextually relevant ad copies led to a notable increase in clickthrough rates. With ads designed to resonate with the newsletter’s tone and subject matter, users found the advertisements more appealing, which in turn boosted interaction and effectiveness for both publishers and advertisers.
Streamlined user experience
The development of a user-friendly frontend using ReactJS, combined with robust AWS infrastructure, provided an intuitive interface for generating and managing ad copies. Users could easily input URLs, select models, and view results, while the secure storage and processing through Amazon S3 and DynamoDB ensured a smooth and efficient workflow.
Model Spotlight
Claude 3 Haiku
We chose Claude 3 Haiku as it is the fastest (at the time of release), most compact model of the Claude 3 family and produces near-instant output. In addition to the 200K context window, the Claude 3 Haiku model can generate human-like outputs and was ideal for automating basic repetitive tasks.
We calculated that if LiveIntent wanted to generate 1,000 ads a day, using Haiku would cost around $3.50 (compared to Sonnet which would cost around $42 a day).
The marginal gains in accuracy from Sonnet would not be worth the significantly higher cost. Sonnet is a fantastic model for complex reasoning, but LiveIntent was the perfect use case for Haiku where the prompts were relatively simple.
For example: “Generate an ad relevant to both product X and newsletter Y”.
What’s more, Haiku performed around 50% faster than Sonnet.
Why Firemind
Firemind’s extensive knowledge of AWS services plays a crucial role in their appeal as the chosen partner for this project. With their adept use of services like AWS Lambda for efficient data processing, Amazon S3 for secure storage, and API Gateway for managing interactions, Firemind provides a robust and scalable solution. This expertise ensures that the implementation is seamless and integrates smoothly with LiveIntent’s existing systems, enhancing overall performance.
Firemind’s commitment to creating a user-centric experience also stands out. Their development of a user-friendly frontend, combined with their focus on delivering a streamlined workflow, helps ensure that LiveIntent’s team can easily navigate and utilise the new system. This attention to usability reflects Firemind’s dedication to not only meeting but exceeding client expectations.
Get in touch
Want to learn more?
Seen a specific case study or insight and want to learn more? Or thinking about your next project? Drop us a message below!