How to (logically) test Facebook Ads at an incredible speed

Key Concepts

• Programmatic Creative Testing: A structured, data-driven approach to ad creation using predefined variables and AI to generate and test hundreds of ad variations at scale.
• Unique Selling Proposition (USP): The fundamental promise that differentiates a product in the marketplace.
• Marketing Angle: A specific, laser-focused slice of a USP designed to trigger a particular emotional response or address a specific customer problem.
• Level of Awareness: A framework (by Eugene Schwartz) categorizing customers based on their knowledge of their problem and potential solutions (Unaware, Problem Aware, Solution Aware, Product Aware, Most Aware).
• Market Sophistication: The degree to which a market has been exposed to similar products, dictating how unique or aggressive the messaging needs to be.
• Voice Mining: The process of gathering customer feedback from forums, Reddit, and reviews to feed authentic language into AI models.
• System Thinking: Moving away from "creative guesswork" toward a logical, repeatable framework for ad production.

---

1. The Mental Framework for Programmatic Testing

The speaker emphasizes that ads should not be treated as isolated creative pieces but as combinations of variables. The process follows a logical hierarchy:

1. Context Gathering: Feeding the AI comprehensive data, including brand descriptions, product details, and "Voice of the Customer" (VOC) data.
2. USP Definition: Establishing the core value proposition.
3. Angle Generation: Creating multiple ways to communicate the USP.
4. Messaging Frameworks: Applying proven copywriting structures (e.g., "Desire Future State") to the chosen angle.
5. Audience Segmentation: Tailoring the message based on the customer's "Level of Awareness."

2. The Technical System Architecture

The system is built on a Google Sheets interface connected to a Make.com backend.

• Automation: When a user clicks a button in the spreadsheet, it triggers a web hook in Make.com. This backend uses LLMs (like ChatGPT or Claude) to process the inputs and generate outputs (hooks, angles, or scripts).
• Templated Integration: The system uses the Templated API to automatically overlay generated hooks onto static image assets, handling font sizing and layout automatically.
• Ad Deployment: The system pushes these generated assets directly into Facebook Ad Manager, allowing the algorithm to determine the winners.

3. Step-by-Step Methodology

1. Input Data: Populate the spreadsheet with brand/product info and "gold nuggets" (customer quotes/pain points).
2. Generate Angles: Use the AI to create variations of a USP.
3. Generate Hooks: Select a USP, an angle, and a messaging framework. The AI generates multiple hooks based on these parameters.
4. Human Review: A human marketer reviews the AI-generated hooks to select the most resonant ones.
5. Creative Assembly: Use the Templated tool to generate static ads with the selected hooks.
6. Testing & Iteration: Push ads to Facebook. Once winners are identified, use the system to iterate on those specific angles or expand into long-form video scripts (VSSLs).

4. Scaling to Video (VSSLs)

For high-performing hooks, the speaker suggests transitioning to long-form video ads:

• Scripting: Use the AI to generate a full script (Hook, Lead, Body, CTA) based on the winning hook.
• Production: Use InVideo (AI) to import existing B-roll footage. The tool segments the script and automatically matches the best B-roll clips to each sentence, significantly reducing production time.

5. Key Arguments and Perspectives

• From Chaos to Logic: The speaker argues that most agencies fail because they rely on "creative" intuition. By treating ads as a system of variables, marketers can remove the guesswork.
• The Role of the Human: AI is not meant to replace the marketer but to act as a force multiplier. The human remains the "editor" who ensures the output aligns with the brand voice and customer sentiment.
• Iterative Scaling: The goal is not just to launch ads, but to find winning combinations that can be iterated upon indefinitely.

6. Notable Quotes

• "An ad is not only a piece of written copy or a nice picture. It's a segment of different predefined parameters."
• "It's not AI testing thousands and thousands of ads without control; it's like it generates something and then you can select... that's the human touch."
• "If you start thinking of ads in something that is not chaos, it's logical."

7. Synthesis and Conclusion

The programmatic creative testing system transforms ad buying from a subjective, chaotic process into a scalable, engineering-like workflow. By leveraging LLMs for copy generation, automation tools for data flow, and structured frameworks for market research, businesses can test hundreds of variations daily. The ultimate takeaway is that winning ads are not "created" by accident; they are discovered through a systematic process of testing variables, analyzing performance, and iterating on what resonates with the customer.