Week 17: Precision Targeting, HTML Systems, and Lessons from Our First Live Campaign

Precision Targeting, HTML Systems, and Lessons from Our First Live Campaign

This week, Entonomy crossed an important milestone: moving from building infrastructure to testing autonomous execution in the real world. We’ve now entered the phase where ideas meet live data, and the results are already reshaping how we build.

🔁 From Business Plan to Deployment: One Workflow to Rule Them All

We completed the core Venture Main Workflow in n8n. This system:

• Accepts structured business plans via form

• Generates configuration parameters

• Builds HTML assets (emails and landing pages)

• Pushes assets to GitHub Pages

• Stores a full venture object in Supabase

In short: with one input, we now produce fully configured, deployment-ready ventures — no manual touchpoints required.

🧠 Targeting, Reimagined as a Multi-Agent System

Our most complex upgrade was deploying a three-agent targeting system:

1. Demographics Agent — defines age and gender

2. Education & Work Agent — enriches targeting using Meta's API to pull degrees, majors, employers, and positions

3. Interest-Based Agent — constructs nuanced interest/behavior/income/lifestyle targeting using Meta's interest libraries and validation routines

Each agent operates with clarity and precision, dramatically improving targeting reliability and modularity. All outputs are stored in a standardized flexible_spec JSON format.

📏 Audience Validation: Don’t Launch What Won’t Scale

We integrated Meta’s audience estimate API to eliminate audience segments that are too small to be meaningful. This was a major unlock — we found that some combinations that “looked good” were actually too narrow to justify ad spend.

Now, weak audiences are automatically pruned before launch.

📧 Refactored HTML Creative System

Our original HTML generation approach was fragile. So we rebuilt it. Now, all HTML assets use a structured skeleton with 40+ dynamic keys that agents populate contextually. This makes templates reusable, maintainable, and consistent across ventures — all while preserving design fidelity.

🎯 First Live Campaign: TopHop

We ran a live campaign for TopHop. It didn’t perform well — and that’s exactly what we needed.
The original targeting logic was too broad and lacked coherent filtering. That campaign directly informed the agent-based targeting overhaul and our new audience size validation strategy.

🛑 Roadblock: GitHub Account Suspension

Our GitHub account was unexpectedly suspended, taking down the hosted assets required to run new campaigns. Until this is resolved, campaign deployment is paused. We’re actively working on migrating to AWS (S3 + CloudFront) for hosting to avoid future single points of failure.

Key Learnings

• Agent decomposition leads to modularity and higher targeting accuracy.

• Real-world audience estimates expose flaws in theoretical configurations.

• Skeleton-based HTML vastly simplifies creative system management.

• Cold traffic punishes poor targeting — precision matters more than volume.

• Redundant hosting is essential — GitHub can’t be a single point of failure.

• Human-like reasoning is still necessary for validating complex targeting strategies.

What’s Next

Next week, we aim to:

• Launch live campaigns for 20+ ventures in parallel

• Stress test our workflows at scale

• Migrate asset hosting to AWS

• Monitor conversion pipelines (from Meta to Stripe)

• Refine early failure detection in targeting validation

If successful, Entonomy will reach a major milestone: autonomous campaign execution across a portfolio of ventures, fully independent of human intervention or GitHub.

Stay tuned.