Chapter 6: The Scaling Framework
The thesis was proven. This chapter is about how I scaled it for $500.
El Pico Navideño proved the thesis. The next step was scaling it systematically without burning money on audiences that wouldn’t stick.
The first decision was not to set an AVD target upfront. I let the market tell me what good looked like. El Pico Navideño established the baseline for long-form content. A separate January test — a Funk track with a car drifting through Mexico City — established the baseline for short-form. Once I had both numbers I knew what I was optimizing toward.
The bandwidth problem nobody talks about
Scaling across LatAm has a structural wrinkle that most campaigns ignore. In many markets telecoms offer free access to Meta apps but cap data for YouTube. That means a broad targeting approach wastes money on users who can’t actually watch long-form content on mobile data.
The fix was simple — use AVD as the filter. Users who watch with high AVD have either premium data or Wi-Fi. That single metric naturally filtered out low-quality traffic without needing complex technical exclusions. It’s how 18k subscribers became genuinely engaged viewers rather than cheap volume.
The AVD kill switch
I ran broad regional targeting with genre-based affinity layers to push the algorithm toward the right audience. But the real efficiency mechanism was ruthless budget control.
If a region’s AVD dropped below baseline I didn’t optimize. I cut the bid by 80% immediately.
That one rule turned the campaign into a self-correcting system. Only high-performing regions got budget. The algorithm learned fast because I only fed it clean signal. The result was a 2:1 organic-to-paid watch time ratio — for every paid hour the channel earned two organic hours for free.
The numbers:
Total subscribers acquired: 24,000 Net retained: 18,000+ Total spend: ~$500 Revenue goal: zero
The goal was never monetization. It was to validate an AI content engine against real audiences at the lowest possible cost. $500 to prove a system works across 20+ markets is not a media budget. It’s an R&D investment.
The honest risks:
The algorithm is a landlord not a partner. Google can shift ad serving, update ranking logic, or deprioritize the content category at any time. That’s a real dependency.
High AVD doesn’t always convert to subscribers. When growth stalled despite strong watch time the fix was rotating the visual creative — new image or video chassis — to force a fresh engagement response. That worked consistently but it’s a manual intervention in what should be an automated system.
The bigger unknown is whether this scales beyond music. The mechanism is proven. The portability is still being tested.