Here's the rule that runs this whole operation: the title and the thumbnail come first. Before a single frame is shot, we decide what the cover promises. Then the entire production exists to make that promise true.
It sounds backwards. It's not. A video nobody clicks doesn't exist. So the click is the product, and the click is decided by packaging — the title and the image, read in about one second, at 120 pixels wide, next to twenty other thumbnails screaming for the same tap.
20 to 50 covers per video
We don't make one thumbnail. We make dozens — 20 to 50 — and then we test. Not "which do I like" in a Slack channel. Actual votes from actual humans. Two phases:
- Macro round. Three genuinely different directions — different angle, different face, different object. Not three shades of the same idea.
- Micro round. Take the winner and iterate the small stuff: shirt color, background object, how much of the face. Stop when the deltas flatten.
That's exactly what the A/B Lab automates. Two covers, one vote, live results, and a ledger that remembers every finding so next time you're not guessing.
One face, one emotion, one absurd thing
The aesthetic isn't an accident. Saturated, high-contrast, face-forward. One readable emotion — lately "determined" beats "shocked," which we only learned by running the test 30 times. The subject cut out with a hard edge against a simplified, hyper-colored scene. One scale cue: a wall of cash, a giant door, an island of gold. Little or no text, because text doesn't survive at 120px.
The tone underneath all of it: spectacle plus generosity. The money is always being given away, never flaunted. That's the emotional payload, and it's the reason people come back.