Unlock the Modular Meta Ads Method

Ever since Meta leaned hard into AI-driven targeting, we’ve noticed that the more ad variations we test, the better the results. As long as the ad checks some basic boxes (clear audio, person talking), quantity beats quality.

The trick is to throw a bunch of ideas out there, see what sticks, and then double down on the winners with higher production value.

The problem is most people don’t have time, gear, or anyone confident on camera. So we built a solution.

Instead of filming full ads, we break everything down into short, standalone clips. 5-second hooks, 5-second benefits, and 5-second calls to action. But to make that work, each piece has to be filmed very intentionally so they flow together naturally. Same framing, tone, pacing, eye-line, lighting. Otherwise the final ads end up feeling super awkward and choppy.

If done right though, it lets us mix and match clips to make tons of unique ads, all from one shoot.

This week we had a 2-hour shoot with a local service business, and things got out of hand. We ended up with 14 hooks, 8 benefits, and 13 CTAs. Our goal was to make 50 ads, but when our Python script mixed and match all the clips together…suddenly we had 1,456 videos.

It’s a goofy amount of ads. You wouldn’t want to run all of those at once (unless your budget was massive). But now this client has 6+ months of content to test, and we can quickly learn which messages actually work before going all-in on higher quality production.

Here’s our (simplified) process:

1. Write solid scripts ahead of time—think mix-and-match lines, not full monologues.

2. Film the clips (this is all we need from clients). Use a teleprompter if possible—it keeps delivery consistent.

3. Edit and organize clips with a solid naming system. Something like H_CleanGrill_v1 or C_FreeConsult_v2_AI helps keep things sane.

4. Caption the clips—makes a huge difference in performance.

5. Recombine using code (we built a custom tool for this).

6. Upload to Meta via bulk import (we also use a script to auto-generate CSVs for this).

7. Track results by maintaining a simple clip database. We use Notion to tag, log, and report on which components are driving performance.