Things to Turn Off

I read Will Atherton's guide on running profitable TikTok ads for apps. He scaled one of his apps from a hundred dollars to ten thousand MRR in a month, on five-dollar CPMs, while everyone else was on Meta paying eight times that. The piece is a long step-by-step from pixel setup to scaling, and most of it I want to keep.

The hook formula will get most of the attention. The part that quietly does the work is the list of things he tells you to turn off.

Disable iOS goal-based budget increase. Disable Pangle and Lemon8. Disable auto-select creative assets. Disable every automatic enhancement. Use campaign budget, not ad-group budget. Pick one CTA. Pick one country. Pick one language.

Read it together and the whole playbook is a refusal. TikTok offers you a dozen helpful toggles, and the right answer to almost all of them is no. Every toggle is the algorithm trying to spend your money on the easiest possible win, which is rarely the win you wanted. The defaults are tuned for TikTok's revenue. Once you accept that, the long checklist stops feeling like superstition and starts looking like a contract you're refusing to sign.

I like the rest of it too. Pixel and MMP before creative, because no creative survives broken attribution. One Smart+ campaign instead of nine forked ad groups, because the algorithm needs concentration to learn. After three days at fifty dollars a day you have a hundred and fifty dollars of data and either a winner or you don't. If the campaign was set up right and the ads still flop, the ads are the problem. That kind of clean ownership is rare in marketing writing.

His scaling rule is the same shape. Increase budget thirty percent every three days. Faster than that and the algorithm loses its grip. CPAs balloon, you panic, you cut the budget, the algorithm gets more confused. Slow scaling is stable scaling.

This rhymes with what Koby Conrad said about hook and placement. Two things matter, the rest is support. Will's version is the operations side of the same coin. Two things matter: the pixel and the creative. The rest is a long list of toggles you should leave alone.

The thing I keep coming back to is the discipline of the no. Most people running ads burn money in the first week because they let the platform run them. They take the suggested placements, the suggested budget pacing, the suggested creative variants, the suggested everything. The algorithm learns nothing because the input was a thousand small accommodations. Will's playbook is one campaign, one country, one CTA, one bid strategy, six to twelve ads, fifty dollars a day, and a list of refusals as long as your arm.

Set up the boring infrastructure. Make the bet small and clean. Turn off everything that wasn't the bet.