How much should founders budget for ad testing, and when should they kill or scale a creative?

A workable default is 10 to 20 percent of spend reserved for testing, giving each new concept about 3 to 5 times your target CPA before a keep-or-kill decision, with a 60-30-10 split between proven winners, variations, and new swings. AI changes the volume side (you can now afford 15 to 30 variants a week) but not the discipline: most founders kill ads too early on noise or scale them before the signal is clean. Several sharp operators now argue separate testing campaigns are obsolete and you should launch straight into scale campaigns.

A quick orientation. The real value is below: resources worth your time, from people who've actually done it.