Turn off the ad if your observed cost per trial is above the curve. Below the curve, your results could still be noise.
$
Kill threshold (75% confidence)
Target CPT
How to read this: If you've observed 3 trials and your CPT is $51 or higher, there's a 75% chance the ad is genuinely worse than your $30target - kill it. As you collect more trials, the threshold tightens toward your target because the data gets more reliable.