“the further you are away from the metrics you want you should proportionately increase the radicalness of the testing and experimentation you try...if you have total trash like you're not losing anything just try something nuts”

When Metrics Are Far Off Benchmark, Increase The Radicalism Of Your Experiments

Jacob's calibration rule for founders: the scale of your experimentation should be proportional to how far you are from your targets. If conversion is in the ballpark, run conservative A/B tests. If it's terrible, incremental tweaks are statistically equivalent to nothing — try a completely different paywall, a different business model, or a different app entirely. The time cost of trying something radical when things are broken is far lower than the time cost of grinding through marginal improvements.