UGC enters the same structure as everything else: launched into the testing share of budget, roughly 30% at steady state, read per angle and concept, with winners graduating into the scaling pool that holds about 55% of spend.
Because each video ships with hook variations, the readout is granular: sometimes the video wins, sometimes only its third hook does, and that hook then gets grafted onto the next batch. Nothing is a one-off; everything is a data point in the angle grid.
