A workable starting framework is three to five distinct angles per campaign, each with two to three variants such as different hooks or creators, rather than dozens of near-identical edits of the same idea. Testing angles, not just edits, is what actually produces signal.
Cut losers fast based on hook rate and thumb-stop performance in the first few days, then double down on what's working with fresh variants of the winning angle rather than the winning exact ad, since GCC audiences fatigue on repeated creative quickly.