Net-new concepts vs iterating winners: iteration has a half-life

Net-new concepts vs iterating winners: iteration has a half-life
Across 38 accounts, 90d, on a proven winner:
— Iterations (hooks/edits of the winner): hit-rate 35%, but lift shrinks each round — by iteration 4, +5% at best
— Net-new concepts: hit-rate 15%, but winners outperform iterations by 1.6x ROAS
Iteration is cheap insurance; net-new is the only path to a step-change. Pure iteration plateaus your whole account.
Read: budget ~70% to iterating winners for stability, 30% to net-new swings. All-iteration accounts slowly decay to the mean.