The creative test grid: stop guessing which variable matters
I wasted months changing five things at once and learning nothing. Then I built a variable grid.
The setup
Nutra offer, native traffic, flat results from random creatives.
The move
Isolate ONE variable per round. Round 1 = hook only. Pick one image, one body, write 4 different headlines. Everything else identical.
— Round 1: headlines (4 variants)
— Round 2: take winning headline, test 4 images
— Round 3: winning headline + image, test 4 angles in the body
— Round 4: winning combo, test 3 CTAs
Each round runs to ~2x payout spend per variant, then you lock the winner and move down.
The numbers
Headline round alone swung CTR from 0.6% to 1.4%. By round 3 the compounded combo hit $5.20 CPA vs the $9 I started with. Total testing spend: ~$310 across four rounds.
The lesson
The headline usually carries 60-70% of the lift. Test it first, in isolation, before you touch anything else.
What I'd do differently
Fewer variants per round (3, not 4) to spend less reaching a verdict. Numbers illustrative.
Arb Files
@ArbFiles