"Test 50 creatives to find a winner" is how I burned $2,800 on noise

"Test 50 creatives to find a winner" is how I burned $2,800 on noise

The advice sounds disciplined: flood the account, let the algorithm pick. I did it. It taught the algorithm nothing and taught me less.

The setup
E-commerce offer, paid social, $60/day budget. I launched 50 distinct creatives in one campaign because a popular thread said volume wins.

The move
After a week of flat results I stopped, did the math on impressions per creative, and felt stupid.

The numbers
$60/day across 50 creatives = roughly 30-40 impressions each per day before the algorithm consolidated spend onto 3. Statistically, 47 of those creatives never got enough data to mean anything. I spent $2,800 over the test and 80% of it funded ads that never had a chance to prove themselves.

The lesson
Creative volume without budget-per-creative is just expensive randomness. The algorithm doesn't "explore" 50 things on a small budget — it picks fast and starves the rest.

What I'd do differently
Fewer, funded. I now run 4-6 creatives with at least $15/day each so every one clears 1,000+ impressions before I judge it. Same total budget, ten times the signal. Numbers illustrative — the starvation math is not.