"Test creative until you have statistically significant results, then scale the winners based on that."

That seems like good advice on first glance. How can you disagree with methodical testing that's attentive to good statistical principles?

Because it's almost impossible to actually do without burning huge amounts of money on Meta, for one, and for two, it's unnecessary anyway, since Meta's machine learning can use better methods of forecasting your creative's performance than frequentist-style stat sig tests.

Yes, this one is getting into the weeds. Hopefully it helps clarify more of why you absolutely positively should not run auto-bid creative testing campaigns on Meta.

EPISODE HIGHLIGHTS

EPISODE SPONSOR

Virtual assistants can be helpful. Virtual professionals can transform your business. Get connected to incredible ecommerce talent from the Philippines with More Staffing by visiting ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠www.morenow.co⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠.

REFERENCED IN THIS EPISODE

FOLLOW UP WITH ANDREW

EPISODE MUSIC