One of the challenges w/ campaigns like this is that waiting for statistical significance can quickly become very expensive.

If we’re running 4 creatives in an ad set w/ CPA of ~$100, it could take a few thousand dollars before you can make a call on whether the creatives are performing or not. I’ll share how we’re approaching this - would love to get any inputs from you guys: -> If an ad or ad set or campaign gets to$200 in spend(2x the CPA) with no purchases OR 3 days in a row with no purchases, pause it.

In the above:

a. if an ad performed well in the past, then duplicate it and try again.

b. if an ad set/campaign performed well in the past, then add a new creative into the ad set/campaign to reset the learning phase.

**

**

I’m curious: what else should we be thinking of or doing?

It sounds like you're doing a lot right. The biggest thing I would suggest is to adjust your account structure to retain learnings within campaigns.

Your example of taking a high performing Ad Set and adding an Ad to reset the learning period caught my eye. In a case like that, you could instead keep the high performing Ad Set running, duplicate it and add whatever new creative you want to the new one, and use Campaign Budget Optimization to maintain your hard-won performance while allowing it to evaluate your new creative.

But the principle that's based on is keep Campaigns live whenever you can, use CBO to your advantage, and add instead of change whenever possible so you retain your learnings for as much of your spend as you can.
by (410 points)

+1 vote