Very interesting, love that you share metrics from your ad campaigns!
Deborah Kay

It can be tricky to identify broader targeting that works, but also well worth the testing. Once you’ve found a few combinations that work you’re setting yourself up for success.

I generally aim to test at least 8–12 audiences across 90 days.

Deborah Kay, you mentioned testing the beauty/cosmetics interest…how many others have you tested?

A few that come to mind:

  • Competitor brand interest
  • Purchase behavior — beauty shoppers
  • Salon interest
  • Publication interest (Refinery29, InStyle, Allure, etc.)
  • Compliment interest — certain fashion brands or trends? What does your target audience wear?
  • Compliment interest — natural remedies, organic food (if the USP is something along those lines, obviously depends on USP)
  • Demographic — mothers?
  • Top GA Affinity or In-Market segments?

Breakdown your target audience and try to hit each segment based on who they are, what they like, where else they shop, etc.

Another key variable on performance is the conversion event you’re using to optimize your campaigns. If it’s set to ‘Purchase’, try changing it to one funnel step above (i.e. AddToCart). This is especially worth testing if your conversion event (per ad set) sees less than 100 event fires in 30 days.

Re: How long do I run campaigns for?

If I’m looking for traction (i.e. new audience) that I run it as long as needed for some semblance of statistical significance (see ‘cascading significance’). Aim for 95% significance, but if you need to move fast 90% should be fine.

*NOTE: This is more a matter of spend than time. The higher your daily spend, the quicker you can conclude the test.

If I’m scaling an audience that’s performing…well, I have some audiences that have been purring for months.

Re: Will my serial ads hit the same subset of the larger audience?

Yes and no. For each ad, Facebook will try to find the best people based on the targeting options, each separate ad’s CTR, and each separate ad’s engagement (likes, comments, shares).

So each ad could be showed to the same person, or it could be shown to someone different (based on the CTR/engagement of the respective ad in the auction). Additionally, there are other advertisers bidding on the person Facebook decides to show your ad to — so even after the match you might not win the auction for that specific person.

At any given moment, each ad is rolling a die for it’s own version of the audience.

So yes, there will be overlap. And also no, they aren’t guaranteed to keep hitting the same people — every moment is a roll of the die.

*NOTE: I don’t have any insider insight on this and the above is based purely on my current understanding of the Facebook algorithm (see “how Facebook ads really work”).

Re: When should I kill my ad to test a new idea?

First, prioritize your ideas based on impact and learnings. Imagery and overall message should be a higher priority then minor copy tweaks.

Then, run the new version against your current ad. To ensure as clean a test as possible, don’t turn off the original ad. For a super clean testing, at we generally set up one ad per ad set. This allows us to duplicate the ad set and switch out the ad, ensuring that the targeting and daily spend is the same.

After you’ve spent enough to make a call (see ‘cascading significance’ above), turn off the underperformer. Then test a new ad.

Show your support

Clapping shows how much you appreciated Edwin Plotts’s story.