I’m running a paid ad strategy for a small clothing brand. It is a prospect campaign and the ROAS is on 3 to 1.5 depends on the ad. We have 7 ads in each ad set and the campaign has been running for 30 days now. My questions is, which metrics do you optimize for and do you guys have tips and tricks to know when to kill and ad. Thanks. 🙂

Hi Philip! On the Ad level, it’s common to kill Ads based on the most costly CPC or lowest engagement. On the Ad Set level, you can use CPA to determine which audience group doesn’t work. Regarding optimisation, IMO it comes down to your campaign’s goals and objectives. Is it conversion, action or reach focused? Hope this helps 🙂

I'm running CBO campaigns to make a simple setup, today I have set a cost cap on the campaigns and that is a little under the average CPA (purchase). Should I always go with the ads that have the highest ROAS and kill the low ROAS or how do you optimize?

Hey Philip, if you’re letting Facebook’s algorithm distribute your Ads based on your campaign goals, then the short answer is YES, kill Ads with the lowest, statistically significant ROAS or the highest CPA after a 7-14 days period or when the Ad Sets learning phase is completed, whichever that comes first.

Hey Philip, not sure if you’re running these for your business or you’re hired to run them for someone else. But in general I optimize for ROAS and kill the worst performing ads and then in their place start new ads to see if they can beat my current best performers while slowly adding more budget to my top performs (I personally add no more than 10% spend per day while scaling, others go up to 20% per day but I like to be slower in my scaling). Where this might not be the right strategy is if you/client are more concerned with total dollar value of sales more than ROAS, in that case maybe you’re okay with a lower ROAS as long as it’s higher total amount of sales. In that case then you’re looking you’re getting more complicated b/c you need to scale multiple ads at the same time to see which can generate highest total sales number balanced against how much you’re willing to sacrifice ROAS.

Okay, that makes sense, I’m going for the highest ROAS so that makes sense. Would you split up the audiences into male vs female or make a new campaign for each? Also do the learning phase stops when you scale or add new ad sets and ads to the campaign? Thank you so much for you tips. 🙂

It depends on what you're testing for. If you're testing for gender conversion rate, then create two Ad Sets within the same campaign, keeping all other variables constant.

For clothing, yes, split your ads by gender but ALSO tailor your ad copy and images/video to the gender. And if you are using CBO I would make them seperate Campaigns so that you can optimize each on it’s own. It may be that your brand performs better with one gender vs the other and that is good info to get, but if they are under the same campaign on CBO then facebook will just downplay the low performing gender which probably isn’t your goal. Learn phase starts again for each new ad/adset you add but not when you just change budgets.

Hey Scott, could you elaborate why the recommendation to make them separate campaigns if one is trying to determine which gender (on an Ad set level) converts better on a CPA (acquisition) basis? If you're trusting Facebook's algorithm, wouldn't you want to let the CBO tell you which type of audience (Ad set level) converts better, and thus, automatically allocate higher weightage to the higher converting Ad set (aka the higher converting gender for this case)?

@Randy Chai if I was just selling a product that would be applicable to my entire user base (let’s say I’m selling a kitchen knife and I don’t care the gender of the person who buys it, I just want to know what is and is not selling knives) then you’re right, I want the FB algorithm to do it’s thing. However clothing businesses are a little bit of a different beast, b/c even if one gender purchases more clothes from the brand, if it is a brand that offers both genders of clothes they’re generally not going to make the business decision to stop selling Men’s cloths entirely just b/c the women’s clothes perform better. A brand selling both genders will generally try to optimize their designs/styles etc for each gender. So in effect, if you let the FB algorithm just do it’s thing, if women’s clothes sell better then FB is just gonna optimize for that and it’s gonna shut down the Men’s adsets and then the brand is left with a whole bunch of unsold Men’s clothing simply b/c FB was getting better ROAS on women! Most likely they still want to sell Men’s clothing, even if the ROAS isn’t as high, so you’d want to split the two into separate campaigns so each gender get’s it’s fair chance at selling as well as possible, even if one is better than the other. The FB algorithm is the best algorithm that has ever been designed for selling stuff, it’s truly mind boggling, that being said, since it is just a cold-blooded selling machine you often need to nudge it in the right direction based on the larger business needs and do a little bit of “gaming the system” to get the algorithm doing whats actually best for you.

Thanks, Scott! That thought process makes a lot of sense. As I’d assume that females buy more volume of clothes than males (potentially spending more too) which will then lead the CBO with a ROAS goal toward a “false truth”

Exactly.