How to optimize - no Facebook learning phase

[OnyxX]

Like many of you, I have watched all of Andrew's FB ad videos multiple times like he is a prophet and they are scripture. And, of course, I've noticed that he has changed the way he does things over time. It's hard to tell what is the right approach at this point. He had an optimization video a while back in March 2020 where he said to look in the breakdown and divide adsets up to target the lowest cost demographics (even by gender and age). If you've ever tried doing that, it decreases the audience size and increases the cost. So it doesn't work. I saw that he admitted this in a later video.

So how does he optimize then? I can't see any other way to do it other than delete countries or turn off low performing ads and adsets and start new ones with different audiences.

Which brings me to my point. All other Facebook marketing strategists talk about the "Learning phase" and that every time you edit your adset, you reenter the learning phase. You supposedly need 25 to 50 conversions for your adset to be optimized and exit the learning phase. If you edit it, it starts learning all over again. My facebook does not show a learning phase (even with Delivery view on). Each ad goes straight to active. And I keep editing them!! Cuz I thought that was optimizing them!!!!!!! Cuz Andrew said to!!!

You are supposedly supposed to have a budget of $10 a day per adset to allow Facebook to optimize (I just learned). We've seen Andrew's Facebook back office with a kajillion adsets per campaign in the past with test adsets. He recently did a video with just one adset. I can't seem to get my conversions below $.40, so I'm wondering if it's because I have too many adsets and I keep messing with them and FB has to keep going back to the learning phase each time. I HOPE that is the problem.

My Facebook does not show a learning phase ever, so I did not know anything about this. Apparently, those of us without it have a glitch. I've never seen it.

How on earth is he optimizing his ads if you aren't supposed to edit them during the learning phase?

I did get one maintenance campaign down to $,20 by just leaving it alone. I feel like I missed something really big!

What is everyone else doing?

[Regret Regret]

I start my campaign with one ad set. That way all the budget is used on it and it optimizes quickly. I put 6 ads in it and use broad targeting. After I get 50+ conversions I turn off the dud ads and duplicate with other targets.

I duplicate about four ad sets and just let them run. It’s been my experience that fb algorithm will shift the budget away from bad ads so I just don’t touch them.

one exception I made was adjusting the countries targeted after seeming listener/conversion ratios. I know everyone says don’t target India, but I would add Egypt, Tunisia, and Morroco to the list as I was seeing conversions but no listeners. El Salvador and the Philippines were getting some listeners but not a great listener/conversion rate. This threw my ads back into the learning phase which costs more in the short term, but I was early in my campaign so it was worth it. Next time I’ll start with those countries off and hopefully wont have to adjust any ads.

[OnyxX]

Yeah, I read through the forum last night and saw your answer on another post. I just can't believe I missed this part of it, but I guess everyone would assume you would just see "Learning phase," but ours never had that. I have enough history with my ten (!) adsets that I chose the 4 best performing ones and will let them ride for now. I've noticed that once Facebook has a lot of history with one adset, it's reluctant to give traffic to a new adset, especially if the audience is smaller. I also started with 6 ads on each adset and different ads performed better with different audiences, so that might be one reason to start with different ads and adsets in the beginning. I tried to force some adsets to use the best performing ad and cost went up if I turned off the cheaper one for that adset, so my four adsets each have two or three different ads now.

But for countries, one way to solve the country issue is to put them all in one adset and put a budget limit on it, this way you still have the opportunity to target them on release day or the first week and you can turn them off afterwards without sending your other adsets into the learning phase. Most of the traffic will go to India, so you can even put India by itself and the rest in another one. I did track them for a while, and only about 25% of conversions for India turned into listeners. But you never know. We had a listener in Eqypt put our song on repeat and it was getting 200 streams a day just in Egypt.

Thank. you for your answer!

[georgik]

lemme show u da wei https://www.facebook.com/business/m/power-five

[georgik]

also heres my list of FAKE CONVERSIONS:

India

Morocco

Tunisia

Nicaragua

Egypt

Philippines

El Salvador

Italy

[OnyxX]

Dude, that's kind of weird that you would get on Andrew's forum and promote another service...ha ha. No, I'm just asking what approach you take with a new ad campaign. Andrew's video from March 2020 said to go in and "optimize" adsets by narrowing the parameters by turning off placements or even targeting specific age ranges and gender, and I did that and my ad costs went up. I'm just learning now that I was restarting the learning phase over and over by editing the adsets too much. I'm guessing that he has since learned that, too.

I found a post by Andrew on the forum last night where he said to wait until each adset has around $5 in the ad spend before turning one off, and ideally, $5 each in all of the adsets with optimization turned on. I'm going to try that. He doesn't think you need to wait until you have 50 conversions with a music marketing campaign to know if it will be a loser.

I'm already seeing my ad costs going down just leaving everything alone. Boy do I feel dumb, but I had no idea there was a "learning phase." I've just been following Andrew's videos and he mentioned it a few times but didn't really explain it.

I also read somewhere that you shouldn't change your budget by more than 20% each day, but I've seen his spreadsheets jump from one amount to another. I wonder if he stopped doing that, too. He said in a recent video that you can just choose worldwide and nothing else and let the algorithm find your audience. There was a band on Reddit who did that and had great results. Then you could just have one adset. I think my targets have been too narrow, that's been another problem. I just learned that turning on expansion is a good thing. I thought it made the audience less specific, but I guess it doesn't really matter if you can just use worldwide and get good results (if you wait long enough).

[georgik]

the link i posted is facebooks new way to advertise. large audiences, cbo + other things. read it. if you play by their rules they will reward you with cheap conversions. thats how i do it and im currently getting 20 cent conversions from the usa. i have 1 adset per creative, targeting the same audience, with CBO on. here is my audience and results. keep in mind its in CAD not USD https://imgur.com/a/BeFjwJX

the only thing u have to watch out for is botty countries. facebooks algorithm does the rest

[georgik]

fear of bots btw is why i only target rich countries. plus, spotify pays more for those streams

[OnyxX]

Oh, okay. But back to my original question, from release day on, do you just set up your ads and then leave them and let Facebook do it's job or do you turn off low performing ads? It looks like your audience is a large group of artists. So you have the same large group for each adset and a different ad (video/creative) for each one?

I have left my adsets alone now for two days and the CPR is already coming down. I have four Adsets running and FB is ignoring one, and only giving a small amount of traffic to the other two. There are three ads/creatives on the one that is getting all the traffic and only one of them is getting most of it. There won't be anything to turn off, from the looks of it.

I will go back and look at the link you sent again. It didn't have Facebook's branding at first glance. Thank you for your feedback.

[georgik]

after release i run a/b tests until i find the best converting creative. then i go into my dts campaign, duplicate one of the ad sets (which are identical barring the creative) , and replace the duplicated creative with the a/b winner. thats it. i let it run forever, letting facebook do the rest.

in my case, each adset/creative pairing in my cbo campaign is a different song. look up georgi krastev in the ad library for context

[OnyxX]

How long do you run the initial test? I got the first impression from Andrew's videos that he was turning off adsets and starting new ones with different audiences the first day, but now I'm not so sure since one day of data doesn't account for day-to-day fluctuations. It would make more sense to wait at least a few days, knowing what I know now. I don't have the "learning phase" mode showing to follow, so I'm just guessing based on the number of conversions. My ads last week seemed to settle on around $.50 per conversion after a day or so and then I was messing with them after that to "optimize." I think I've been unknowingly restarting the learning phase doing that. I'm at $.45 after two days, so this time I'm just going to leave it alone and see what it does. I had all the data from last week to choose the best performing ads and audiences manually. Most traffic is going to one adset and it only has 25 conversions so far. I'm keeping my budget low because we wasted so much money last week with a high CPR. I'll inch it up as the results get better.

You have the same audience with each adset? I copied an audience from another campaign initially that had gotten down to $.20 and kept it running while I had the same audience with the new song in a different campaign. They both went up to $.65 (in different campaigns running at the same time). I realize now that I was bidding against myself and driving both ad costs up, so targeting the same audience with different adsets usually isn't recommended. I turned them both off. I'm surprised that is working for you, unless I'm misunderstanding what you're doing.

[georgik]

i love Andrew but he does a/b testing completely wrong. you are supposed to use this https://www.facebook.com/test-and-learn/learn otherwise your results could be skewed by any one of a million things


"realize now that I was bidding against myself and driving both ad costs up, so targeting the same audience with different adsets usually isn't recommended. I turned them both off. I'm surprised that is working for you"

it is not possible to bid against yourself at your budget. if 2 of your ads end up in the same auction, facebook automatically takes the better-performing ad and ignores the other
i believe you can target the same audience in different ad sets in the same campaign if you are using the campaign structure i described previously

[OnyxX]

Well, other Facebook marketers take the same approach with testing the audiences first with all else being apples to apples. It's not really A/B testing when you are testing different audiences with the same set of parameters. A/B testing of creatives is supposed to come second, after you find the best audience, so it sounds like you are on the second step.

I have tried dozens of audiences with three different songs and always seem to end up around $.40 on the lifetime ad cost. Tried lots of different ad types that different music people said work well for music. Always looking for the key. And, yes, I've been doing CBO, conversions, landing page, etc.

Regardless, I let my adset run until 50 conversions and it is still around $.40, but I will have to see what it does going forward. It just broke 50 today. Other people without the learning phase showing have also said they aren't getting super low rates. Supposedly, it is still learning, but we can't see it. I don't know. The three other adsets I have running have lower rates, but very few results compared to this one. From my experience, if I turn off the one that is at $.40, the rates on the other ones will go up when impressions are forced their way. I will keep leaving it alone and see if it will come down on its own. I don't know what to do at this point. I really wonder if not having the learning phase is hindering the optimization on our account.

[georgik]

oh i see what you mean. my advice is to not test audiences. use as big of an audience as you can. you can even just leave interests blank and facebook's algorithm will do the rest. i have been having great success with detailed targeting expansion.

dont worry about the learning phase. its not nearly as important as people make it out to be. the BIGGEST factor for conversions BY FAR is your creative. if you want to you can drop your artist name and i can look at your creatives in the ad library and critique it

[OnyxX]

Well, of the four adsets we now have running, one is very big. Just Electropop with expansion (270 million), and it's getting all of traffic, but it's not the cheapest. There is one more narrow ad/adset that has gotten $.03, but only has 100 impressions and 3 results for that ad (a penny a piece!). Lady Gaga with expansion was also very big and not cheap. Same with Taylor Swift, Selena Gomez, etc.

We have tried different lengths of videos, and shorter videos seem to do better. We also tried introduction videos, mic vides, etc. and album artwork (with a moving background) seems to do the best. We split tested album artwork videos with different parts of the song, and then split tested the winner part of the song with different types of ads featuring the singer. The album artwork has been the winner so far. We made sure it is clear with sound off that it's a song and on Spotify, and we use a Hyppedit landing page. I did have the link pointed to the profile the first week and there were a ton of followers, but not that many saves, so I switched it to the song and followers have gone way down but saves are up.

If you have more ideas of ad types that work well, I am all ears.

I also tried making a new campaign from scratch today with Electropop, one adset with CBO on, and it went straight to active also, no learning phase. We also have no stats for quality, conversion, and engagement ranking for any of our ads. I know I've seen this before in the past, but there isn't any now for any ads, not even old ones that had it before.

Before we released this song, we experimented with ads promoting our two singles last summer and fall: audiences/ad types/lengths, etc. and then re-released the second single in the fall to do a dry run with promoting with FB ads. A album artwork video with a moving background did the best with that one, so we started with that ad type for this release.

I know it could be the song, but our save rate is 61%, even from algorithmic traffic like Release Radar. It could also be that our target audience is narrow, art pop and appealing to fans of Grimes, Aurora, Melanie Martinez, Bjork, etc. All the weirdos. Maybe it's just that. Too eclectic to appeal to a broad audience. But wouldn't Facebook figure that out? I'm just not seeing our traffic going to the cheapest results in volume.

Would you turn off an adset converting at $.40 after 4 days or give it more time? We've gotten $.70 before, so I'm happy it's not double! I could try even broader and just do pop in a separate campaign and see how it does. The narrow adsets have been cheaper, just not with volume.

The related artist on Spotify (only one so far) is HANA, and she's not real big. Very niche. Maybe the fan base is just too niche. Do you think that could be a factor? Or should Facebook be able to find them anyway?