the entire meta ads algorithm explained in 21 mins

Added: 2026-05-12

[00:00:00]Hey guys, Mark here. In this video, I'm going to be explaining how the entire meta ads algorithm actually works. A lot of people think that running ads on Facebook is just all random or that if your creative is good enough, then everything else will just magically fall into place. Well, running profitable ads on Facebook is both an art and a science. The art is the creative, but the science is exactly what I'm going to be explaining in this video, how the algorithm actually works. And more specifically than a science, it's actually an equation. And this is the one equation that is actually going to determine how profitable your ads are going to be. Now, before I break down this equation, I need to explain to you the four-step process that Facebook uses to actually show your ads. The one update that broke everybody's ad account. And at the end of this video, I'm going to put in a little something special to explain what this actually all means and how the game really works.

[00:00:46]All right, so before we get too technical, this is what you have to understand. Facebook basically goes through a four-step process to determine which ad is going to show for any particular user. Every time somebody opens up their phone on Instagram and they scroll and see an ad, behind the scenes, there's literally a 200 millisecond window where Facebook is doing all these incredible calculations and machine learning decisions to ultimately decide to put your ad in front of the user or somebody else's. So in that one 200 millisecond period, there is literally tens of millions of ads that are competing for that one impression. And these are the four steps right here. First you have retrieval.

[00:01:23]Second is light ranking. Third is heavy ranking. And the final piece is the auction. So let me break this down for you. So first up is retrieval. Retrieval is the most data inensive portion of this entire process because as I said, there's tens of millions of ads, tens of millions that are competing for one individual spot. The process of retrieval is essentially Facebook going through and ultimately deciding is the ad right now even remotely relevant to this user. Basically like is it even in the right ballpark? Now this is also the portion of the process where Facebook actually analyzes all the aspects of your creative to determine how to narrow down the audience. So you may have heard this before but this whole concept of your creative is your targeting. This is where Facebook is analyzing everything about your creative to determine who to show it to. This includes, but is not limited to, your hook, your format, the demographic of the users that are actually on your ad. So, for example, if you're running an ad that is featuring a woman that is 55 years old, Facebook takes note of that and is going to use that to power the targeting to skew much more towards women and skew much more towards older women specifically. Of course, all taken into context with what the creative is actually saying. It's also looking at your copy. It's looking at your script if you're running video ads. It's looking at your thumbnail.

[00:02:41]It's even going and looking at your landing page to determine generally who is this ad trying to target. This is a non-exhaustive list. I'm really simplifying this behind the scenes. It's a pretty substantial, it's literally an algorithm in order to determine which variable is going to correlate to which piece of targeting. And this is probably a good time to mention the Andromeda update. And drama was was an update to the sorting process of the retrieval steps specifically mostly just because there was a massive influx of AI creative that was absolutely flooding Facebook to the point where their old model of sifting through all these creatives was not becoming very effective. This is not the only reason for the Andromeda to update. But I believe this is one of the biggest ones simply due to the fact that AI has massively increased creative production for all advertisers. And now that's why Facebook is mostly looking at creative to determine targeting, not what you have going on within your adsets. That's why you've probably seen a decline in audiences that are actually able to sustain high budgets profitably. Like I rarely use audiences like interest and if I do, they're pretty capped in terms of budget or, you know, lookalikes or even custom audiences. Like I don't set any real [ __ ] targeting when it comes to my retargeting. I don't put any custom audiences in there. I literally just still run completely broad and most high skill advertisers that I speak to they're doing something relatively similar. Maybe one exclusion of purchasers but like other than that very very broad. So if that was kind of the old way of doing targeting on Facebook the new way is to do a lot of broad targeting is to do a lot of advantage plus campaigns and it's usually within reason. within reason. Giving Facebook control. Like the reason that I say it's within reason is because if you give Facebook all the control, the algorithm is good. Don't get me wrong. But holy [ __ ] the amount of suggestions that Facebook gives you that are horrible pieces of advice and are just going to drain your [ __ ] bank account. It is obscene. And it's always been this way, too. They're still a company at the end of the day, and you are a customer. You are paying Facebook. So, they still want to get your money. I'll save that for later in the video. Now, the second step is going to be light ranking. This step barely deserves a description. It's basically taking the now millions of ads instead of tens of millions and reducing it down to thousands. There's really not a whole lot more to say about the light ranking process. Now, the heavy ranking is the next step, which you guessed it, if it's going from thousands of ads, you're now trying to get to a couple hundred. Now, at this point, this is like your real legitimate competitors.

[00:05:04]Like this is the stage where competitors are actually something that you should take into consideration because you don't have tens of millions of competitors. You just don't you probably only have a handful. At least in the eyes of Facebook, you only have a handful. There's only 10 million advertisers total on Facebook, which may seem like a massive number, but also think about we are in e-commerce specifically. At least I'm in e-commerce specifically, which immediately narrows it pretty substantially. Now, the really important piece about the heavy ranking step is that this is where that equation actually comes into play. So I want to take a second to talk about that because I remember Facebook released this. They have since withdrawn it, but they released this documentation to explain how they actually assign a specific value for every single ad when they are fighting for an impression. And the one that has a higher value obviously is going to win the impression. The one with the lower value is going to lose the impression and get served a lower quality impression. All right, so this is the equation I was talking about.

[00:05:57]This is actually directly from Facebook.

[00:05:58]Somebody swiped it and it is still on their website but it's not on Facebook.

[00:06:03]But this is how it's actually working underneath the hood. All right. So the total value, this is what you want to aim on maximizing is equal to this piece first of maximizing advertiser value plus maximizing consumer value and consumer experience. So the piece to maximize your advertiser value number one is going to be your bid, which we'll get to in a second, times your estimated action rate. Since 99% of us and 99% of people watching this video are running conversion campaigns, that essentially means what is the estimated action rate to get a conversion. And if you look down here, you can see they actually break it down specifically for a conversion optimized ad, which is the estimated click-through rate times the estimated click to conversion rate. So the way to gamify this, of course, is you want to have really high CTR.

[00:06:48]Historically, Facebook just loves ads that have really high CTR. And the reason why is because that is the the most pure indication of how relevant that one particular ad is to the user.

[00:06:59]So Facebook is going to reward you typically over time and as you spend more with lower cost to advertise, lower CPMs and ideally higher quality traffic too. So this is kind of like the first piece of that. The second piece is going to be you want to have a high they call it like click to conversion. It's basically just your conversion rate which is out of all the clicks that you had, how many actually purchased. And again, if you're running a e-commerce business, this is going to be like a legitimate purchase because that's typically what you're optimizing for.

[00:07:26]But if you're in, you know, the info industry, for example, you're optimizing for calls. What is the ratio between clicks to calls that you're getting or clicks to applications or insert your conversion here, it doesn't really matter. Facebook views a lot of this largely the same, at least in the context of this equation. The way I like to think about this is the estimated click-through rate. I think it's actually more than the click-through rate. I think it's a culmination of all your soft metrics. So that is your CTR, but that's also your CPC, your CPMs, your hook rate, your hold rate, cost per 3 second video view, like all these metrics that are associated with getting somebody engaged and getting them to click. And then the second half of this is how good are you at monetizing that click, taking that click through your funnel and turning somebody who's a prospect into a buyer. Pretty straightforward. And this last piece is basically Facebook trying to save their own ass from advertisers that are incredibly aggressive and cause a terrible user experience for the users.

[00:08:19]Because if this piece just did not exist and we didn't care about the user experience whatsoever, advertisers would just flood the platform and completely just destroy Facebook as a platform because you just have shitloads of scammers. Not like there isn't scammers on Facebook anyways, but there would be way more. Like it would be like 50/50 instead of 8020, which is roughly what it is right now. So this is Facebook saying, "Hey, I want you to have the most relevant ads to the user, which is CTR, all the soft metrics, and I want you to actually get conversions for all the clicks that you're getting, which is the conversion rate. And I don't want you to piss off all of our [ __ ] users in the process." And that's the last piece. All right, so we took it from thousands of ads and now we're down to a few hundred. In fact, sometimes it could be even less than a few hundred. But now we have the final step in the process, which is going to be the auction. And the auction is exactly what it sounds like. It's all the advertisers that are left competing for that one impression.

[00:09:13]And basically, who can have the highest bid, of course, with all the context of that equation that I said previously, that's going to be the one that wins the impression. So, majority of people watching this video right now are probably running what's called, you know, lowest cost bidding or also known as like auto bidding, which Facebook is just doing all the bidding for you, trying to get you conversions for the lowest cost possible. This is, like I said, auto bidding. You can also do manual bidding, which can come in a few different ways. Manual bidding, you can do something called cost caps, which is where you set a cap on the maximum amount that you're willing to pay to acquire a customer. You can also do bid caps, which is similar except it's the maximum amount that you want to bid to acquire a customer. And there's some other stuff that Facebook is always releasing as well to help you optimize, you know, maximum value, target rowass, stuff like that. But I'll tell you this, there's a lot of people that are really obsessing about cost caps or bid caps or whatever the hot, you know, manual bidding method is at the time. I can tell you for a fact, I have seen ad accounts. I have ran ad accounts that are spending multiple five figures per day. I have seen ones spending six figures per day all on lowest cost and they're good ad accounts. They are profitable. Other people literally only run bid caps. You can see both sides of the spectrum. But the point is, you should choose what is actually going to solve the constraint in your business for you. Because I talk to people all the time. They're like, "All right, in order to get from like 1K a day to 10K per day, I just got to like media buy my way there." And like there's a little bit you can do. Don't get me wrong. I'm not the guy that's saying like media buying is dead. I'm defending media buying. At the same time, you have to think like an entrepreneur, not like a media buyer. And just to show this entire process visually. All right. This is how Facebook actually picks ads in under 200 milliseconds. So this is the first step. tens of millions of ads competing for one impression. After that, we go through the light ranking.

[00:10:57]Right now, you can see there's substantially less. Say like roughly 5,000. Next step, there's just a few hundred. Now, we're going to go into heavy ranking, which is boom. Now, we're on the final step, which is actually competing. Who is going to have the highest bid? There's very few advertisers left at this step until finally your ad is actually served. You have the highest bid, you win the impression. So, again, retrieval, light ranking, heavy ranking, and the auction as the final step there. By the way, guys, this Thursday, May 14th, at 700 p.m. Eastern, I'm going to be holding a free master class on how I create and systematize all of my AI ads from scratch to create winners predictably.

[00:11:34]Damn near 100% of my ad accounts right now are running some form of AI ads, and those are the ads that are scaling best.

[00:11:40]So, what I'm going to share with you is something I've never shared publicly live, but now that changes. So, if you want to be a part of it, I'll leave a link below in the description. Click the link, register, and I'll see you there.

[00:11:50]Okay, so now you know how this filtering process works. How does that actually affect your ability to create winning ads consistently and more importantly predictably turn out winners into your ad account? So, there's a few other things that aren't part of that four-step ranking process that are still really important to understand so that you can have the most profitable ads as possible. All right, first I want to talk for a second about the learning phase. Okay, so the learning phase according to Facebook themselves is basically a phase that your ads go through, your adsets go through, your campaigns go through, but especially at the ad level to acquire conversions so that it can learn about your customer and then you kind of exit the learning phase and go into like a more mature phase where you're getting more predictable, consistent results. Here's my beef with this whole [ __ ] term.

[00:12:34]Like that's the theory. Let me tell you the [ __ ] reality. Your ads are always learning. They never stop learning.

[00:12:39]There's not a learning phase and then they get out of learning phase and they're just going to be good forever.

[00:12:42]Everything dies at some point. It's just the circle of ads. So if you are obsessing over the learning phase, just stop. Of course, the ads initially are going to be a little bit more volatile, a little bit more bumpy, and sometimes you see improvements over time with more consistency, but sometimes you don't.

[00:12:58]Sometimes the solution is literally just to make better ads. So spending too much time on the learning phase is not super smart. The main thing I want to say here is like if you're not spending a lot of money on ads, maybe you're spending, you know, 50 bucks a day, a few hundred bucks per day, even up to like $1,000 per day, I think that's kind of like the barrier. If you see your results being like up and down, like it literally looks like a freaking sine wave, that is normal, okay? That's 100% normal. That's how the game works. [ __ ] I experience inconsistencies. I spend way beyond 1K per day, and most advertisers do.

[00:13:30]Facebook is an emotional ass platform, man. And it's a bunch of emotional human beings like you and me. Do you really expect it to be the most consistent thing on the world? If you want super duper consistent results, like go advertise on Google and uh you know, enjoy your time. All right, Google's cool. It's got its own issues, but with Facebook, inconsistencies is part of the game, but especially at lower levels of spend, you'll see people just bitching about, oh, I was at 3x rorowass and now I'm not profitable. I was like, well, how much do you spend? It's like a 100 bucks. I'm like, come on. What are we doing here? Like, what do you what do you expect? So the solution there is to when in doubt zoom out look at larger time frames with larger amounts of data and you'll see what's really happening.

[00:14:07]Instead of being like oh today was really good next day was really bad. I don't care. Look at 7 days. Look at the most amount of ad spend that you can and that's when you'll make the best decisions and not these emotional decisions where you're just going with the [ __ ] flow of Facebook taking you for a ride. Nobody wants that. All right. Next I want to talk about signal quality and your EMQ and conversions API. So, Facebook performance is often times dependent on the data that you're sending back to the ad account and back to the Facebook servers so that it can learn about your customers and reduce your cost to advertise long-term hopefully. And with that, that means that you have to be sure that you're sending the best quality data you possibly can to Facebook because the better quality data that you send them often times they're going to be able to learn more about your customers and send you better customers in the future, make your ads more efficient long term. So, you can go into your events manager and go to the purchase event specifically and you'll see something called the EMQ, which is your event match quality, and it'll be on a scale of 1 to 10. I'm not saying it's a direct correlation, like 10 out of 10 is super profitable ads, like five out of 10, you're going to be dog [ __ ] I actually think this whole concept is overrated, but I think generally speaking, it's a good thing to try to send good data to Facebook. And you can do that through conversions API, which is simply sending your data back server side and not only on the browser.

[00:15:22]For those that have been advertising for long enough, you know this became a big deal around iOS 14, which was launched in 2020, but it actually went into effect, I think, early 2021, and everybody's ad just kind of went to [ __ ] because I don't look at any conversions after the click. Like, I don't look at my CPA on Facebook at all. I always use thirdparty attribution tools, you know, whether that's triple whale, whether that's high. There's so many of these out there. Like, you can build one custom now. [ __ ] you can vibe code one probably in a few hours and have your own solution that you don't have to pay for at all. The issue is because of what the update that iOS launched all these users can now opt out of tracking from Facebook. So everything becomes really inaccurate. Now if you're able to use like conversions API any server side tracking you can improve your EMQ and get better data into the Facebook servers to help you optimize with your ad account your ads a lot better than you are right now. So to summarize this point try to send Facebook good data and also take all the conversion data with a grain of salt. Okay let's go. Now the third thing here is going to be the whole concept of creative diversity. So because of Andromeda, there was this really weird narrative where people were basically saying, "Oh, there's kind of like two types of ads that you can test." You know, you can test something completely new. We'll just call that like a new concept, a net new concept, or you can test, you know, a variation.

[00:16:32]You can only do one of the two, which it's really not that simple, but let's just say for these purposes, this is what it is. And people like, "All right, all you need to do to survive Andromeda is just like pump out more new concepts and do less variations." No, variations are still highly, highly effective. What Andromeda did is let's say like this is within an ad set and we're looking at the ad level and each one of these is a new ad, right? You have ad one, two, three, and four. So every time I launch creatives, I'm actually launching variations. So this entire ad set per se will be one net new concept, but ad number two, ad number three, and ad number four will all be variations of that original net new concept. So it's not necessarily variations of the winner. Like it's difficult to get a variation of like your top ad in an ad account to outperform the top ad. It'll do decent. It's typically not going to outperform, at least in my experience.

[00:17:22]But variations are still highly effective because you're basically getting more at bats at the concept level. Technically, you're testing a new concept, but you can test one concept with three different hooks. And it turns out that this variation actually ends up being the winner. And this original like net new concept is actually the loser.

[00:17:39]But if you said variations are stupid and they don't make any sense, then you never would have figured that out and you would have just concluded that this net new concept was a loser when in reality you're a loser for believing that. I swear I could rant on all this stuff for so much longer. I'm trying to stay focused here. Let me distill all this technical crap into something that you can actually take away. Because basically what we just did is we went through all the super autistic scientific mathematical explanations of how the Facebook ads algorithm actually works. But let's take our [ __ ] hat off for a second and think like a Chad.

[00:18:15]Okay. Three things you should take away.

[00:18:17]All right. Number one, your creative is the targeting. Stop trying to [ __ ] around so much with the adsets, targeting custom audiences, look alikes.

[00:18:24]That can be a little experimentation sandbox, but for the most part, run broad. The creative is a targeting.

[00:18:29]Number two, make good creatives that talk people from their doom scroll. Of course, easier said than done, and that's not even what this video is about, but if you just have damn good creatives. Like, that's literally half of the [ __ ] formula is just having that high CTR, high soft metrics. Of course, excluding the conversions, like that's half the [ __ ] battle right there. And then third is going to be don't [ __ ] up the user experience. Now, I have a portion of this video that I intentionally left for the end of this video for people that are genuinely serious about figuring out how Facebook works. And all the stuff that I've said so far is what we'll call, you know, buy the books. A lot of it is in, you know, Facebook's own documentation or documentation that they had years ago and they have since removed. But let me tell you how the game really works.

[00:19:09]Facebook is a for-profit company based in America. Facebook is a massive, massive company. We can assume they probably don't do everything by the books. So, what I'm really leading to here is Facebook doesn't play by the rules. I could go down the list of what this actually means, but at the end of the day, as I said, Facebook is a for-profit company, and like most companies, they are trying to get the most amount of money from the largest advertisers. And then all the smaller ones that cause like 99% of their headaches, they're not going to treat them super duper well. So, a lot of you have probably launched ad accounts and got them disabled, have ads rejected, have ad accounts banned, have business managers banned. Maybe you've had profiles banned. But you have to understand that Facebook is a company and they're a company that is public.

[00:19:52]And what happens when you run a public company is you have to answer to your shareholders, which means depending on how the earnings reports come out regarding Facebook, Facebook is either going to constrict their guidelines, their restrictions, and this is where you see ban waves come through, which is where everybody's just getting their ad account banned, everybody's dealing with business manager bans, page bans, and then a month later it's like totally fine. That's all due mostly to earnings reports that are coming out directly from Facebook. So depending on how those look, they will either be way more laid-back, way more chill, or super duper strict. That's why saying, "Oh, you can't run like this type of format of ad on Facebook. One person runs it gets approved. Another person runs an identical creative and gets it rejected." Why is that? Well, Facebook don't play by the rules. All right? So, what do you do about this? I mean, I'm not going to get too much into specifics because I don't know if I can. I think the main thing anyone can take away from this is just question everything. Like when I was talking about the Cappy and the signal quality and the EMQ score, technically speaking, like it makes sense logically, theoretically, if you're able to get higher quality data back to Facebook, they should be able to optimize better. However, I haven't seen a direct correlation between that. I have had really really good server side setups to where my EMQ was like 9.8, 9.9. The little thing on my freaking ad account was like, matter of fact, I'll show you. I'm literally at a zero at a zero opportunity score and this ad account is incredibly profitable. So all the [ __ ] from Facebook, they're still wanting to get money from you. So just because they say something or it's in their documentation or like their CEO, CMO, head of advertising, it's like this is what advertisers should do. [ __ ] it.

[00:21:30]Think for yourself. Question everything.

[00:21:32]And I want to take a moment of silence for everyone that listened to all the Facebook suggestions at the ad level that said they would lower your cost per result by 11% and then their ads went to [ __ ] RIP. Don't listen to Facebook.

[00:21:44]Other than that guys, I hope you enjoyed. Thank you so much for watching this video. See you on the next one.

[00:21:47]Peace.