The LFG Show
Talking with movers and shakers who grew up with nothing and worked their asses off to achieve success. Let's 🤬 Go!!!!
The LFG Show
Meta’s Andromeda: Spend $200k Daily Consistently ft. Dimitri Vulpe
The ground shifted under Meta buyers—and if you’ve felt sudden CPA swings, throttled delivery, or terrifying burst spend, you’re not crazy. That’s Andromeda. And in this session, Dimitri Vulpe breaks down what actually changed inside the auction—live from the stage at Affiliate World Bangkok.
We get real about why creative clones and overlapping angles now get punished, why endlessly tweaking interests is a losing game, and why small, scattered budgets confuse the model instead of training it. Then Dimitri walks through the exact structure that pulled accounts out of the spiral and made scaling predictable again.
It starts with clarity: ultra-clean accounts and ABO-based testing that isolates angles properly. One ad set. One clear proposition. Creatives that align tightly with that angle. That separation gives Meta clean signals, reduces internal competition, and stops self-throttling before it begins. When an angle proves repeatable—strong CTRs, healthy CVRs, and stability under modest budget increases—it’s stress-tested, scaled deliberately, and rolled back fast if performance slips. Predictability wins. And Andromeda rewards it.
For scale, we move into bid-cap CBOs with intentionally inflated daily budgets. The goal isn’t to spend the headline number—it’s to unlock more of the auction. When executed correctly, CPMs often drop instead of spike. Guardrails like max ad set spend caps widen reach without risking catastrophic burst spend.
From there, it’s horizontal scale: multiple domains, pages, ad accounts, and operators. That diversification reduces CPA volatility, isolates issues like flags or reviews, and keeps delivery consistent day after day. Dimitri also breaks down how AI now powers the creative pipeline, why broad targeting beats interests, and where retargeting still fits in a post-Andromeda world.
If you’re battling volatile CPAs, dead CBOs, or random spend spikes, this talk gives you a simpler structure, respect for angle ceilings, and a framework to scale with intent—not hope.
Huge shoutout to the sponsors who power The LFG Show:
🔥 Ringba – The leading pay-per-call tracking and routing platform, giving media buyers real-time call data, smarter optimization, and full control over performance.
📰 NewsBreak – One of the largest native discovery platforms, helping advertisers reach massive, high-intent audiences outside the Meta and Google echo chambers.
👉 Subscribe for more real-world media buying frameworks
👉 Share this with a teammate who needs stability
👉 Leave a review and tell us your biggest Andromeda pain point
Straight from the stage at Affiliate World Bangkok.
No money. No honey. Let’s freakin’ go.
Right, but the road to the book brother Wow Good morning Bangkok!
SPEAKER_00:How are you guys today and how was your night? Are you still hangover? I'm very honored to be here. It's my first time in Bangkok and my first time at the affiliate world Asia. I want to introduce myself a little bit, but I don't want to I don't want to go down the rabbit hole of how many things I did because I'm obsessive about business. And to keep things short, I've been running ads long enough to have PTSD flashbacks from Pixel 1.0. It was back in 2013 when the first Pixel event on Meta was released. I've accidentally turned a few small tests in multi-million dollar problems that I still manage today. And I built a portfolio of companies because I don't like having hobbies like a normal human being. So I guess I still spend more on ads than my mom thinks it's legal every single month. And I think at a certain point, apparently someone at affiliate World Asia thought it was a good idea to have me as a speaker. And now today I'm here to explain you why Meta feels like an abusive relationship. And everything was fine until this was released earlier this year. How many of you know what is the Andromeda update within the meta environment? That's it? Do you even guys run ads on on Meta? How many of you still run ads on Meta today? Well there's quite a few. Okay. So I need to help you guys understand what's Andromeda then. So it's a new release that was planned back in October 2023. So 2024 and started as a release early this year with the final update back um a final update in October this year. At first, only some of ad accounts had the rollout happening until the full 100% of them in October. You have to understand that before Meta was behaving differently. That's because the it was the pre-AI era. With the AI happening and the magnitude of ads going on and tools you guys have at your disposal, and you pumping out so many different creatives and videos every single day, Meta had something like a bottleneck, like a hardware problem. Imagine trying to play Call of Duty on a 2010 laptop. You're not gonna go very far. So they had to come up with new ideas and new new methodologies of how their model works. And also, you have to understand there's not only one algorithm, there's many models. Models that work together to optimize your lead event, which could be purchase leads, submit trials, add-to-cart, initial checkout. There's they're all different. There's not just one. So what they started to do is bottlenecking the auction system within an internal environment, decide which was the winner ad that was that should be served to the impression of the user in the real world scenario. And that's when they throttle you in order for you to stop releasing so many bullshit ads that are clones of whatever you guys are doing. Otherwise, the system just breaks and it's not working any longer. So before we have the old way where the meta behave like a winner takes it all arena, you dropped a single ad, all the budget on that, and it's still optimized very, very well. Earlier this year, when this stopped happening, we were all like, oh, we're fucked. We need to find a new job soon. Except there's strategies to optimize this too. With the new way, the as I told you, Meta had to rebuild the engine. It's an X-Gen model that it's designed to interpret massive creative diversity uploaded every day. So let's talk about how we managed to um we managed to solve this situation. Because what happened to us was that nothing was working anymore. We were like, okay, we have a big team, something's going on, it's something fishy. Our creatives have a CPA of$20 a day and$100 the day after. Something is going on. We felt that the system was broken. And we had we have a great relationship with Meta, reps in Dublin, their tech team. We were in constant communication. We were like, what's going on? And slowly and surely we understood how to optimize our ad accounts and our structure. So we now use four strategic phases within the optimization environment that we use in our company. So at first we have the signal layer, then we move to the stress layer of the signal layer, and then we have the scaling layer, which is the most important part. And then we have this what we call distribution layer. This is just some phenomenal names for something very basic. So actually, we solved the problem by simplifying everything we do and work actually much less than before. So we went back to how we find angles and prospecting creatives that actually work in a discovery area that rely on the ABOs. So we stopped using CBOs because CBOs were throttling our ad creative angles that were pumped out to the audience. Because of how the new model works. So what we do is extremely simple, simple. We keep the ad accounts very, very clean. At most, we have two or three campaigns inside the ad account itself. You won't see any of our other ad accounts with tens of different campaigns, random tests, random scenarios. We don't do targeting anymore. We don't, at most, we disable some Advantage Plus options that are very invasive on the ecom side and also on the Legion side. But it's actually very basic. We need to go back to the basic where creative is extremely king, and we need to divide how we test angles. So we would in our in our environment we create different ad sets that are angles, which means I know if you do neutra, you will have mom that works at McDonald's and did shit all day. You would have different creatives for that angle inside that ad set. Then you would have another ad set with a total different angle with the creatives inside that ad set that actually are very correlated between them. But still you need to make them different. So what we do now is this we have an ABO campaign, it has different ad sets, it could be tens of them with a dedicated budget. So in each asset, though, we test different angles, and inside the angles, we have a variety of creatives for that specific angle script, hook, or you name it. The important part though to make this system work is that all the angles need to be extremely different. Because otherwise, Andromeda will still throttle your assets because there's an internal auction happening also at the asset level before they pump out the budget that you want to spend on said angle. So why not CBO for testing? That's because the method doesn't work anymore. And the CBO mixes intent because it decides where to where should allocate the budget, and what you would see is an extremely volatile CPA and birth spending issues. How many of you had the past six to eight months birth spending issues on the CBO campaigns at a either small or big volume? One, two, three, four, five, six. That's a lot. Okay. I bet you have no idea why this is happening. This is happening because the new model has an internal auction system. This wasn't a thing before. It's based on the entity ID. And actually, which is funny enough, is that inside Meta, the developer department team, they have no idea what's the entity ID and where does this stop? It's ad ID, it's ad set, it's ad account, it's BM, or it's organization child ID, or is the org ID. Nobody can answer that. So we need to be extremely careful of how we set up our structure. Because this is happening when you have a CBO that stops spending in the real-world auction and starts pumping out new creatives from a new asset that didn't spend for the entire day, and you will randomly burst two to three thousand, five thousand, ten thousand dollars in an hour. Sorry. Once we do that, we have a small and easy process of how we graduate the ABOs that are winning. We consider 20 to 30 conversion at 200 ROI, a good creative that has a great CTR and a conversion rate on the entire offer. It holds budget when increase because that's the strategy we use next to still scale the ABOs, but still keep them in one single campaign. And when the performance is repeatable, which means that if we add new ads inside that ABO that are for that specific angle, they still convert. So this is the second phase, which we call the stress layer. You have to understand that every ABO, every creative, every angle, they have a certain ceiling of how much you can scale. So what we do, we just bump the volume and the budget gradually, slowly by little. Every couple of hours we increase the budget to not shock the system because Andromeda doesn't love that. Meta loves when everything is linear, when everything is predictable, and when conversions are happening at a very, very, very fast pace. Okay. What happens when we raise the budget on the ABO and stops converting? We just go back. We just go back to the budget it was converting before, and Roma. If you didn't fuck up your ad accounts and you started testing random shit without being actually very, very, very specific on how you have the architecture of your ad account. So if you go back, usually this starts to convert well as it was before. Now, I want to talk, I want to walk you through how we scale when we have an ABO that actually is winning out of maybe a dozen 20, 30 ad sets that you might want to scale. You don't want to use all of them. You want to use chunks of four to five max on the a CBO bit cap campaign. It doesn't have to be accelerated, it has it needs to have a very inflated budget. And this is how we do it. We have four or five proven angles that are extremely different. We set a CBO with a huge budget, and we have a bit cap that maintains the constant, the constraint. We use something called inflated method. We go anywhere from 150,000 to 800 or even million dollars a day of budget on the CBO on one single campaign. And we don't aim to spend that amount. I mean, we'd we'd like to, but that's not possible. We do that to actually have access to the entire auction system of Meta. When you spend little, Meta knows that you want to spend little and gives you access to only a certain amount of auction that you can bid to. When you have an extremely high budget, what you will experience is that CPMs drop half. You might be experiencing very high CPMs if you do this method, you probably will see half the price. Just because you're actioning the entire ecosystem and you get also views that might not be interested, but if your offer actually is very good, they will convert. So what we do next is we set a guardrail. Because when you're working with a very high budget, it's very dangerous. I we've had days and hours where we probably burst our ad spend by$50,000 an hour with this method. So what we do, we set a guardrail with a max ad set uh limit where we raise it little by little, every now and then every hour, based on the results of the BitCAP campaign itself. This protects everything while we uh stabilize the CPA. And sorry, this is a recap of how we structure it. This lets Meta explore carefully what's happening inside the auction system. We have a CPA that stabilizes and we inflate the budget of the limited asset that we set. This goes anywhere from$50,000 to$100,000,$200,000 a day if your offer, e-comproduct, brand, lead gen method, whatever you guys do, actually is kicking in very, very hard. So this works only if your offer or your company or your brand has a very good track of history. Okay. Why do these big budget works after Andromeda? It's because Andromeda likes very high volume amount of budgets, it's it has a predictable structure, and it has a very large data set patterns. So instead of shrinking all the assets, all the creatives, testing ages, genders, you need to give the system a lot of freedom to explore and optimize. It becomes very stable once you do that. With small budgets, that's not gonna happen. And the system itself gets very confused. What we do next is what is called uh horizontal scaling. That's actually very easy, but it's very hard to put in in place because you need to be in the position of creating multiple assets. You would need multiple domains, multiple profiles, multiple ad accounts, multiple pages. That's because the risk of going out of business because the system still hallucinates today, is very, very high. So, what we do is we isolate whatever winning offer we have, we replicate it across many different assets and media buyers. This allows us to put us in a position where we can maintain a very, very consistent delivery throughout the days of the year. So I would advise you to do that. Maybe for e-commerce it's not gonna be easy to do that, but if your e-commerce becomes an offer for a lot of affiliates, that's very, very, very good. Why does vertical scaling alone fail? And that's where most affiliates struggle, is because their assets are very limited. You get into throttling issues like the one I explained earlier. You have page issues, like flags, spamming. You get domain flags that you might not be aware of. That's when you get a lot of rejections on meta when you launch new campaigns. Even with new assets, new pages, new, it's maybe the domain that is flat. You have very volatile CPMs because you have no um, you get flagged, so you get high CPMs. You have random reviews, could be manual, could be AI. It's mostly AI lately, and it's very, very aggressive. And you get account restrictions. How many of you had any ad account bans? You can be honest, you know, there's no there's no police here. Okay, and profile restrictions. I had a lot of troubles with my profile being restricted from advertisement. So I guess apparently matter rewards this kind of behavior and system that we built because each account is internally clean and it has a fast learning process and curve. The distributed spend has lower variance for us, and the bad days of each of one ad accounts that get flagged are very isolated. They're not contaminated to each other, they never poison the system. And this is when the CPA stability actually increases, and the auction diversification uh improves the delivery. And to be frank, this is all year round for us, and this is how we manage to be spending from January to today an average of$200,000 a day. And we apply this system on different business models. It could be paper call, it could be Legion, it's e-com, it's SaaS. So it works. I advise you to apply, and I advise you to use the strategy to simplify your ad accounts and to scale your business. So to have a final recap of what I've said so far is test your ABOs, keep it simple. Try to raise the budgets to the ABO and surf them until you until you reach the ceiling point of that particular angle. If it doesn't work, reverse go back, reduce the budget. It's that easy. You don't have to be Einstein today to run ads. And then when you're ready and you have enough conversions, you can scale with the bed cut method that I showed you. There you have the guardrails I've told you to be to be using, because otherwise you will get burned and you will be spending budgets that you are maybe not uh being. To handle. And then we have the horizontal scaling through multiple assets. Maybe it's like ABC for someone, but maybe someone didn't thought about it. So that's food for thought. So I told you what I needed to tell you, and I'm gonna finish with this sentence that is buying on meta is like navigating a woman's feelings. The answer lives in the signals you didn't measure, the timing you didn't respect, and the creative you didn't test. Thank you for today. I um I hope you had very good insights, and I hope this was helpful. For us, resetting all the balances we had with Meta actually worked better than before. And we thought we had a failing business back in March. We struggled to make profit, even though we had still very high volume, but the problem stayed within the methods. We had to reset everything we knew about Meta, we had to rethink our creative process, we have to, we had to knew how to integrate the new AI models of creative processes that we integrated in the company, and we had to teach our media buyers to media buy again. They became prompt engineers, they became better architects of their ad accounts, everything became very consistent. At first, I was extremely, extremely frustrated about like why is Mark hating on me? Why is he releasing every now and then new models? Why is this happening? And then it clicked me. If I were a business owner like him, maybe one day I'll be, I would still need to find a way for my company to cost less. Hardware is extremely expensive, CPU is extremely expensive, processing is extremely expensive, with so many brains around the world using my platform and pumping a lot, a lot of more creatives inside my ecosystem, my ecosystem stops performing and starts breaking. So I now understand exactly why he did that. Why the company, I don't know if it was him, probably. Uh I understand why this had to be happening. With the amount and magnitude of ads will pump out with automatic tools that we will experience in the next decade. Probably we will not be still having media buyers within a few years. And the system has to protect the environment because this could handle a trillion of new ads, but there's only a amount of eyes in the world that can see the ads, which is an average of 36 times per day. So if you have a million viewers, it's gonna be 36 million impressions per day. The ads will be a million times more. So we need to throttle that. We need to make the system work, otherwise, everything's gonna break. So I hope this left you with a little bit, a little bit of insights of how Andromeda and Meta works. Meta team told that they're working on a fix. It's a it's actually a bug that was acknowledged of the birth pending issues, but I'm sure they will not be able to find it. Thank you. Hope you had a nice day.
SPEAKER_01:You know how I don't I trust this guy. He dresses better than me. Come on, look at this. But hey, you're not allowed to go anywhere yet. Okay, have some QA's. Okay. So, guys, put up your slide around and uh put in the questions and we have a shit ton already. So, first question, Mr. Dimitri. If you have many products with different target audiences, is it better to run separate campaigns for each audience or can they be combined under one campaign?
SPEAKER_00:Based on my experience I had so far with the new methods, they still need to be combined within the same campaign. And you need to be able to diversify how you target. I mean, you don't even have to target, they just need to be different assets inside the same campaign with you if you either use catalog, different creatives, flex, or whatever you're comfortable with. I would still suggest you to keep it very, very simple and very clean.
SPEAKER_01:And uh how many campaigns do you suggest? And uh how do you suggest to scale those campaigns?
SPEAKER_00:As I told you, it's just a couple of campaigns at max three on your ad account. So I would suggest you using ABOs to test them. You don't have any limits on how many ABOs you could test in your ad account until they die. So it could be, I've seen examples of people running 10 ads and people running 100 assets inside the same campaign.
SPEAKER_01:And um what qualifies as a different angle? Can you give some examples of different angles?
SPEAKER_00:I can give you an example. So, first of all, you need to understand what angles are is angle is something that is very unique as a proposition of marketing, advertisement, than the other. It could, as I as I mentioned you earlier, it could be a woman, a script about a woman working at McDonald's, eating fast food and want to lose weight for neutra. Same example applies for still neutra. Another angle could be a dad that is uh I don't know, food truck driver, and still it's like shit and needs to lose weight because otherwise his wife leaves him. So that's a different angle. Different angles mean different topics and how you market to towards your audience for your product.
SPEAKER_01:And you know, we have to talk about AI, right? So, how much of uh AI are you actually using in your angles in your content creation?
SPEAKER_00:How much we use it?
SPEAKER_01:Yeah, how much AI do you use?
SPEAKER_00:Probably I'm gonna be uh I'm gonna be saying if it's UGC for e-commerce, it's 50%. If it's b-rolls, Legion, if it's SaaS, it's 100%. We don't use any more like real real clips. We create them with AI. We use a stack of different AI models that we combine together to get a good result in our in our ads. So let's talk about the technical stuff.
SPEAKER_01:What percentage of your spend is uh ABO versus uh BitCap?
SPEAKER_00:So the bit cap is actually where you do the real money. There's no percentages that are very strict. You can still scale ABOs at very high budget, but the bit cap is going to actually scale your business. So it could be 80-20, 70-30, 90-10. It really depends on the performances you guys have.
SPEAKER_01:And for your assets, are you choosing specific interests or did you let them in broadcast?
SPEAKER_00:Yeah, yeah, yeah. From our experience, you can't run interests anymore. They they they will simply don't work.
SPEAKER_01:So you're basically using broad, yeah.
SPEAKER_00:And do you use flexible format ads or are they working like flexible forms? We try to disable most of it of the new formats that Meta tries to push on us because they're still very inconsistent. We still test it and use it in some certain cases, but I wouldn't start using those if you don't have um a track of history and conversions on your ad account. It's it didn't work for us.
SPEAKER_01:So someone asked also why don't you prefer Costcav?
SPEAKER_00:Probably because I'm a bit cop lover. We use cost cap. I'm not gonna lie. I myself tend to use bitcops much more because they're more predictable.
SPEAKER_01:So I like this question. Do you like tofu, mofu, pofu contents? According to the stage of the photo, how we can respond and folks involved.
SPEAKER_00:I know what it is. I I love the question. It it conflicts a lot with the new strategies we have to use on Meta for our business. So I would say whatever you're comfortable with, try it. If it doesn't work, use my method and let me know how that goes.
SPEAKER_01:And one more last question. Do you still think retargeting works? Yeah.
SPEAKER_00:Awesome. Yeah. We still use retargeting for e-commerce. Okay, I'm gonna say it has Meta has an internal retargeting method when you don't use retargeting. They they still show your ad to the same client that clicked on your ad because it might have a high intent of converting. So it has an internal, uh, an internal retargeting method. But we use retargeting across different traffic channels.
SPEAKER_01:Thank you so much, Dimitri. There is a lot more questions, but thank you. If you want to ask more questions, we'll be here and hanging out. So again, give a huge matter of love for you. Thank you so much. Thank you, brother.