Full transcript
0:00the original structure. stands
0:03for ad set budget optimization, and it's
0:05the original way every Facebook ad
0:07account was built. Here's how it works.
0:09You create a campaign. Inside that
0:11campaign, you create ad sets.
0:14And each ad set gets its own budget that
0:16you set manually. 50 bucks a day on ad
0:18set A, 50 on ad set B, 50 on ad set C.
0:22Meta cannot move that money around.
0:23Whatever you set is what gets spent.
0:26You're in complete control. For years,
0:28this was the only way to run ads. If you
0:30wanted to test broad targeting against a
0:32look-alike, you'd put each one in its
0:34own ad set with the same budget. Clean
0:36fair test. The reason still has a
0:39place in 2026 is isolation. When you
0:41absolutely need a specific audience to
0:43get spend, is the only way to
0:46guarantee it. Retargeting is the big
0:48one. If you've got a small custom
0:50audience of cart abandoners, CBO will
0:52starve it because the algorithm sees the
0:54bigger audiences as cheaper. [music]
0:56forces the spend. Same for clean
0:58creative testing.
0:59You want every concept to get a fair
1:01shot? Put each one in its own ad set
1:04with the same budget. No algorithmic
1:06favoritism. The trade-off is efficiency.
1:08You're choosing certainty over
1:09optimization. And in an Andromeda world
1:12where the algorithm is genuinely better
1:14at finding buyers than you are, that's a
1:16real cost. is no longer the default.
1:19It's a specialty tool. CBO, the
1:21algorithm, takes over. CBO is campaign
1:24budget optimization. Meta renamed it
1:26Advantage Campaign Budget in the new
1:27interface, but it's the same thing. And
1:29it changed everything. Instead of
1:31setting budgets on each ad set, you set
1:33one budget at the campaign level. Meta
1:35then decides in real time how to
1:37distribute that money across your ad
1:38sets. Ad set A converting cheap? Meta
1:40floods it. Ad set C dying? Meta cuts it
1:44off. You don't touch anything. The
1:45algorithm runs the wallet. And here's
1:47the wild part. Studies show that in a
1:50typical CBO campaign, 20% of your ad
1:52sets eat 80% of the budget. That's by
1:55design. CBO isn't trying to test fairly.
1:58It's trying to win, which is why CBO is
2:00incredible for scaling and brutal for
2:02testing. If you've already got proven
2:04winners and you want Meta to push spend
2:06toward whatever's hottest, CBO does that
2:08better than you ever could manually. But
2:10if you load it up with five new untested
2:12ad sets, Meta might spend 95% of your
2:14budget on one of them and leave the
2:16others with nothing. Zero usable data on
2:18the losers. And here's what most
2:19beginners miss. Changing budgets on a
2:21live CBO campaign resets the learning
2:24phase. Pausing, adjusting, targeting,
2:26all of it. Meta needs stable conditions
2:28to optimize, so you set it up, you walk
2:30away, you let it cook for at least seven
2:32days. There's also a guard rail most
2:33people skip. You can set ad set minimum
2:35spends inside a CBO campaign. That
2:37forces Meta to spend at least that much
2:39on the ad sets you care about. Used
2:41right, [music] you get the best of both
2:42worlds. Most six-figure per day accounts
2:44in 2026 run their scaling spend through
2:46CBO. It's the engine. is the test
2:49bench. 111, the consolidation structure.
2:52111, one campaign, one ad set. Every ad
2:55you've got stacked inside that one ad
2:57set. This is the structure that took
2:58over after iOS 14 broke everything. The
3:01whole point is signal consolidation.
3:02When Apple let users opt out of
3:04tracking, Meta lost a huge chunk of the
3:06data it relied on to optimize. Suddenly,
3:08campaigns split across 10 ad sets
3:10couldn't generate enough conversions per
3:12ad set to actually learn. Each ad set
3:14needs roughly 50 conversions a week to
3:16exit the learning phase. Fragment your
3:18spend across too many ad sets and none
3:19of them ever get there. So, the smart
3:21move became collapsing everything down.
3:23One broad ad set, 20 plus ads inside.
3:25All the signal pooling in one place. And
3:27for a while, this was the structure. But
3:29111 has a ceiling. Once you load more
3:31than maybe 10 or 15 ads into a single ad
3:33set, Meta picks its favorites. Two or
3:35three winners. The rest barely get any
3:37impressions. You're paying for them to
3:39exist, but they're not getting at bats.
3:41And the moment your winners fatigue, you
3:42don't have backups warmed up. The whole
3:44account stalls. That's why 111 started
3:47losing ground around 2024. Meta's own
3:49data showed one ad set with 25 creatives
3:52generated 17% more conversions at 16%
3:55lower cost than five ad sets with five
3:57creatives each. The principle was right,
3:59consolidate the signal, but the
4:01execution had to evolve. 155, the
4:03testing hybrid. 155, one campaign, five
4:06ad sets, five ads inside each ad set.
4:09This is the structure most serious media
4:10buyers actually run in 2026, and the
4:13math is simple. 111 consolidated too
4:15much. You couldn't test. So, instead of
4:17giving every concept its own ad set or
4:20stacking everything in one pool, you
4:21split the difference. Each ad set holds
4:23one core concept. Inside that ad set,
4:26you've got five variations, different
4:28hooks, different opening lines,
4:30different formats. The ad set is the
4:32concept test, the ads inside are the
4:34variation test. You get five at bats per
4:36concept. Even if your strongest hook is
4:38a loser, your second or third might be
4:40the actual winner. Meta picks between
4:42them automatically. This is the sweet
4:43spot for accounts spending anywhere from
4:45a thousand bucks a day up to about
4:4620,000 a day. Enough segmentation to
4:48learn something, enough volume per ad
4:50set to actually exit the learning phase.
4:52And when you run this on top of CBO at
4:54the campaign level, Meta routes the
4:56budget to whichever concept is
4:57converting cheapest. The trap people
4:59fall into is over segmenting. They make
5:01it 11010 or 1255, and suddenly the
5:04budget is too thin per ad set. None of
5:06them generate enough data. Everything
5:07stays in learning forever. Three to five
5:09ad sets is the practical ceiling under
5:11Andromeda. Anything beyond that is
5:13fragmenting your signal for no payoff.
5:15Keep it tight. Let the algorithm do its
5:16job. If you're loving this breakdown,
5:18make sure to subscribe and hit the bell.
5:20We've got more coming. Let's keep going.
5:22ASC, the black box. ASC stands for
5:25Advantage Plus Shopping Campaign, and
5:27this one is different from everything
5:28else on this list because ASC isn't a
5:30structure you build, it's a structure
5:32Meta builds for you. You don't create ad
5:34sets, you don't pick audiences, you
5:36don't choose placements. You upload
5:37between 20 and 50 creatives, set a daily
5:40budget, set your existing customer
5:42budget cap. That's it. Meta handles
5:44every other decision. Targeting,
5:45placements, bidding. Internal Meta data
5:47shows ASC delivers 17% lower cost per
5:50acquisition than manual campaigns. And
5:53for e-commerce accounts pulling in more
5:54than 50 purchases a week, ASC has
5:56basically become the default. It's the
5:58campaign type Meta pushes hardest
5:59because it gives them the most control.
6:01And here's where you have to be careful.
6:03ASC works incredibly well when you have
6:05clean conversion data, an existing
6:06customer list uploaded, and a real
6:08product catalog connected. It tanks when
6:10any of those are missing. The existing
6:12customer budget cap is the one lever you
6:14absolutely have to use. Without it, Meta
6:17will spend 40 or 50% of your budget
6:19retargeting people who already bought
6:20from you. Your reported ROAS will look
6:22great. Your actual new customer
6:24acquisition will be a disaster. The
6:26recommended starting cap is around 25 to
6:2830% for most accounts. Drop it lower if
6:31you want pure prospecting. The other
6:32thing to know is ASC needs volume to
6:35work. The official threshold most
6:36agencies use is 50 purchases a week.
6:39Below that, stick with manual. Above
6:41that, test ASC in parallel and see if it
6:43beats your existing setup. ASC is a
6:46tool. It's powerful when fed properly
6:48and expensive when it's not. Testing
6:49plus scaling, the two campaign split.
6:51This is where most agency accounts in
6:532026 actually live. The two campaign
6:56split. You run one campaign that exists
6:58purely to test new creative, low budget,
7:00maybe 100 or 200 bucks a day. Brand new
7:02ads get launched here every week. Some
7:04flop, some show promise, a few become
7:06winners. Then you run a second campaign
7:08whose only job is to scale those
7:10winners. Higher budget, sometimes much
7:12higher. Often using CBO so Meta routes
7:14spend to whichever proven winner is
7:16hottest. The testing campaign is the
7:17factory. The scaling campaign is the
7:19distribution line. Creative flows one
7:21direction. Test it cheaply, validate it,
7:24then migrate it to where the real money
7:26lives. And the reason this dominates
7:28among high-volume creative shops is that
7:30Andromeda made creative volume the
7:32number one lever. Brands testing more
7:34than 20 new ads a month see roughly 65%
7:37higher ROAS than brands testing fewer
7:39than 10. You need a structure that lets
7:40you launch volume without polluting your
7:42scaling environment. If you just dropped
7:44every new untested ad into your main
7:46scaling campaign, you'd reset the
7:48learning phase constantly and tank
7:49performance. Separating testing from
7:51scaling solves that. The trade-off is
7:53what people call the migration tax. When
7:55you copy a winning ad over to scaling,
7:57it usually dips for a few days before
7:59stabilizing. Some people use post ID
8:01duplication to preserve the social
8:03proof. Others accept the dip. The
8:04pipeline only works if you keep feeding
8:06it. Duplication, vertical scaling.
8:08Duplication is one of the oldest scaling
8:10tricks in the book and it still works in
8:122026. Here's the play. You've got a
8:14winning ad set spending 200 bucks a day
8:17and crushing it. You want to push more
8:18money through, but you've learned that
8:20big budget jumps reset the learning
8:21phase and tank performance for a week.
8:23So, instead of doubling the budget on
8:25the existing ad set, you duplicate it.
8:27Make an exact copy, same creative, same
8:29audience, same everything, new budget.
8:31Now you've got two ad sets running the
8:33same thing, each at 200 bucks. You can
8:35duplicate again and again. Suddenly
8:37you're spending a thousand a day on what
8:40was a 200 a day winner. No learning
8:42phase reset. It works because Meta
8:44treats each duplicate as a fresh entity
8:46in the auction. The original keeps its
8:47momentum. The duplicates start their own
8:49learning, but they're using a creative
8:51that's already proven to convert. The
8:52downside is auction overlap. You're
8:54competing against yourself. Sometimes
8:56one duplicate eats another duplicate's
8:57lunch. Sometimes the new ones never
8:59stabilize and there's a ceiling. Past a
9:01certain point you're just splitting the
9:02same audience pool into smaller slices
9:04and CPMs climb. Most experienced media
9:07buyers use duplication as one of several
9:09scaling tools, not the only one. But for
9:11accounts where you've got one clear
9:12winner and you need to scale fast
9:14without breaking it, duplication is
9:15still in the playbook. Omni, the
9:17consolidated account. And then there's
9:19the structure that's emerging at the
9:20very top. Call it Omni, call it the
9:22consolidated account, call it the
9:23Andromeda end game. The idea is brutal
9:26in its simplicity. One campaign, one ad
9:28set, 20 to 50 creatives stacked inside,
9:30broad targeting, Advantage Plus enabled.
9:33That's the entire account. No
9:34retargeting campaign, no prospecting
9:36campaign, no separate testing setup,
9:38just one massive funnel with the
9:39algorithm doing all of the work. And the
9:41reason this is showing up on accounts
9:43spending six figures a day is that
9:44Andromeda fundamentally changed what
9:46targeting even means. Your creative is
9:48the targeting now. Meta reads every ad
9:50you upload, the hook, the visual, the
9:52voiceover, and it decides who to show it
9:54to based on what the creative is saying,
9:56which means the more diverse creative
9:58you give Andromeda, the more types of
9:59buyers it can find. Brands running this
10:01structure are testing 20 plus new ads a
10:03month. Their creative library is the
10:05entire strategy. The campaign structure
10:07is just plumbing. And here's the catch.
10:09This only works once you have proven
10:11creative volume and clean signal data.
10:13If you copy this structure as a beginner
10:15with three ads and no track [music]
10:17record, you'll burn through your budget
10:18in a week with nothing to show for it.
10:20The bigger principle to take away is
10:21this. Don't pick a structure because
10:23it's trending. Pick the structure that
10:25solves the actual constraint in your
10:26account. If you can't produce enough
10:28creative to feed Omni, you have no
10:30business running Omni. The structure
10:31follows the operation, not the other way
10:33around.