YT YouTube2Text

YouTube2Text → Transcripts → How To Structure A Meta Ads Account At Every Spend Tier In 2026

How To Structure A Meta Ads Account At Every Spend Tier In 2026 — Transcript

by Blue Sense Digital · 21,727 words · 3,138 segments · language en · Watch on YouTube

Open in transcriber Summarize with AI Download SRT/VTT

Full transcript

0:00If you run Meta Ads as an e-commerce
0:01brand, this is the only video that you
0:03need to watch on Meta Structure. Whether
0:05you're just starting a new ad account or
0:06you're spending a million dollars a
0:08month on Meta, we'll be running through
0:09every level of structure, everything you
0:11need to know so that you can maximize
0:13efficiency on the platform. There's
0:14three claims on account structure that
0:17are all true at the same time. And by
0:19the end of this video, you'll understand
0:21why three of these claims all make sense
0:23contextual to your ad account. Claim
0:25number one, consolidation will always
0:27beat segmentation within any meta ad
0:30account structure. So if you have fewer
0:32campaigns, fewer adsets, you'll
0:33generally get better performance, but
0:35it's not necessarily the right thing to
0:36do. Number two is that media buying
0:39actually still matters at higher ad
0:41spend levels. Now, if you're spending
0:42$100 a day, it's probably not that big
0:45of a lever when it comes to growing the
0:46account and growing the business. But if
0:48you're spending $100,000, $200,000,
0:50$300,000 a month, maybe you're buying is
0:52still going to give you a 10 to 20%
0:55efficiency lift if you do it correctly.
0:57And then third is there is actually no
1:00universal perfect account structure that
1:02fits every business. The reason why
1:04there is a thousand YouTube videos on
1:05how to structure better ad campaigns is
1:07because there is a thousand different
1:09types of businesses. And so dependent on
1:11the nuances within your particular
1:13business, it will change how you think
1:14through structure. I'm going to be
1:16giving you through the way that you
1:18should think through the problem so that
1:20you yourself can make the structure
1:22yourself. Over the next two hours, I'll
1:24walk you through how Meta decides where
1:26your money is going. The three spend
1:28playbooks if you're doing under 50k a
1:30month in spend 50 to 250 and 250 plus.
1:33The difference between ABOS and CBOS and
1:34when you should use each. Three settings
1:36that you should be turning off right
1:38away as you finish this video. How to
1:40design high performing adsets. How much
1:42spend should go towards existing
1:44customers and different audiences?
1:46What's the testing budget look like? How
1:48do you math that out? How do you back
1:49propagate from your goals? And then
1:51number four is I'm going to give you
1:52four diagnostic questions that you
1:54should always be asking yourself. Meta
1:55is optimizing for one thing in 2026 and
1:58honestly forever. It's revenue per user
2:00per minute. Now, there's two ways that
2:03Meta can grow this number. Number one is
2:05they increase ad inventory. And so they
2:08simply put more ads onto their platforms
2:10or they'd buy more platforms where more
2:12ads can serve. Now you've actually seen
2:14this be the case over the course of the
2:15last six years. I remember six years ago
2:17I would get one ad in every five to six
2:19posts. Now sometimes you get triple ads.
2:22You're scrolling and you'll get an ad,
2:24another ad, and then another ad all in a
2:26row. And that's meta increasing the ad
2:28inventory so that they can maximize
2:30revenue on the platform. And then the
2:32second is if they can't increase ad
2:34inventory any further if the platform's
2:36starting to get not enjoyable to use
2:37because there's just so many ads. Well,
2:39the other thing that they can do is they
2:40can just increase the cost for the
2:42advertisers to serve, which is your
2:44CPMs, your cost per thousand
2:45impressions. And so the same amount of
2:47ad inventory, but let's make it more
2:49expensive for everyone. And you see this
2:51in year-to-year CPM inflation. We have
2:53over 150 million in ad spend connected
2:56to our business manager. and we can go
2:58and do an aggregated ad report and look
3:00at what CPMs have looked like over the
3:02course of the last three years. And it's
3:03cyclical with obviously Black Friday,
3:05but every year it goes up and what you
3:07end up seeing is there's about 20 to 30%
3:09inflation in CPMs. Now, obviously
3:12there's a little bit of natural
3:13inflation to the dollar that you need to
3:15factor out of that, but still CPMs are
3:17going up faster than inflation is. And
3:19the reason for that is that Meta has to
3:22continue to publish good quarterly
3:24earnings because they're a publicly
3:25traded company. And the way that they do
3:26that is they need to increase revenue on
3:29their biggest product, which is the ad
3:31product. Now, this seems all very doom
3:32and gloom and like, "Oh, Meta's against
3:34you. Everything's becoming more
3:35expensive. It's a terrible platform.
3:36Don't spend on it." That's not
3:37necessarily true because if Meta
3:39increases CPMs, everyone just becomes
3:41unprofitable. So, they can't do that.
3:42They can't just make all their
3:43advertisers unprofitable or else people
3:45will stop spending. And so, what they
3:47have to do when they increase CPM is
3:49they also have to increase expected
3:51conversion rates or ROI of the platform.
3:54So they need to effectively more
3:56efficiently pull dollars out of users on
4:00the platform and transfer them to
4:02advertisers so that you get conversion
4:05rates that outweigh the CPM increase.
4:07The reason why all this matters and the
4:09reason why this context matters is that
4:10every time Meta rolls out an update,
4:12whether it's Andromedor, whether it's
4:13Gemmin, whether it's one of the smaller
4:15updates that you've never even heard of,
4:17what they are trying to always do is
4:19increase the efficiency of ad serving so
4:22that then they can make it more
4:23expensive for you to actually place.
4:25Now, a core principle of setting up
4:27structure on the account is
4:28understanding that what you see in the
4:30platforms is not necessarily what Meta
4:32is optimizing. And that's where you get
4:34a lot of poor decision-m being made
4:36within account structures or within
4:38decisions as to where budget should
4:39flow. There's a lot of stuff that you
4:41see that's not necessarily reality. And
4:44so an example is what you see in the
4:46platform is lastclick attribution.
4:48Meaning if a user clicks on multiple ads
4:50and then buys the purchase the
4:52conversion value will just go to the
4:54final ad that they touched. But in
4:56reality Meta is optimizing across
4:59multi-touch attribution. Meta knows that
5:01just optimizing towards the last click
5:03is meaningless when there was all of
5:05these prior clicks and prior
5:06interactions that led up to the
5:08conversion. And so what you end up
5:09seeing is that if you have three ads
5:11here and the clickfunnel looks like this
5:13and then they ultimately buy, you on the
5:15surface will go, "Ah, this is the ad
5:17that's performing killer." But Meta will
5:19still be distributing spend here because
5:21it sees that a click actually occurs
5:23here, click occurs here, and then the
5:24final click occurs. And so in the back
5:26end, it's likely allocating one/ird
5:29credit to each one, which is why spend's
5:31getting distributed in the way that it
5:33is. So another example of what you see
5:34is that one ad gets all the credit. But
5:37the reality is that Meta is optimizing
5:39across a chain of impressions. What you
5:41see is ad level return on ad spend. So
5:44you're going down here and looking at
5:45rorowaz on these individual ads, but
5:47what met is actually optimizing for is
5:49your CPA target at the adset level.
5:52because the bidding and optimization
5:54actually is inherited from the adset,
5:56which is why you don't set cost caps at
5:58an ad level. If you're going to set cost
5:59caps or any kind of more complex bidding
6:02approach, you're going to do it at the
6:03adset level instead. Now, if we go back
6:05to this concept of sequencing where a
6:07user might pathway through multiple ads,
6:09all of the ads, let's say, are spending
6:11$5,000, but they're at very different
6:13returns. When you look at this on the
6:15surface, you make an obvious decision if
6:17these are all sitting under the one
6:18adset, which is let's turn this ad this
6:20ad off. Let's put all the spend here.
6:22Now, besides sequencing, besides the
6:25fact that, well, these ads are probably
6:27doing a little bit of heavy lifting, we
6:28need to think about why is Meta
6:31distributing spend to these two ads,
6:33well, there's something called the
6:34breakdown effect. And this is public
6:36documentation by Meta. You can go and
6:38Google for it right now and pull up
6:39their actual uh page on this. And the
6:41idea is that when you use any of the
6:43breakdown features in Meta Ads, you can
6:46go to breakdown and you can click on age
6:47or you can click on gender or you can
6:49click on placement type or you can click
6:51on region. Okay, there's all these
6:52different options as to how you can
6:54segment and break down the data. What
6:56you will often see is weird stuff that
6:59doesn't make sense. You'll look at a
7:01breakdown, for example, on placements,
7:03and you'll see that stories are at a 4x
7:06rorowaz feed is at a 3x, but the feed is
7:10getting all of the spend. It might be
7:12holding 80% of all the spend in the
7:14account. And you go, why is this
7:16happening? Seems pretty easy to fix
7:18this, right? we just launch a dedicated
7:20campaign that only places on stories
7:22because stories perform better. But it's
7:24a flawed assumption because you were
7:26looking at blended data that is not
7:28counting in the incremental impact of
7:31pushing more spend through that
7:32particular channel or through that
7:33particular placement. And so if we
7:35simplify this back to the ad example up
7:37here, the reason why this ad is getting
7:40the same amount of spend as this ad is
7:42because Meta has tried to spend more
7:45here. It's tried to push it up past,
7:48let's say, $300 a day to 320. But when
7:51it does that, there's zero incremental
7:54returns. You don't make any more money.
7:57But over here, when this was, let's say,
7:59a lot lower down at $200 a day, Meta
8:02went, "Oh, we can't put spend here.
8:03We're getting no returns." So, let's
8:04see. Even though this has a lower base
8:06return, do we get any incremental
8:08returns here by putting more spent? And
8:10it goes on layers 250 in, and we make
8:13more revenue. Now, it's not at a great
8:15return. Maybe it's at a 2x or something,
8:16but this is at a 0x and so budget goes
8:19here. Same thing applies at a story and
8:21a feed level. So, Meta will try to put
8:24more budget through stories. Obviously,
8:26it's got a better return, but when it
8:28does it, the incremental ROI is so poor
8:32that it would rather just put spend into
8:34the feed. So, the distribution of spend
8:36majority of the time at a breakdown
8:39level is actually accurate and you
8:41should trust Meta. Now, there are cases
8:44where that isn't the case, and that's
8:45where you have to be a good media buyer
8:46and understand what breakdowns matter,
8:47what don't, and how you should be
8:49thinking through distributing spend. But
8:50majority of the time, if you're a
8:52beginner and you get too deep into
8:54breakdowns, you'll just make a bunch of
8:56segmentation decisions that actually
8:58aren't commercially aligned with what
8:59the platform wants you to do, and you'll
9:00get worse performance. So unless you are
9:02like an expert six year, sevenyear in
9:04media buyer, you should not be
9:06implementing changes based on complex
9:09breakdowns or your limited understanding
9:12of return on ad spend and ad. Now it
9:14wouldn't be a Facebook ads long form
9:16video if there wasn't a mention of the
9:17buzzword Andromeda. So here's a 30
9:19second explanation which is that
9:20previously you would choose interest,
9:22you would choose audiences and then you
9:24would load up a bunch of creative and
9:25that creative would serve to that
9:26interest. Now you load in your creative,
9:29you leave everything broad and meta
9:31looks at the creative and based on its
9:34understanding serves it to a relevant
9:36audience. That's effectively the
9:38retrieval system chain. Now you might be
9:39thinking, well does that mean interests
9:41are dead? Does that mean lookalike
9:42audiences are dead? We shouldn't use
9:44these anymore. Generally speaking, yes,
9:46interest targeting is super flawed. And
9:48if you actually go down this rabbit
9:49hole, you'll learn a bunch of reasons as
9:50to why interest targeting is not good.
9:52I'll give you one of them, which is that
9:53interests don't take into consideration
9:56intent. And so if I said I hate dogs on
9:59a Facebook post, Meta would group me
10:02into being interested in dogs. And so if
10:04you go and target the dogs interest, you
10:05would target me. But I actively went out
10:07there and said, "I hate dogs." And I
10:08commented on a bunch of posts. And so it
10:10isn't intent driven. It's simply
10:13interested. And so because of that, Meta
10:15put out a report or someone put out a
10:17report saying that 30% of the people
10:20that are inside of an interest group
10:22aren't meant to be there. They were
10:24incorrectly assigned. And so interest
10:26groups in itself is a poor way of
10:28categorizing users. Now, if you really
10:30want to go a little bit deeper on this
10:31topic, the way that interest targeting
10:34worked conceptually is it was
10:35effectively labeling. And so you would
10:39be interested in a pets post. And once
10:40again, that just means an interaction,
10:42comment, a like, something. Meta would
10:44tag you with pets. And now anyone that
10:46wants to target pets would target you.
10:48Very rudimentary. When you think about
10:50Meta being a trillion dollar business,
10:52you're like, "What? They're just putting
10:53labels on people based on the post that
10:55they interact with. That doesn't seem
10:56very sophisticated and it's not, which
10:58is why it doesn't perform as well as
11:00targeting broad. What happens when you
11:02target broad? Instead, you can think of
11:04it as if it's a vector space. Now, I'm
11:06just drawing three axes here, but in
11:08reality, this is like 10,000 dimensions.
11:11But what happens is that when I go and
11:13interact with a pets post, I get plotted
11:15on the graph. And then depending on what
11:17I interact with in real time, I will get
11:19moved around on this chart in a certain
11:21direction. So if I interact with cats as
11:23an example, I might get moved up in this
11:25direction. But if I interact with dog
11:27posts, I get moved over in this
11:30direction. And then what ends up
11:31happening is people will naturally
11:33cluster through this threedimensional
11:34space. And then when you go and target
11:36broad, instead of targeting people with
11:38labels, you're just targeting this big
11:40space and like sprinkling your ad out
11:42all over the place and going who
11:43interacts with it. And there'll be these
11:45hotspots. And over here, the people in
11:47this area actually start buying from
11:49you. and Meta goes, "Okay, well, let's
11:51just target this area." And then it will
11:53go and target this area right here. And
11:55then as you start to scale, what happens
11:57is the area increases around this spot.
12:00And so you target colder and colder
12:01audiences. And this is why most ads
12:02fatigue. This is why most campaigns die
12:04when you try to scale them up because
12:06you go from a hyperspecific target
12:08demographic and you try to scale out of
12:10it. And these people out here actually
12:13aren't convinced enough to buy your
12:14product cuz your ads might not be good
12:15enough. Now, the caveat with this whole
12:17model is that it isn't
12:18three-dimensional. it's like 10,000
12:21dimensions and so it can be much more
12:23specific in the actual clustering of
12:25users. So the first core thesis to
12:27understand is that consolidation beats
12:30segmentation. Really important concept
12:32to always be thinking about anytime
12:34you're structuring a meta account. Now I
12:36want to run you through the timeline of
12:37why this used to not be the case. Why
12:39it's counterintuitive and why a lot of
12:40agencies particularly legacy agencies
12:43are still hyper segmenting all over the
12:44place and it's probably ruining your
12:46performance. Now, what's also really
12:48unfortunate, as a little side note, is
12:50that the legacy outdated agencies are
12:53typically the very cheap agencies, which
12:55means they're the agencies that work
12:56with smaller businesses. And so, as a
12:58product of that, a lot of smaller
13:00businesses that might be watching this
13:01video that are just starting or maybe
13:03spending $10,000 or $20,000 a month on
13:05Meta, if you're with an agency and
13:07they're cheap, they're probably running
13:09legacy structures. And so, this is
13:11probably relevant for you so that you
13:14can push towards what actually works.
13:16these days. So 2018 was actually the
13:18first time I opened up a Metarat account
13:19and I was spending my own money at the
13:21time. Now back here you could segment by
13:24interest. In fact, it was favorable.
13:26Interesting was actually one of the
13:29biggest levers in the account back here.
13:30Now obviously creative still mattered.
13:32Obviously the website like all of these
13:34things played a part, but there was an
13:36additional lever which is that if you
13:38could take a creative and you could find
13:40the interest that it performed on, you
13:42could achieve scale. and creatives would
13:44not perform on some interests and they
13:46would perform on others. In fact, you
13:48could duplicate adsets with the same
13:49interest, same ads, and sometimes
13:50they'll work, sometimes they won't. And
13:52so, there was all these different
13:53nuances in the account that would
13:56actually allow you to exceed
13:57performance. And the reason being is
13:59that the platform was built around this.
14:01How it used to work is every time you
14:03would launch a new adset, the adset
14:05would go out and it would serve to a
14:07thousand random people and based on the
14:09initial intent signal, so who would
14:11click, who would interact, potentially
14:13who would buy. Meta will then zone in on
14:16those types of users within the interest
14:18group. And so you might have had that
14:20when you launch this adset, for whatever
14:23reason, moms age 40 to 50 interact. And
14:25so it starts going after that audience.
14:28You could launch the exact same ads set
14:30a second time, same ads, same interest,
14:33but on these thousand people, for
14:35whatever reason, men aged 30 to 35 are
14:38the ones that interacted because it's a
14:40very small sample size. So, you can have
14:41bias in these small sample sizes out of
14:43the gates. As a product, this adset
14:46starts going off and optimizing towards
14:47those types of audiences and you end up
14:49with very different results on each and
14:51you end up with very different
14:52audiences, too. And that was just a
14:53product of the way that the ad serving
14:55worked at the time. And so you were
14:58favored to Seg. It was a good thing to
15:02have a ton of adsets where you were
15:03constantly doing testing across
15:05different interests across different
15:06audiences, across lookalike audiences so
15:09that you could try to squeeze out more.
15:11Then what ended up happening? Well, Meta
15:14improved the ad product, right? We're
15:15looking at an 8-year time horizon here.
15:17Some stuff happened. So iOS 14 popped up
15:20which substantially impacted the ability
15:22for Meta to actually do interest groups
15:24because of a lot of the interest
15:26grouping was done using offsite pixel
15:28data from blogs from whatever they were
15:30interacting with on the internet. So
15:32they had to start to change the way that
15:34their retrieval system was working as
15:36well as their ranking system for users.
15:38Then AI started to pop up over here and
15:40that's when Meta became Meta. Went from
15:43Facebook to Meta and Zuckerberg went and
15:45made that big play into the metaverse.
15:47bought a ton of chips off Nvidia. They
15:50ended up not being able to do anything
15:51with those chips. So what did they do?
15:53They rolled them into inference to be
15:55able to better train the actual ad
15:56serving platform. And so then the ad
15:58serving platform started to improve even
16:00further and they started to be able to
16:02squeeze more efficiency. But then as a
16:04product of that consolidation was
16:06preferred during iOS 14.2. Just a side
16:08note, Meta lost they said about 30% of
16:11the signal that they were using for
16:13targeting at the time, which means that
16:15they had to compensate for this by
16:17finding better buyers by not relying on
16:19interest targeting signals and instead
16:21being able to use the creative and go
16:23broad. Now, the fundamental reason why
16:24the platform now prefers consolidation
16:26is because conversion data is siloed at
16:30the campaign level. This doesn't just
16:31apply to Facebook. It applies to Google.
16:33It applies to Tik Tok. It applies to
16:35Pinterest. It applies to all of the ad
16:37platforms. They're all structured in the
16:38same way if you haven't noticed, which
16:40is you have campaigns, you have adsets,
16:42and you have ads. On Google, you have
16:43campaigns, you have ad groups, and then
16:45you have app. This is the same case on
16:46all the platforms. And the reason being
16:48is it's the way that the platform silos
16:51out data in targeting. I won't go too
16:53deep into it, but I think it's important
16:55to understand at least the architecture
16:57of an ad account, whether you're a
16:58beginner or whether you're an expert,
16:59because most people never even think
17:00through this problem. Um, to be able to
17:02better understand structure. So, you
17:04should always be challenging everything.
17:06Why does this exist? Why are they doing
17:08it in this? Why do campaigns and adsets
17:11exist? Why are there adsets? Why aren't
17:13they just campaign? In fact, why isn't
17:15there just one camp? Why isn't the
17:16platform just ads and you just input
17:19your ads and there's no structure up
17:21here? like why does all of this
17:22structure exist? Why have they
17:24introduced this additional complexity?
17:26And it's because you do need a level of
17:28segmentation within the account to
17:30better commercially align with the
17:32objectives of most businesses. The
17:33reason why campaign segmentation exists,
17:36the reason why you can make multiple
17:37campaigns is because businesses often
17:41need to flow budgets through different
17:44campaigns over time. And so you might
17:46have a promotion that rolls in that then
17:48rolls into something else. Since you
17:49need segmentation in the actual KPIing
17:52and reporting number one. Number two is
17:55different business units exist in a lot
17:57of businesses which is that you might
17:59have a category for pets and then you
18:02have a category for children. As a
18:05product of that you don't want the
18:07campaign getting confused and not
18:08understanding who to target when you
18:10have slightly different demographics.
18:12And so as a product was that you want
18:13the ability to be able to segment the
18:15account based on the actual segmentation
18:18of personas within the business and this
18:20allows you to do it. Now one of the
18:22downsides of that is that conversion
18:25data here is not shared at the same
18:28level as it is at the adset and ad
18:29level. And so is there some conversion
18:32data sharing going on here? For sure,
18:34campaign one is pulling from some of the
18:36signals of campaign two and vice versa,
18:38but they're not directly sharing all of
18:41the conversion data with each other,
18:42which means that if you start to
18:44introduce a bunch of campaigns in the
18:46account and you have, let's say, four
18:48campaigns for a small ad account with
18:50not much budget, what you're actually
18:52doing is taking all your conversion
18:53data, let's say you're getting 100
18:55conversions a month, and you're
18:56splitting it up across four campaigns.
18:58So now you have 25 25 25 25 25. You're
19:00inherently going to get worse results in
19:03the account due to segmentation because
19:06now you have less conversion data that
19:08these campaigns can optimize. And
19:10anytime you have less conversion data,
19:12you will see worse results due to small
19:14sample size bite which simply means that
19:16if you don't have a lot of conversions,
19:17meta doesn't have a lot of signal. And
19:19so it doesn't really know who to touch.
19:20The more signal you have, the more
19:21conversions you have generally actually
19:23the better the efficiency is. Which is
19:25why you often actually see that when you
19:27can crack through 100 or 200 conversions
19:30a month, the ad account starts
19:31performing better rather than worse. And
19:33it's because there's enough signal that
19:34Meta now understands who to target
19:36properly. And you can get out of that
19:38rut of Meta not really understanding who
19:41to target because there's not enough
19:42conversion data. So you always generally
19:44want to be consolidating up at the
19:46campaign level cuz every time you
19:48segment, you're chunking out the data.
19:50Same thing kind of applies at the adset
19:52level, not as much. There is more data
19:53sharing that occurs here than at the
19:55campaign level. Number one. Number two
19:57is that all the data at the campaign
19:58level also inherits down. So you don't
20:00really have to worry as much about
20:02segmentation at the adset level. You can
20:03kind of segment all you want. It's not
20:05going to impact performance that
20:06heavily. You can consolidate up. It's
20:08not going to perform impact performance
20:09that heavily. We're talking about really
20:115 to 10% efficiency swings here based on
20:14hyper segmentation versus consolidation.
20:16It's noticeable particularly at scale,
20:18but it's also going to be context
20:19dependent on the actual. Then at the ad
20:21level, obviously you're going to have
20:22massive segmentation because you can't
20:24stack ad. Now you can technically can
20:26stack ads and the name of this always
20:28changes. So by the time you're watching
20:29this, it might be different. The moment
20:31it's called flexible ads, which is where
20:32you can load in like five different
20:33creative on the same ad. The reason why
20:35we generally don't like these is because
20:37you don't get visibility into data
20:39insights. And so you can go and load
20:40five creative up into one ad, which is
20:43all well and good. Nice. It's
20:45consolidated. Meta prefers
20:46consolidation. So it seems like the
20:48right idea. But the issue is when this
20:50ad performs well, we don't know what
20:52creative is actually performing well. Is
20:53it creative 1 2 3 4 5? Which of the
20:56creative that we put in here is actually
20:58lifting performance? Because we want to
20:59do more like that. We want to make more
21:01ads like that. But if we don't have the
21:02ability to read the insight, well, it's
21:04kind of useless. Yeah, we got
21:05performance, but now we don't know what
21:06to do with it. There's also a couple
21:07other benefits of consolidation over
21:10segmentation, and this is mainly letting
21:12the machine decide where to distribute
21:14spend. Now, this can be a bad thing and
21:16it can overweight your account into a
21:18heavy degree of risk, but can be a good
21:20thing if you're just trying to maximize
21:21efficiency in the short term, which is
21:23that when we're talking about
21:24consolidation, Meta is picking where to
21:26distribute your spend. If you just have
21:28everything sitting in a campaign that's
21:30a CVO with a bunch of adsets with a
21:31bunch of ads, Meta is just going to go
21:33and swing budgets around based on what
21:35it believes is most efficient. when you
21:37have segmentation and let's say an
21:39structure. So you're choosing where
21:40budgets go at the adset level. Well, you
21:42are deciding where the budget goes and
21:43inherently meta will always in most
21:46cases distribute budget more efficiently
21:48than you will. However, Meta will also
21:50therefore distribute budget on an 80/20
21:53Purto sprint, which is that 20% of your
21:55ads will get 80% of the spend. Now, kind
21:57of annoying if you're making a 100 new
21:59ads a month and none of them are getting
22:00any spend and they're not getting
22:01tested. Also kind of annoying if all the
22:03spend goes into just one creative and
22:05maybe you're paying influencers for
22:06partnership ads that are getting no
22:08spend or maybe you know that creative is
22:10going to fatigue and now you're like
22:11we're in a risky position because we
22:12have no backup ads that are doing well.
22:14And so this can be good but it can also
22:17hurt you. Whereas on segmentation you're
22:19going to be forcing spend across the
22:20structure in a way that you want which
22:22technically might derisk you might put
22:24you in a better position might
22:25facilitate better testing. But if you're
22:27a novice and you're not doing this
22:29thoughtfully and methodically and
22:30carefully, well then you can just end up
22:32burning a bunch of money here. And so a
22:33good way to kind of think through this
22:34problem is that if let's say you're an
22:36in-house business, $10 million a year
22:38and you're making an in-house media
22:40buying hire. If you're hiring someone
22:42that is in their first year and they
22:43have kind of no idea what they're doing,
22:45I would prefer they run this structure
22:46cuz I would actually prefer Meta
22:48Distributes budgets over them because
22:49they don't know what they're doing. If
22:50you know what you're doing and you're
22:52experienced and you understand weighing
22:54risk against profit and efficiency and
22:56understanding the changes that need to
22:58be made, I would much prefer they run
22:59this structure or somewhere in the
23:01middle so that we can still have control
23:02over where spend's getting forced
23:04carefully and thoughtfully, but we're
23:06not just burning money. There's also the
23:08extreme of this which is there is a
23:10circle on the internet that is focused
23:12around consolidate everything. Just run
23:15one campaign, one adset, throw all the
23:17ads under it. That's all you need. You
23:18don't need an agency. You don't need any
23:20kind of media buying. You don't need any
23:21kind of segmentation. You don't need
23:22structure. Just consolidate. Just upload
23:24ads into a campaign and you're good.
23:26That's a terrible idea for 90% of
23:29people. Now, 10% of brands can get away
23:31with that and they'll be okay. But
23:32there's three reasons as to why you
23:34don't want to just consolidate
23:36everything. Number one is different unit
23:38economics. And so, if you're a large
23:40business with a lot of SKUs, generally
23:42the unit economics or the gross margin
23:44will change across different categories.
23:46you will have 70% gross margin on this
23:48particular product range but on this
23:50particular product range maybe you have
23:5160%. Or you will have grade A, grade B,
23:54grade C, grade D inventory. So on your
23:55grade A inventory you have incredible
23:57unit. On your grade C inventory you also
24:00have incredible unit. Your gross margin
24:02is the same. It's great but it's not
24:03selling. And so because of that you
24:05generally have to discount it hard which
24:07compresses the actual margin post
24:09discount. And so if we put both of those
24:11products or both of those ads into the
24:13one campaign, we would end up with
24:15likely a lot of spend getting
24:16distributed to the heavy sale items. But
24:18the heavy sale items is not where we
24:20want our budget to go because we have
24:21compressed margin there. We're probably
24:23going to end up with worse contribution
24:24margin and it's also just bad branding
24:26to be heavy pushing sales messaging
24:28through our entire funnel and existing
24:29customers. So when you have different
24:31unit economics across the product
24:33portfolio, you actually want
24:34segmentation. Number two is you might
24:36have different audiences. And so an easy
24:38example of this means that you might
24:40have an activewear brand and the
24:42activeware brand sells to both men and
24:44women. Now if you go and put all the men
24:45and all the women ads into one ads set
24:47into one campaign, the adset is likely
24:49going to get confused and in fact you
24:51see this on ad accounts all the time.
24:53This is an easy call out anytime I audit
24:55a brand that sells to both men and women
24:58through core different product ranges.
24:59So there's a men's range and there's a
25:00women's range is that you just go down
25:02under the adset level and you just take
25:04a look at all the ads. You use the
25:06breakdown feature. You look at gender
25:07split and what you'll end up seeing is
25:09that women ads are going to men and men
25:11ads are going to women. Now men ads
25:13serving to women isn't as bad of a idea
25:16because women do a lot of purchasing for
25:18their partners and you end up seeing
25:20directional purchase behavior from women
25:22to men but you don't see it as much in
25:24the other direction dependent on the
25:26category and so you generally don't want
25:27to be pushing women's active wear ads to
25:30men. Will you get a return? Sure. People
25:32will buy. You still get a return on it.
25:33Is it the most efficient use of your
25:35capital in the business? Absolutely not.
25:36You should serve your women's active
25:38wear ads to women. And so if you just go
25:39and put all of these in the one ad set
25:42because targeting primarily exists at
25:44the adset level. That is where you
25:46choose the targeting. That is where the
25:47targeting generally resides for most of
25:49what's occurring under it. It gets
25:51confused. It doesn't know where to serve
25:52them. So it's just serving them to
25:53everyone. Number three is high average
25:55order value or low conversion volume
25:57accounts. We have some clients that we
25:58work with where average order value is
26:01$10,000. on a $10,000 average order
26:04value. The issue is is that you just
26:05don't do much order volume. And so this
26:07brand as an example, I think they do 50k
26:10a day or something like that. It's only
26:12five orders a day. Not a lot of purchase
26:14volume at all. And so what ends up
26:15happening in an account like this is
26:16that if you have a CBO as an example,
26:19and this is a bunch of ad sets. So this
26:21is ads set one, set two, etc. is that
26:24Meta will distribute budget at the adset
26:27level not based on purchases, not based
26:30on row, not based on actual signal that
26:33we care about. Instead, because there's
26:34nowhere near enough signal, there's only
26:36like five conversions max getting
26:38attributed into the account per day.
26:40What ends up happening is Meta needs to
26:42go upstream in its optimization. And so,
26:44it goes up and starts looking at CTRs
26:46and CPCs and hold rates. And now sure
26:48these are pre-intent pre-click signals
26:51as to something that might infer future
26:53performance. But unfortunately the
26:55reality is when you look at very large
26:56data sets CPCs are not correlated with
27:00returns except for the extremity of
27:02bounds. And so if CPCs are super high
27:05yes ROI will be super low. If CPCs are
27:08super high that actually doesn't
27:09correlate with ROI. So what you end up
27:11saying when you graph this out is you
27:12just say something like this, which is
27:14that cost per click has almost no
27:16correlation or a very weak correlation
27:18to actual return on the ad. And so if
27:20we're optimizing budget distribution
27:22based on cost per click or CTR, we're
27:25optimizing it towards a metric that has
27:27nothing really to do with the actual
27:28core objective, which is to drive
27:30revenue. And so in this case, we likely
27:32don't want to use a CBO. We likely don't
27:34want the campaign to distribute budgets
27:36how it is. And we want to think through
27:38a structure that's going to look very
27:40different from what you would expect in
27:42any other account which is sort of
27:44getting us towards the position of every
27:45account looks different depending on the
27:47actual commercial objectives of the
27:48business. What do the unit economics
27:49look like? What do the audience profiles
27:51and personas look like? What does the
27:52average order value and conversion
27:54volume look like? Because all of these
27:55things will change how we approach
27:57consolidation or segmentation. I'll give
27:59you a tactical example of where
28:01consolidation on an audit I did a few
28:03weeks ago was very relevant and not a
28:06good idea which was this was a furniture
28:08brand about 20 to 30 million they had a
28:10cold campaign which was consolidated
28:13that's all they had and then it was a
28:15CBO and they had a bunch of adsets down
28:17here now the issue was 57% of the meta
28:21budget was going into the top adset now
28:23what was this adset it was just a
28:25dynamic product ad now the issue is
28:27return on ad spend looks really good on
28:28On the surface, it was like a seven
28:30rorowaz, but once you break down to
28:317-day click or you break down to
28:33incremental attribution, either one gave
28:34the same rate, returns was actually a
28:361.9. Everything else was performing
28:38better than this on an incremental or a
28:407-day click rate. And so, this was a
28:42good example of where them just
28:44consolidating up and putting DPAs in
28:46with everything else. Not a good idea,
28:47particularly in that industry, cuz
28:48you'll always end up with overspend into
28:50an ad type that doesn't actually perform
28:52that well on cold audiences. But let's
28:54talk about what in media buying actually
28:56died, what doesn't matter anymore, and
28:58then what still matters. Because there's
29:00a lot of stuff that is irrelevant these
29:01days. And when I say media buying still
29:03matters, I'm talking about a very
29:05specific subset of things within the
29:07account. There's a lot of stuff that you
29:08can push to the side. What 99% of people
29:11shouldn't be doing is daily bid tweaks.
29:13Going into the account using cost caps
29:15or bid caps and just changing bids every
29:17single day. Now, that used to be a
29:19strategy 4 or 5 years ago that a lot of
29:21people did quite well with, but these
29:22days, you're really moving the needle on
29:24a variable, on a lever that has nowhere
29:26near as much leverage as other things
29:28that you could be focusing on. Now,
29:29could you technically squeeze 5% more
29:31out of an account using daily bid
29:33adjustments every single day and
29:34dropping an hour a day into this? Sure,
29:36probably. Could you spend the same
29:37amount of time just making better
29:38creative and double the account or
29:40triple the account? Definitely. And so
29:42if we look at return on capital being
29:44time allocation, you're way better off
29:46just making better more creative than
29:48going in there and just tweaking with
29:49bids all day. Now what still matters is
29:51account structure and account
29:52architecture. So thoughtfully thinking
29:53through how we're segmenting and setting
29:55up the account in alignment with the
29:56commercial objectives of the business.
29:57The next one for smaller accounts, this
29:59is applicable on very large accounts,
30:01but for smaller accounts, daily budget
30:02pacing. So changing budgets every single
30:04day in alignment with let's say expected
30:06conversion rates across the week or just
30:08in alignment with daily performance, not
30:10a good idea. You don't want to be going
30:11in and tweaking budgets all the time.
30:13You're going to mess around with
30:14campaigns. You're probably making these
30:15decisions on very limited sample sizes
30:17of data. Once again, was this a strategy
30:19that you would do back in 2019 2020?
30:21Yeah, for sure. Is this something that
30:22matters these days? Not at all. What
30:24still matters? Concept level
30:26segmentation at the adset level. We're
30:28going to go into this in a lot more
30:29detail later on in this video, but this
30:31matters a lot. What doesn't matter on
30:33the other hand is interest targeting.
30:35Very strongly of the belief this
30:36shouldn't be in your account. You're
30:38wasting resources. you're wasting tests
30:40on something that doesn't even really
30:42work anymore. What's a better use of
30:43time is instead of going in and doing
30:45interest testing and interest targeting
30:46is you can do page testing instead. So
30:48split testing different landing pages,
30:50split testing CRO on the website. This
30:52matters a lot. You can make an
30:54inflection in conversion rates. That's a
30:55much better test than going and just
30:57messing with interest targeting on the
30:58platform. I'm also going to throw in
31:00here lookike audiences don't matter
31:02anymore. They don't work. I could once
31:03again talk for five minutes about all
31:05the deficits of lookalike audiences and
31:07why you should just be going broad
31:08instead. What does matter on the other
31:10hand is thinking through and controlling
31:13percentage or dollar allocation to
31:16existing customers. This is something
31:18that you should really be thinking about
31:20and controlling over time. Before I
31:21start to give you the actual account
31:23structure that you should be using at
31:24each level of spend, I want to quickly
31:26talk about page strategies. We as an
31:28agency only work with eight and nine
31:30figure businesses. And so as a product
31:32of that, the content is more tailored to
31:33very large businesses and people that
31:35are in a position to work with us. If
31:36you're small, this is going to be kind
31:37of irrelevant. If you're big, this is
31:39going to be super relevant, which is
31:40that Meta caps the amount of ads that
31:43you can have live on a page at about 3
31:45to 500 ad. And so what this means is
31:48that if you launch a lot of creative
31:50every single week, every single month,
31:51you very quickly pass this cap. If
31:53you're even a somewhat big account, you
31:55then have two options. Option number one
31:57is you just start turning ads off
31:59because you can't have more than 500
32:00active ads in the account, which is
32:02really annoying. Or option two is you
32:03make more pages. And so I'm going to
32:05give you three options here that we
32:07recommend to all clients when we run
32:08into this position. And we run into this
32:10position with pretty much like every
32:11single client because all of our clients
32:12are launching enormous amounts of
32:14volume. Number one is you create
32:15multiple duplicate pages. So someone
32:18that does this quite well, and I've done
32:20a reel on this on Instagram, is Gruns.
32:22They have like 15 Guns pages and the
32:25only difference is that the logo is a
32:27different color and then they're just
32:28running ads to a bunch of these
32:29different pages. Now, personally, if I
32:31was running an e-commerce brand, I would
32:32just do this. I would just spin up a
32:34bunch of pages for the sake of ads. I'm
32:36not really too concerned about
32:37consolidation onto a single page, but I
32:39know some people are. In which case,
32:41your other option is whitelisting or
32:43partnership ads. And so, you just have
32:45to lean more into running ads through
32:47other people's handles that aren't
32:48yours. Now, worth noting, there is a
32:49core difference between whitelisting and
32:51partnership ads. Partnership ads is
32:53where you're using an influencer, a
32:55creator, someone that actually exists
32:57out there, and you're running the ads
32:58through their page. Often, you'll have
33:00to pay them to run the ads through their
33:01page if they have somewhat of a
33:03following and they have leverage.
33:04Whitelisting on the other hand is you
33:06are taking a third-party page that you
33:08can create and you're running ads to it.
33:10So in this case, rather than Grunes
33:12making another page called Grunes,
33:14instead they might make a page called
33:17fiber health magazine or something like
33:19that and then they're running the same
33:21ads. I personally would change them a
33:23little bit to be more advertorial and
33:24orientated around this third party
33:26framing, but you could run the same ads
33:28and instead you're running it through
33:29the handle fiber health. And now the
33:31obvious advantage there is that it looks
33:32like it's coming from a third party. It
33:34looks less like you're being sold to.
33:35It's a really good strategy that we use
33:37across quite a lot of clients and it's a
33:38way that we can decrease the meta ad cap
33:41on the primary page. And then lastly,
33:43you can also do regional pages. Now, if
33:46I was running a brand, I would do all
33:47three of these. I would have
33:48whitelisting. I would have as many
33:49partnership ads as I can. I would be
33:51having multiple different primary pages
33:53and I would be going into regional
33:55pages, which is that when you sell in
33:56multiple countries, you hit the meta ad
33:58cap way faster. The reason being is that
34:01generally you should segment your
34:03campaigns based on country. And so you
34:05will have a UK campaign, you will have
34:07an Australia campaign, you will have a
34:08USA campaign. If each of these has a 100
34:11ads in each, you hit the cap. Even if
34:13they're the same ad, you hit the cap,
34:14which is really annoying. And so the way
34:16that you fix this is you have a Grun's
34:18UK, Gruns Australia, Grun's USA, or
34:20whatever your brand is. In that way, you
34:22have new caps per country. And so you
34:23don't run into this issue. So the third
34:25core claim I made at the start of the
34:27video was that there is no universal
34:29account structure. So, what are the
34:31rules and what are the mental models
34:32that you should use to be able to think
34:34through building your own account
34:36structure? Well, there's two
34:37non-negotiables, which is number one,
34:40data integrity. Now, what this means is
34:42that the structure that you use, more
34:44specifically, the structure must be
34:46readable and actionable. So, whatever
34:49you're running, it needs to produce data
34:51that is readable and actionable. Most
34:54people create account structures, set up
34:56their Meta account, set up their Google
34:57account, Tik Tok account in ways that
34:59aren't readable and therefore aren't
35:01actionable. And so when we try to make
35:03decision loops where we look at the data
35:05and then we make a decision and then we
35:07make time go by and we make another
35:08decision, etc. This isn't a viable
35:11option because the account wasn't set up
35:13in a way that had data integrity. Sounds
35:15really obvious, but you'd be surprised
35:16if you don't think through this as a
35:19core non-negotiable of the structure.
35:20you end up leading yourself into a
35:22position where you regret the way that
35:23the account's structured and you have to
35:24do a restructure. Then the number two
35:26non-negotiable is commercial alignment.
35:29The structure needs to actually fit
35:31towards what the business sees and
35:34thinks about its products margin and
35:36goals. If a campaign ends up hiding a
35:38product or hiding a category that needs
35:41spend or it's going to turn into grade
35:42inventory, that needs to be presented
35:44within the account structure. If there's
35:46a product portfolio with a low margin
35:48base that actually can't support paid
35:50media acquisition, well, that needs to
35:51be reflected within the account
35:53structure. If there's a particular goal
35:55across a category that needs rep
35:57prioritization or KPIing separately,
35:59that also needs to be baked into the
36:01account structure. There's five
36:02questions you should be asking yourself
36:04when you're building out the account
36:05structure. Number one is how many
36:07product categories do we have? The more
36:08product categories, generally the more
36:10segmentation will need to be introduced
36:12if there's different goals across those
36:13categories. Number two is how wide is
36:15the margin spread? Is all the margin
36:17relatively to the same across the
36:19different product categories or does it
36:20vary substantially category to category?
36:22Number three, do we have multiple
36:24personas that don't overlap? Now, it's
36:26fine to have personas that overlap and
36:28are kind of the same person, but we're
36:30tweaking the audience slightly. But if
36:32we're talking about very different
36:33personas that don't overlap, that will
36:35need to be introduced to some degree
36:37into the structure. How many regions are
36:38we selling in? As I said before, when
36:40you're selling in multiple regions, you
36:41generally want campaign segmentation so
36:43that you can control spend distribution
36:45and sell through rates. If you're
36:47holding inventory in particular
36:48countries and they have their own
36:50revenue targets, well then you need the
36:52ability to be able to distribute spend
36:54and the only way to do that is to
36:56segment. And then lastly, what's the
36:57creative throughput? If you're putting
36:5910 ads in the account per month, there's
37:00honestly not much segmentation that can
37:02be done. If you're putting 2,000 ads in
37:04the account per month, it's a very
37:05different volume of creative that's
37:07flowing through. So we need to think
37:08about how is that creative actually
37:09going to be introduced into the
37:11structure methodically have spend
37:13allocation and then give us the ability
37:14to move it into a scaling structure. So
37:16let's go through two actual examples two
37:18different brands very similar in terms
37:20of revenue similar in terms of gross
37:22margin but the account structure out of
37:23the back of it is going to look very
37:25different. So we'll build out what the
37:26account structure will look like for
37:27brand A and brand B. So brand A is in
37:30CPG. So they're selling consumables.
37:32There's only one product. They're doing
37:335 million a year. 70% gross margin which
37:35is fairly indicative of CPG. There's
37:37only one persona so far, which is
37:39actually a good thing. Most people at
37:40this revenue level will try to squeeze a
37:42bunch of different personas out, but
37:43they're only at 5 mil a year, so they
37:44only really need one persona, and that's
37:46Busy Moms. And they sell an AU in US. In
37:49my opinion, they should only be selling
37:50an AU at this revenue level, but they've
37:52decided to expand. Brand B is in
37:54fashion, 50 SKUs, same revenue, slightly
37:57lower gross margin, which is likely
37:58indicative here of them needing to
38:00discount constantly. Their gross margin
38:02shouldn't be this low in fashion, but
38:03it's normally a product of the fact that
38:05they're taking a lot of product to
38:06discount. They've got four personas.
38:08This is fairly typical in fashion,
38:10having three to four core personas, but
38:12they're selling an AU, NZ, and UK. So,
38:15how do these end up getting structured
38:16differently? On brand A, we likely just
38:19want one advantage plus cold campaign.
38:22We want to be doing creative testing at
38:24the adset level based on concepts. For a
38:27majority of spend should be going into
38:28the one concept and the one persona
38:30that's actually performing well at the
38:31moment. There should be one of these for
38:33the US and then we should have another
38:34one for AU. We want existing customers
38:37excluded from both of these cold
38:39campaigns, particularly in CPG where
38:42there's hopefully going to be a lot of
38:43returning customer revenue that's going
38:44to overattribute into cold and the
38:46frequency on these campaigns will end up
38:48being driven up and it will just
38:49retarget. So we want existing customers
38:51excluded. doesn't mean we shouldn't have
38:52any spend towards existing customers.
38:54There's likely a bit of incrementality
38:56here in allocating a little bit of spend
38:58to them. And so we're going to throw in
39:00a retargeting campaign on existing
39:02customers, but it's going to be at a
39:04very low spend to the point that we just
39:06want to keep frequency at below a seven
39:08on a 30-day rolling period. So this will
39:11end up at this size of business being a
39:12very small campaign probably spending
39:14$20 a day. So that's brand A. What about
39:16brand B over here? Due to the higher
39:19complexity of SKUs, there's probably
39:22multiple categories here. In fact,
39:24there's three. And so, we're going to
39:26build out campaigns for each different
39:30category, particularly because they have
39:32different purchase orders, different
39:33revenue goals, and they need to sell
39:35through all of the categories or else
39:37they have to go to sale and a road gross
39:38margin. So, they're going to have a
39:40pants campaign, campaign, dress
39:42campaign. Now, for AU NZ, we're just
39:44going to consolidate this into one
39:46campaign. We're going to target
39:47Australia and New Zealand together. The
39:49nice thing about New Zealand as a
39:50pairing country is that it will
39:52automatically max out at 7 to 10% of
39:54total spend. So we don't have to worry
39:55about it overspending and most New
39:58Zealand orders will always get fulfilled
39:59from an Australian warehouse with
40:01relatively fine shipping craze. And so
40:03we don't have the need. Now this is
40:05already a lot of segmentation for a
40:07brand of this size. And so we also need
40:09to be thinking through do we really want
40:11the UK in a separate campaign because
40:14that's going to mean we have six
40:15campaigns now because we have to
40:16duplicate the whole structure. We really
40:18have two options here. Either option
40:20number one is yes, we do that. But
40:23rather than having six campaigns because
40:25that introduces way too much
40:27segmentation in an account of this size,
40:29instead we move the segmentation down to
40:32the adset level. So we have one
40:34Australia campaign, one UK campaign, but
40:37then at the adset level, we have pants
40:39adsets, tops adsets, dress adsets, and
40:42that's where the segmentation of
40:43categories occurs, and we make sure it's
40:45an so we can control budgets here.
40:47Or option number two is we keep this
40:50structure, but we just consolidate all
40:53the regions into the campaigns. Now, why
40:56or when would you want to do that? It's
40:58if we are shipping the UK orders from an
41:01Australian fulfillment center because in
41:04that case it doesn't matter other than
41:06our targets and our goals as a business
41:08whether the UK revenue goes up, goes
41:11down, goes sideways because we don't
41:13have any holding costs of inventory in
41:16the UK market. All of our stock is held
41:19in Australia. So we just need to sell
41:21through the stock regardless of where
41:24we're actually selling to. Now, there is
41:26a bit of complexity here because it's
41:28fashion, which is that Australia is
41:29southern hem, UK is northern hem. And
41:31so, they're going to be different
41:33seasons. And so, what's selling well in
41:35Australia is not going to sell well in
41:36the UK and vice versa, unless there's
41:38transseal products. And so, in this
41:41case, you are probably making purchase
41:43orders for the UK season, in which case
41:47you do need to move inventory. So,
41:48there's this nuance in understanding
41:50what does this region actually look like
41:51within the business? Are we holding
41:53inventory? Is there risk? do we need to
41:55hit particular sales targets? Because
41:57that information is then going to infer
41:59into the account structure, which is why
42:01understanding the business and the
42:02complexity and the goals is so important
42:04when we're trying to translate into a
42:05structure. I'm likely going to go with
42:08consolidating this down having the
42:10segmentation at an adset level and we
42:13have an Australia cold campaign and then
42:16we have the same thing in the UK. The
42:19reason being as well is that the UK
42:22likely sits on a different subdomain and
42:25so we need all of the UK ads driving to
42:28different URLs than where the Australian
42:30ads are driving. And so because of that,
42:32we need the separate campaigns and
42:34therefore we're going to move the
42:37category segmentation down to the adset
42:40level. Then in fashion, what does do
42:43incredibly well is dynamic product ads,
42:46particularly in retargeting. So, we're
42:48going to throw in a DPA retargeting ad
42:51that's going to go to engaged audiences,
42:53which is just like your 90-day website
42:55visitors. And then we're also going to
42:56throw in a final campaign, which is your
42:59retargeting on existing customers, which
43:02is going to contain two adsets. We're
43:04going to have new arrivals in there. So,
43:05we're constantly putting new arrivals in
43:07front of the existing customer base. And
43:09we're also going to have a DPA. We're
43:10going to have it as two separate adsets
43:12because if we consolidate all the spend
43:13will just go to the DPA and we'll never
43:15be able to actually serve new arrivals
43:16to existing customers. These campaigns
43:18are obviously because they're cold going
43:20to have existing customers excluded
43:22which is why we need this. Um targeting
43:25existing customers in fashion is also
43:26much more incremental than any other
43:28category that we've seen. So it actually
43:29is worth the spend allocation which is
43:32why you have on the surface two brands
43:35at the same revenue level. But when it
43:38then comes to understanding the business
43:40and all the complexities, the gross
43:42margin profile, the personas, the
43:44countries, the skew count, it then
43:46translates into a very different account
43:48structure. All right. So, if you're
43:50spending sub 50K a month, I'll give you
43:51a generic structure that most people can
43:53get away with. Will it be ideal? Will it
43:55be custom to the business? No. I've gone
43:57through all of the reasons as to why you
43:59should ask yourself those five
44:00diagnostic questions to be able to
44:02better understand how you should
44:03structure yourself. But if you just want
44:05something generic, if you need to throw
44:06something up, here's what you should do.
44:08Number one is a dedicated testing
44:10campaign. Now, at this spend level, this
44:12is probably all that you need. Now, what
44:15this actually looks like is either an
44:17or a CBO. And we'll go through later
44:19which one you should actually choose
44:21with adets sitting under it. Three to
44:23five creatives per adset. Now, you can
44:25go way more than this if you have the
44:27creative volume to support it. Just most
44:29people at this spend level probably
44:30don't. And the creatives are being
44:32segmented based on concept. Now concept
44:35for people that haven't watched any
44:36other video of ours is the intersection
44:39right here in the middle of an angle, an
44:42offer, and a persona. And so you're
44:43putting a persona, angle, offer
44:45together, and then you're making
44:46creatives under that. And that sits in
44:48an adset. When you then go and make more
44:50creatives for the concept, you have an
44:52option. You can either launch it in the
44:53existing adset or you can launch it in a
44:55new one. In terms of which you do there,
44:57if the adset is not performing, throw it
44:59in. Try to get the adset to perform.
45:01Throw more creators in. If it is
45:02performing, if it's hitting KPI, if it's
45:04doing well, don't touch it. Number one
45:06rule of meta media buying is never touch
45:08something that's working. If it's
45:09working, don't touch it. Do something
45:10next to it. Don't ruin the thing that's
45:12working. So, if it's working, don't
45:13touch it. Launch the concept as another
45:16adset. So, you might end up in a
45:17position where you have 10 adsets and
45:19six of them are for just one concept and
45:21it's a bunch of new creative that you've
45:23made over time and everything's working
45:24because that ends up being the case.
45:25This one concept ends up outshining
45:27everything else. Now, it's not that
45:28important at this scale of business, but
45:30you do want existing customers excluded.
45:32We don't want our testing getting skewed
45:34around based on existing customers
45:36coming through a few ads and making
45:38something look better than it actually
45:39is. You also want to make sure that your
45:40attribution setting is on 7-day click so
45:43that you have data integrity in the
45:45numbers that you're reading or else you
45:46also end up in a position where some
45:47things might look better than others
45:49just because it's claiming conversions
45:50that has nothing to do with it. Then you
45:52can layer on a scaling campaign. Now the
45:55idea here, the premise of having a
45:57scaling campaign is that one of these
46:00adsets performs well. So let's say this
46:02adset at the top is performing well
46:04above your target return of 5x and all
46:06the other soft metrics are really good
46:08too. You start scaling this up. You put
46:10more spend more spend over time and
46:13you're ramping it up and then eventually
46:14return on ad spend drops to a point that
46:16is unprofitable and so you need to pull
46:19back spend. Now, when you do that, you
46:20effectively find an equilibrium where
46:23you find the daily spend level where
46:25that creative can keep sitting there,
46:26continue turnurning over results that
46:28you're happy with. You can't push it any
46:29further. Anytime you do, results drop.
46:31So, you have to come back. And because
46:32of that, you're kind of not happy with
46:33it. You're like, "Ah, we could only take
46:34this creative or this set of creatives
46:36to $300 a day." But obviously, I want to
46:39spend way more and I want to scale. So,
46:41what do you do? It's at this point where
46:43people often turn and go, "Well, we
46:45can't scale this adset anymore. It's not
46:47working. So, how can we take these
46:49creatives and just launch them elsewhere
46:51within the account structure to be able
46:53to get more spend through them? And
46:54that's effectively where scaling
46:55campaigns are born, which is you take
46:57creatives that have been maxed out, you
46:58can't get any more spend through them,
47:00and you go and just dump them in a new
47:01campaign. And you hope that the new
47:02campaign can get even more spend through
47:05that ad. And a lot of the case, it
47:07can't. For whatever reason, you take the
47:08creative, you launch it in a new
47:10campaign, you keep this on, keep this
47:11spending, but you go and launch it
47:13elsewhere and you can likely get a
47:15little bit more spend through that ad.
47:16And so rather than the ad within the
47:18account holding $300 a day, maybe you go
47:21at this point and you layer it in on a
47:23scaling campaign and a scaling campaign
47:25can get it to hit. And so total spend
47:27across the entire account on this asset
47:29is now $400 a day because it's being
47:31propped up by the scaling campaign.
47:33That's the idea of why you've launched a
47:34scaling campaign. There's absolutely no
47:36reason to have one if you don't have ads
47:38that are working well. Like there's no
47:39point in adding this complexity unless
47:41you have creatives that are doing well,
47:43that have been cranked to their maximum
47:45potential, and then you're just trying
47:46to squeeze even more out of it. A big
47:48mistake I see is that people will have
47:50something that works and they get it to
47:52like $60 a day. So like a meaningless
47:55amount of ad spend and then they'll go,
47:56"Ah, we can't push it any further. Let's
47:58roll it into scaling." This is nowhere
48:00near enough spend to be able to start
48:02trying to scale the ad even further.
48:03Good way to conceptualize this is that
48:05this media buying move of taking a
48:08winning ad and then trying to get more
48:10spend into it using an adjacent campaign
48:12will get you on a good day an extra 20%
48:16daily spend through the app. So if you
48:18have an ad that's spending $60 a day,
48:20this kind of move in the account is
48:22going to get you like an extra $10 in
48:24daily spend. Not worth it. Just make
48:26better ad. Fix other things in the
48:27business. There's probably a landing
48:28page issue. There's probably an offer
48:30issue. You still probably haven't found
48:31a concept that's working. There's like a
48:33million other things to put your effort
48:35towards rather than trying to media buy
48:37your way to an extra 10% of spend on a
48:39really low base. If you're ad spending
48:41$400 a day actually worth just taking it
48:43and putting it in a scaling. It's a
48:45relatively loweffort move. It doesn't
48:46take much time and yeah, you're going to
48:48unlock an extra $80 per day in daily
48:52spend. If you do this across like 10 ads
48:54all at once, you take 10 of your highest
48:55performing ads and you dump them in a
48:57scaling campaign. It might allow you to
48:58unlock an additional $1,000 day in
49:00spend. And that actually is worth it.
49:01That's a decent unlock. it's worth
49:03having this increased degree of
49:05segmentation and complexity in the
49:07account. And then lastly, you have the
49:08retargeting campaign. This is absolutely
49:10an optional campaign. When we talk about
49:12retargeting, there's two different types
49:14of retargeting. There's website visitor
49:16retargeting. So, effectively warm
49:18audiences, people that have shown
49:20intent, they've taken some kind of
49:21action, but they haven't purchased. And
49:23then we have existing customer
49:25retargeting. So, this is someone that's
49:27actually purchased from us. They're all
49:28the way down at most aware in the stages
49:31of awareness cuz they're a customer and
49:33we're trying to effectively get them to
49:34buy a second time or a third time. Now,
49:36for most accounts, 90% of them, you
49:39don't need a website visitor retargeting
49:42campaign. And the reason being is that
49:44the testing or the scaling or whatever
49:46currently exists will allocate a good
49:49amount of spend anyway towards
49:51retargeting website visitors. And you
49:53see this by doing an audience segment
49:55breakdown within the account. So, if you
49:57hit breakdown, you hit audience
49:58segments, you'll be able to see as long
50:00as this is set up within your advertiser
50:02settings in your audience segments. So,
50:04make sure that's set up. You'll be able
50:05to see spend, rorowaz, etc. on new
50:08audiences as well as engaged and then
50:11existing. Now, you shouldn't have any
50:13spend to existing cuz they should be
50:14excluded, but you'll be able to see your
50:16spend towards engaged, which is your
50:17website visitors. And there'll likely be
50:19enough spend here at a high enough
50:21frequency that you don't need dedicated
50:23campaign to force more spend through
50:25that elites. It's okay on existing
50:26purchases cuz you have it excluded here.
50:29You probably want some spend towards
50:31existing customers. It just depends on
50:33the retention dynamics in the business,
50:35the category, and the stage of business
50:37that you're at. If you're a new
50:38business, which at sub 50k a month, you
50:40probably are. You're also probably a
50:41small business, you don't have that many
50:43existing customers. And then depending
50:44on the product portfolio and the
50:46retention portfolio, so is this
50:47furniture where you're like going to
50:49have no repeat purchases really. It's
50:51very low. It's very infrequent. or is
50:52this supplements low average order value
50:54which has super high repeat rates in its
50:56subscription model. Okay, well different
50:58story. We probably want some spend
50:59there. So you need to contextualize this
51:00to the actual business model and the
51:02expectations on repeat purchasing. If
51:04there's an expectation of high repeat
51:05purchasing, put some existing customer
51:07spend in there. Keep it low. Keep it
51:08controlled. Look at frequency as the
51:10measure. If you don't expect repeat
51:12purchases, don't have. Now the reason
51:13why this structure works at this spend
51:15level is number one, it's consolidated.
51:18You probably have not a lot of
51:19conversion volume at this spend level.
51:21And so a product of that means that you
51:23shouldn't have a lot of segmentation.
51:24You want all the conversions
51:25consolidated up. You need structured
51:27creative testing at this spend level
51:29because you probably don't have anything
51:30that's a very large winner or else you'd
51:33be spending more. And so you want an
51:35structure where you can force spend
51:37through adsets and start to learn stuff.
51:39And then number three, at this spend
51:40level as well, you probably don't have a
51:42large product portfolio. You probably
51:44don't have too much complexity in the
51:46business model as it stands. And so we
51:48can keep things relatively simple in
51:50terms of KPIing this structure on the
51:52testing campaign. You want to KPI at the
51:55adset level. Did the adset hit our
51:57target cost per acquisition or our
51:59target return on ad spend over a roll in
52:01window which is reasonable. And what
52:04window in which you read data is
52:05dependent on conversion volume. So you
52:07can't look at three-day windows if
52:08you're only doing three conversions in a
52:103-day window. You can look at two-day
52:12windows if you're doing a thousand
52:14conversions a day. So, it's all based on
52:15contextualizing uh the windows that
52:17you're looking at for per for
52:18performance based on how much volume the
52:19business is actually doing. At this kind
52:21of scale, I wouldn't really be reading
52:22data on any shorter than a 5day period.
52:25And then if the adset is hitting KPI,
52:27budgets go up in 20% increments per day.
52:30If it's not hitting KPI, we diagnose
52:31why. What do we think was wrong in that
52:33creative? Let's then go and make more
52:34ads, more concepts, and let's either
52:36wind that adset down or turn it off
52:38completely. On the scaling campaign, you
52:40want to just be KPIing all the way up at
52:42the campaign level. there isn't any
52:43adset segmentation on a scaling
52:45campaign, at least at this level. And
52:47then on retargeting, you want to be
52:49KPIing based on incremental return on ad
52:51spend by doing an attribution breakdown.
52:52And also frequency shouldn't jump to
52:54above a seven on a 30-day window. The
52:57four common mistakes at this level of
52:59spend is that people test inside the
53:01scaling campaign. There's no exclusions
53:03on the scaling campaign. They kill
53:05testing way too fast. So the window in
53:07which they're looking at conversion data
53:08is way too little. And then they set
53:10budgets on the adset level based on the
53:1250 conversions in a 7-day window rule,
53:15which ends up with just way too much
53:17budget for their particular business. So
53:18you actually need to back propagate
53:20budgets at the adset level based on your
53:22average order value and expected CPA. So
53:24let me quickly break that down for you
53:25so you understand what kind of budgets
53:27you do need to be setting on the adset
53:29level. So when you're looking through
53:30adset level spending, there's really two
53:32factors that you need to think about.
53:33You need to think about time to outcome
53:36of the test and then you need to think
53:38about total budget required for the
53:41test. So what does that actually mean?
53:43Well, let's say that you have a target
53:46cost per acquisition. So you want to be
53:48getting customers at $50 per customer.
53:50You then come up with a new concept. You
53:52make 10 ads under it. We then go and
53:54launch it at the adset level. What we
53:56really want to figure out as quickly as
53:57possible is is that concept working or
53:59not? Now, we need to set a barrier or a
54:02threshold for how much spend are we
54:05going to put through those ads, through
54:06that concept before we decide yes or no.
54:08Now, the general rule of thumb that
54:10almost everyone uses is you take your
54:12target cost per acquisition. So, this is
54:14our target. You times it by three and
54:16that is how much you should spend before
54:18you decide whether to keep the ad on or
54:20off. If you get three conversions in
54:22this time, if you get four, even better.
54:23Keep running it. If you get two, uh,
54:25yellow light, let it run a little bit
54:27longer. Let it run another $50 and then
54:29we'll decide. If you get zero or one,
54:31cut it. Now, that as a rule of thumb is
54:33pretty decent. Now, I would always let
54:35it spend a little bit more. I'm always
54:36willing to be a little bit more gracious
54:38cuz I know that the initial spend is
54:40really trying to learn and figure out
54:42who to target. And then once it figures
54:44it out, the ball gets rolling and you
54:45actually see your efficiency client. And
54:47so, I actually prefer this rule if
54:49possible. And if the client's okay with
54:50it, they're more of a times five. So, we
54:52wait till we spend 250 and then we make
54:54a concrete decision. There's no yellow
54:56lighting. That's either a yes or a no
54:58because we have enough spend volume
54:59hand. Okay, cool. So, let's say that
55:00$250 is what we want to spend. So, we
55:02know the total budget of the test. We're
55:05going to spend $250 on these new
55:06creatives before we decide whether
55:08they're a success or not and whether we
55:09turn them off. Then, we have time. This
55:11is how quickly do we want the test
55:13outcome? Because technically, we could
55:15just set $250 a day as the budget and
55:19we'll know whether this adset and this
55:20group of creatives has worked today.
55:22We'll know by the end of the day. Now
55:23the issue with that is that there are a
55:26couple components that will play into
55:28the time duration that you need to allow
55:30a test to run for. Number one is daily
55:33seasonality. The reality is is that
55:35there is going to be some days of the
55:36week in which you do better and some
55:38days of the week in which you do worse
55:39as a business. If you go and launch a
55:41test on a good day, you will see better
55:43results. If you go and launch it on a
55:45bad day, you will see worse results
55:46irrespective of the actual test variable
55:49which is the creative. And so if we run
55:51tests too quickly, we don't get the
55:53ability to encapture the whole
55:55seasonality of the week and therefore we
55:57can get a biased view of performance.
55:59The second thing that impacts time is
56:02time to purchase. And so for a lot of
56:04brands, people don't instantaneously
56:06purchase the first time they ever see an
56:08ad from. Okay? You don't just serve an
56:09ad to someone and they go, "Oh, great.
56:10Buy." Normally it takes two impressions,
56:13three impressions, four impressions,
56:15maybe a couple clicks, and then over the
56:17course of a 3 to 4 day period of warming
56:19that user up, then they make the
56:20purchasing decision. Now, if we run this
56:22test in one day, well, guess what? We're
56:23going to show the ads to a bunch of
56:25people. They might go, "That's amazing.
56:26These ads are great. I'm now
56:28interested." But then you kill the ads,
56:29you turn the test off, other ads go and
56:32target those users in retargeting. They
56:34all end up converting through other
56:36creative, but you killed the primary top
56:38ofunnel ad that you want to test it
56:39because you only gave it one day. And so
56:41you need to not only give it a little
56:43bit of time because of daily seasonality
56:45that might skew the results, but also
56:47because people take a little bit of time
56:48to buy. True. And this applies not only
56:50to Facebook, but Google and all the
56:51other platforms as well. So because of
56:52that, we need to balance these two
56:54variables. We need to figure out what is
56:55the budget for the test. And then how
56:57much time do we want as a feedback loop?
57:00Ideally, as fast as possible. We want to
57:02know the outcome quickly so we can
57:03continue to iterate because ultimately
57:05the speed of growth in a business is a
57:07function of the speed of cycles of
57:10feedback and learning and iteration. And
57:12so we want to iterate and learn as fast
57:13as possible so that we can grow. But we
57:15want to give it enough time that it
57:16encapsulates daily seasonality and time
57:18to purchase. What does that often end up
57:19coming out to? One week. One week is
57:22generally a good duration to run a test
57:23for. Now if you're a big business with
57:25more stability with not a long time to
57:27purchase and cut that down to three to
57:28four days. If you're a tiny business
57:30with tiny budgets with a incredibly long
57:32time to purchase and massive daily
57:34seasonality, well, okay, maybe we need
57:36to extend that to 10 to 14 days. Okay,
57:38so once again, always going to be
57:40respective to your particular business,
57:42which is why in this whole video and all
57:43the content that we put out, we talk in
57:45frameworks rather than actuals. Rather
57:48than saying this is exactly what you
57:50should do, we say here's how you should
57:51think through the problem. Here's the
57:53two variables that impact it. Now think
57:55through the problem yourself in your own
57:57context and then come up with your own
57:58solution because every solution, every
58:00budget, every adset is different for
58:03every single business. Now as you jump
58:04to 50 to 250k a month in ad spend, if I
58:07was to prescribe you with a campaign
58:09structure, which once again we don't
58:10like doing cuz different based on all
58:12the commercial objectives of the
58:13business, but if I was to, you still
58:14have the retargeting campaign. This is
58:16likely still going to be just existing
58:17customers. Very rare you need to layer
58:19in website visitors here. You might,
58:20probably unlikely. And then in terms of
58:22the cold campaigns, this just builds out
58:24a little bit further. So you're going to
58:25end up with probably more adets at the
58:27adset level because you're going to be
58:28doing more creative testing and likely
58:30have more creative volume which is going
58:31to introduce more adset volume and then
58:33potentially you're also going to extend
58:35out to two maybe three campaigns. Now if
58:36we're looking at this as if it's a CPG
58:39brand with one skew, this campaign
58:41segmentation is actually going to come
58:42from different funnels and different
58:44approaches. And so you might have a
58:46funnel or a persona or an approach
58:48depending on the vernacular that you
58:49want to use to explain it. But you might
58:51have a funnel that is targeted at old 65
58:55plus year olds for arthritis, right? And
58:58this is a supplement that we can pivot
59:00for that. In fact, let's just call it
59:01fish oil, right? Fish oil applies for
59:03old arthritis. All the ads as well are
59:05going to be mirroring this type of style
59:06that works with this age demographic,
59:08which is going to be VSSLs. It's going
59:09to be native statics. It's going to be
59:11testimonials from an older demographic.
59:13and then actually probably throw on like
59:15TV style ads as well that you're
59:17chopping into UGC format. Then you might
59:20have the product repositioned. All of
59:22the ads are now for young people 25 to
59:2435 focused on like brain health or
59:28performance at work or something
59:30oriented around that. Now you could also
59:31get this funnel working. The landing
59:32pages are going to look very different
59:34because you want different type of
59:35demographics and photos on the landing
59:36pages. Um the ads are going to look very
59:38different. Probably the profiles that
59:40you even run this through might look
59:42very different as well. This might be
59:43heavier on whitelisting. This might be
59:45on the native page. But because this is
59:47almost two different businesses in
59:49itself because the funnels look very
59:50different as a product. This is two
59:52separate campaigns. So that's what that
59:53would look like in CPG with one skew. If
59:55we're talking about like fashion with a
59:57bunch of SKs, this just ends up being
59:59some kind of category segmentation. It
1:00:00might be new arrivals segmentation. So
1:00:03you have your normal existing cold
1:00:05campaign that you had at lower spend,
1:00:06but now you just go and bolt on a new
1:00:08arrivals campaign um for the sake of
1:00:10pushing new arrivals and having higher
1:00:12sellrough rates here. This might be a
1:00:14particular product category when we do
1:00:15an LTV analysis, we find out that this
1:00:18particular product category, let's call
1:00:20it pants, ends up with not only a higher
1:00:22average order value on first purchase,
1:00:23but much better repeat rates and
1:00:25retention. And so as a product of that,
1:00:27we make a second campaign just for pants
1:00:30where we want to have a higher cost per
1:00:32acquisition KPI. So we can be more
1:00:34aggressive in acquiring customers
1:00:36through here because we know that our
1:00:38pants category acquires really high
1:00:40quality customers. And so as a product
1:00:42of that, we get a second campaign. So
1:00:43generally when you're going from the
1:00:45sub50k range and then bridging into 50
1:00:47to 250k. What ends up happening if
1:00:50you're a good media buyer or a good
1:00:51performance marketer is additional
1:00:53campaigns will start getting layered in
1:00:55that's has a very specific reason or
1:00:58commercial objective that aligns with
1:01:00the business's fundamentals. If you're
1:01:01just layering in more campaigns for like
1:01:03the seo you're like ah we're spending
1:01:05100k a month right now. One campaign
1:01:06seems weird. Let's do two. Let's do
1:01:08three. you're breaking the fundamentals
1:01:10that I went through before, which is
1:01:11that every segmentation decision needs
1:01:13to be done with the ability to read data
1:01:16and make decision loops through an
1:01:18underlying structure that has data
1:01:20integrity. So, if you're just
1:01:21introducing segmentation for the sake of
1:01:22it, you're breaking all the rules that
1:01:23we've gone to in this case. You want to
1:01:25look at segmentation, but only if it
1:01:26makes sense to better align with the
1:01:28business's objectives. The main key at
1:01:30this campaign level and where the real
1:01:31skill unlock is and where some people
1:01:33just smash straight through this spend
1:01:34level and some people stay here for a
1:01:36very long time is in continuous testing
1:01:38of different concepts so that you can
1:01:41find the 20% of concepts that are going
1:01:43to do really well and are going to be
1:01:44able to hold hundreds of thousands of
1:01:46dollars a month in ad spend. Now we have
1:01:48a YouTube video called Meta Ads creative
1:01:51strategy in 2026 the full system where
1:01:53we spend literally 30 minutes talking
1:01:56about how to create concepts. So, I
1:01:57strongly recommend you go and watch that
1:01:59video. But if you're unfamiliar with the
1:02:01concept, one minute run through. You
1:02:03want to be building out personas,
1:02:04angles, offers, and then ad types,
1:02:07building this matrix out, and then
1:02:08having your ad sets reflect this
1:02:10structure, and then continue to test ads
1:02:13under each concept. An example of this,
1:02:14if you're selling teeth whitening, is
1:02:16that a persona could be coffee drinkers.
1:02:18The angle is whitening without
1:02:20sensitivity issues. The offer is just
1:02:22the product. There's nothing special
1:02:23there. And then the ad type is
1:02:25userenerated content. And obviously
1:02:26under this ad set, if it's coffee
1:02:28drinkers, whitening without sensitivity,
1:02:30we can rotate in tons of different ad
1:02:32types. Like the ad type is flexible.
1:02:33This is infinite. The next angle is
1:02:35brides. And you go, this is very
1:02:36different, right? Coffee drinkers into
1:02:38brides. How do brides relate to teeth
1:02:40whitening? Well, we can frame this
1:02:41around get ready for your wedding in 14
1:02:43days. We can do a bundle offer that's
1:02:45specifically curated for fast 14-day
1:02:48turnaround to get white teeth for the
1:02:50wedding. And then we can do this to a
1:02:51testimonial of an actual person that
1:02:53just had a wedding in the dress. They
1:02:54can do a before and after. they can
1:02:55document the whole process. And the real
1:02:57key here and where this really
1:02:58accelerates is that once you get
1:02:59something that's working, let's say this
1:03:01bride idea does well and this ad does
1:03:03really well. Well, you double down. You
1:03:05start making a ton of creative around
1:03:06this concept and you have a custom
1:03:08landing page that has continuity through
1:03:10it. So, we might actually spin this
1:03:12bundle off and call it like the wedding
1:03:14in 14-day bundle. have an own custom
1:03:16landing page, have all these
1:03:18testimonials on the landing page, build
1:03:19the entire funnel around this angle, and
1:03:21then this can scale to 2, 3x the volume,
1:03:24and we can really saturate this market
1:03:26while at the same time continuing to
1:03:27work on other angles and other persona.
1:03:29And the last one I have here is an
1:03:30obvious one, which is smokers to remove
1:03:32stains. You give them a subscription
1:03:33offer because they're continuing to
1:03:35smoke, so they'll have this issue
1:03:36forever. And so, you want to put them on
1:03:37an offer that's relevant to that. And
1:03:39then this is through founder head
1:03:40talking content because maybe the
1:03:41founder is a smoker and that's why they
1:03:43started the business. So, the reason why
1:03:44concept level testing becomes so
1:03:46important at this tier of spend is
1:03:48because each campaign should be
1:03:49generating enough spend volume, enough
1:03:51conversion volume to be able to support
1:03:53a degree of segmentation of the adset
1:03:55level. It allows you to then start to
1:03:56KPI based on concept rather than just
1:03:59the campaign as a whole, which allows
1:04:01for better directional feedback in
1:04:03creative. And this is really a core key
1:04:05here, which is that this creates a
1:04:07creative strategy because you have the
1:04:09strategy, you come up with all the
1:04:11creative, but then most people just
1:04:12throw it in the account and then they
1:04:14just make more creative and they throw
1:04:16it in the account and they make more
1:04:17creative. But they don't actually look
1:04:18at the account and go, "Well, how is
1:04:20stuff performing? How are we reading
1:04:21this data? How are we drawing insights?"
1:04:23And then how is that informing the next
1:04:25batch of creative that we're making? And
1:04:27so the creative refreshes become
1:04:29targeted. The losing concepts get more
1:04:32attention and the winning concepts get
1:04:34more volume. You can also then begin
1:04:37directing resources upstream to landing
1:04:40pages, retention flows, product
1:04:42portfolio expansion based on concept
1:04:46level testing. As an example, let's say
1:04:48that the bride's persona starts
1:04:49performing incredibly well. What do we
1:04:51do? We increase creative volume. We
1:04:53create a custom landing page. We change
1:04:55the offer accordingly. and potentially
1:04:57as a byproduct of changing the offer, we
1:05:00expand the product categories to meet
1:05:02this audience as we're generating so
1:05:04many existing customers through this
1:05:05funnel. And so we might have a retention
1:05:07win back flow off the back of this that
1:05:09is post wedding. We send them a free
1:05:12gift saying congrats on getting married.
1:05:14Here is an additional offer for you that
1:05:16pulls them into a different product line
1:05:18that meets them at the stage of life of
1:05:20where they are. So you can start to get
1:05:22really creative in terms of the backend
1:05:24retention, flow, sequencing, and product
1:05:25portfolio based on direct feedback of
1:05:28what type of customers are we acquiring
1:05:30and from where. If you end up keeping
1:05:31your concept targeting super broad and
1:05:34you don't niche down in this way, you
1:05:36don't get accurate persona insights and
1:05:38that then bleeds into inefficiencies in
1:05:40the rest of the business. So this is
1:05:42really also telling you who your
1:05:43customer is. Who is the customer? Who
1:05:44are we acquiring? And in what percentage
1:05:46allocations are they? Are we getting 40%
1:05:48of our customers because they're
1:05:49smokers? 40% because they're about to
1:05:51get married, etc., etc., and then we can
1:05:53start to craft the entire business
1:05:54strategy around this. The last reason
1:05:56why concept testing becomes critical at
1:05:58this scale is it supports a large volume
1:06:00of ads. So, at this spend level, you
1:06:02should be launching really between 100
1:06:05to 300 new ads per month. If you want to
1:06:08grow, if you don't want to grow, don't
1:06:09launch that many ads. But if you do want
1:06:10to grow, you should be launching around
1:06:12about this volume. This volume becomes
1:06:15actually quite sustainable for entering
1:06:17into an account structure if you have
1:06:19built out to let's say two campaigns
1:06:21with 10 adsets within each campaign
1:06:23because that's 20 adets which means on
1:06:25the low end you've got five ads per
1:06:27adset on the high end you've got 15 ads
1:06:30per adset very reasonable in fact you
1:06:32could actually have less campaigns less
1:06:33adsets and you're fine here the ad to
1:06:35adset ratio is completely manageable so
1:06:38what then happens at 250k plus pretty
1:06:41much everything at 50 to 250 just more
1:06:44of it. And so when we come to campaign
1:06:45segmentation at this budget, and this is
1:06:47a enormous budget range here cuz I'm
1:06:49saying $250 all the way up to $20
1:06:51million a month in spend. Okay? So it's
1:06:53an enormous range and therefore there's
1:06:55an enormous degree of complexity
1:06:56difference. And so because of that,
1:06:58unlike the other two sections, I'm not
1:06:59going to give you an exact account
1:07:01structure that you should run because it
1:07:02would just be ridiculous. It won't be
1:07:04applicable to anyone at this spend
1:07:05level. And instead, I'm going to give
1:07:06you a bunch of strategies, a bunch of
1:07:08advice, a bunch of tips, and I'm going
1:07:09to run through an example so that you
1:07:11can get as much value as possible. and
1:07:13how you should be structuring at this
1:07:14spend level. Number one is you do want
1:07:16to potentially consider layering in a
1:07:19second ad account. Now, this strategy
1:07:21changes a lot and what I would have said
1:07:236 months ago is different from what I'll
1:07:24say today and what I say today will be
1:07:26different in 6 months. And so, I don't
1:07:28want to give you too much tactical
1:07:29application here cuz it'll just be
1:07:30outdated. But the idea here is that you
1:07:32run the same pixel, but you run
1:07:35different bid logic. So, if you're
1:07:36running maximize conversions in the main
1:07:38account, you might run maximize
1:07:39conversion value or you might run cost
1:07:41caps or big caps in this account on the
1:07:43same page. It could be different page as
1:07:44well. And the idea is that you'll start
1:07:46winning different auctions from this ad
1:07:47account that the primary account just
1:07:49isn't bidding on. And so, it will allow
1:07:51you to get more volume through
1:07:53effectively the same creative and the
1:07:55same pixel and the same page, but you're
1:07:57entering different auctions due to it
1:07:58being in a different ad account. The
1:08:00disadvantage that you could argue is it
1:08:01might increase CPMs and ad costs for you
1:08:03because you might be cross bidding
1:08:04against yourself. Now, if you're using
1:08:06the same page and the same pixel, you
1:08:07shouldn't really cross bid that much.
1:08:09But it's definitely an argument and no
1:08:10one has really been able to prove
1:08:11whether this is the case or not. The
1:08:13main other obvious advantage here is
1:08:15that it derisks the business enormously
1:08:18because if you have a second ad account
1:08:19and the primary ad account gets banned
1:08:21or the billing method goes down for
1:08:23whatever reason, you don't just lose all
1:08:24your new customer acquisition in the
1:08:26business if you're over reliant on meta
1:08:28ads. Instead, you have two ad accounts.
1:08:30One might only hold 10% of the spend of
1:08:31the other, but if the main one goes
1:08:33down, you can just crank up the second
1:08:34one, and it prevents a doomsday scenario
1:08:37where you might not have any revenue for
1:08:39a week. Number two is that at this spend
1:08:41level, you absolutely need a page
1:08:44strategy. As I said earlier, you're
1:08:46going to hit your ad limit without a
1:08:47doubt. This is a guarantee. And so, what
1:08:50is the strategy to be able to avoid that
1:08:52from happening? And I gave you the three
1:08:53options earlier in the video. Number
1:08:55three is that you will generally start
1:08:57to get complexity getting introduced in
1:09:00more media buying tactics at this level
1:09:02of spend. And this is the level of spend
1:09:04where it actually does start to make
1:09:05sense to be playing around with the 3
1:09:084enters because 3 4% when you're
1:09:10spending a million dollars a month is
1:09:11actually quite mature and it could add
1:09:13an enormous amount to bottom line
1:09:14profit. And so having someone dedicated
1:09:16on trying to media buy your way to more
1:09:18efficiency is genuinely worth the
1:09:20investment. This is where like bidding
1:09:22complexity will start to be introduced.
1:09:23So you're not just running maximize for
1:09:25conversions in the account across
1:09:27everything, but you might start
1:09:28introducing bid caps or you might start
1:09:30introducing cost caps or you might have
1:09:31maximize conversion value or target
1:09:33rorowaz that's sitting next to these
1:09:35campaigns. And the reason being is that
1:09:37each different bidding strategy will
1:09:38enter auctions differently with
1:09:40different bids and you will generally
1:09:42win more auctions as you start to
1:09:44diversify the bidding strategies within
1:09:47the account. The reason why I almost
1:09:49never talk about this in any content is
1:09:51because 99.9% of people are not spending
1:09:54over 250k a month and so should not be
1:09:56concerned whatsoever with bidding
1:09:58strategies. Now counter to that 60% of
1:10:00our client portfolio spends more than
1:10:02250k a month. So for us internally the
1:10:05bidding mix actually is a big deal and
1:10:07this actually is something that we need
1:10:08to think about and think through but for
1:10:10most people ignore it. What will also be
1:10:12the case at this spend level is you'll
1:10:13generally have some kind of
1:10:15international market expansion in which
1:10:17case you need to start thinking through
1:10:19the complexity that gets introduced into
1:10:20the ad accounts from that international
1:10:22expansion. Whether you run secondary ad
1:10:24accounts for different regions, how you
1:10:26deal with pages, how you actually deal
1:10:28with the backend on Shopify or however
1:10:30you're hosting the different regions. A
1:10:32general word of advice is that I would
1:10:34advise against segmenting countries out
1:10:37at an ad account level. And the reason
1:10:39being from an agency perspective is that
1:10:41it will increase your costs because
1:10:43working across multiple different ad
1:10:45accounts increases labor dramatically.
1:10:47And so I would always rather work on one
1:10:49ad account than working across seven ad
1:10:51accounts which we have some clients that
1:10:53have seven ad accounts for seven
1:10:55different regions. And it adds so much
1:10:57additional labor and complexity into the
1:10:59management across it. Now the reason why
1:11:01you would have all those ad accounts is
1:11:03really only one reason and it's that you
1:11:05want to get build in the local currency
1:11:08of that region. So if you have a US web
1:11:11presence, you want the ad account to
1:11:12bill you in USD because maybe you have a
1:11:14US bank. If you have a UK presence, you
1:11:16want all of your money flowing through
1:11:17in great British pounds and so you need
1:11:19a separate ad account. That's really the
1:11:21only convincing argument I have seen for
1:11:23introducing segmentation. Other than
1:11:25that, everything else is solvable. You
1:11:26could say, "Oh, reporting is better
1:11:28because we can plug these ad accounts
1:11:29into our dashboards and reporting."
1:11:30Yeah, but you can just add filter rules
1:11:32based on country segmentation or you can
1:11:33just add filter rules at a campaign name
1:11:35level and all of that's solved. Like,
1:11:36you don't need to introduce ad account
1:11:38complexity other than for the reason of
1:11:40just getting build-in local current.
1:11:42Now, let me give you a worked example of
1:11:44an actual real ad account that's
1:11:45spending 25k per day, which is about
1:11:47750k a month. So, closing up on a
1:11:50million a month. And this performs
1:11:51incredibly well for them. Now, note this
1:11:53will not guarantee results for you
1:11:55because your business has its own
1:11:56complexities and its own differences and
1:11:58you should think through everything that
1:11:59we've gone through so far in terms of
1:12:00how to structure it. Or you can
1:12:01obviously always click the link in the
1:12:03description, reach out to us, we will do
1:12:04a free audit as long as you're doing at
1:12:06least $5 million a year in revenue and
1:12:08we can walk you through what that
1:12:09account structure might actually look
1:12:10like and we can provide it to you. So,
1:12:11we have a testing campaign at the top.
1:12:13It's an so all tests are still
1:12:16being done at the adset level. I have
1:12:17actually seen accounts that are spending
1:12:20$300,000 a day in budget and they're
1:12:23still running tests. So, I see a lot
1:12:24of push back, oh, you shouldn't run
1:12:26testing once you're actually a big
1:12:27account and you're spending a lot.
1:12:28That's not true at all. Okay? Like, you
1:12:30can run CBO or It's up to you and
1:12:32there's benefits of each one. It just
1:12:34depends on how you want to manage the
1:12:35account and also what the particular
1:12:37nuances are of that business. ABOS's at
1:12:39this spend level work and you can
1:12:40perform incredibly well and there's
1:12:42reasons why you want to do them because
1:12:43you want to force spend through your
1:12:44tests. and CBOS can also work incredibly
1:12:46well at this spend level. So this is
1:12:47once again up to you, but this ad
1:12:49account runs a testing campaign as an
1:12:50Number two is then a scaling
1:12:52campaign, which is a CBO, one adset,
1:12:54cost caps. Every one to two weeks, the
1:12:56top performers in the the post IDs
1:12:59are taken and they're launched into the
1:13:00scaling campaign. Now, they're not
1:13:02turned off in the testing campaign. The
1:13:03testing campaign is also used to scale
1:13:05at the adset level. There's adsets in
1:13:07here spending multiple thousands of
1:13:08dollars a day. You still scale in the
1:13:10testing campaign. It's just this is a
1:13:12strategy to try to squeeze more out of
1:13:14an existing post that's doing well.
1:13:15Third campaign is a promo campaign. This
1:13:18particular account runs promos every
1:13:19three to four weeks. And so as a
1:13:21function of that, we want it segmented
1:13:22out. Why? Because the promos turn over a
1:13:25lot. And so if you're launching promos
1:13:26in the testing and the scaling campaign,
1:13:28it will disrupt the learnings of the
1:13:30campaign because you're constantly just
1:13:32turning stuff off and launching new
1:13:33stuff in it. You want to leave stuff
1:13:35that's working and you want to add pipes
1:13:37at this level of spend. So we want to
1:13:39add this in separate to not impact the
1:13:41performance over here. This will also
1:13:42generally go after a different customer.
1:13:44So it is a little bit different in
1:13:46prioritizing price sensitive consumers
1:13:48and so as a product of that we do want
1:13:50to optimize a little bit separately. The
1:13:51fourth is an advertorial campaign. So
1:13:54advertorials as an additional funnel for
1:13:56this business started to do very well in
1:13:58the testing. It started to do so well to
1:14:00the point in which it was consuming
1:14:01about 30% of total spend. And so it
1:14:03actually made sense to just pull it out
1:14:05and have it dedicated here. So it can be
1:14:07KPI on its own and it can be looked at
1:14:10as a completely separate funnel within
1:14:12the account. And then lastly, a DPA
1:14:13campaign. This is primarily for
1:14:15retargeting existing customers as well
1:14:17as a little bit of website visitors.
1:14:18This is running on incremental
1:14:20attribution so that we're attributing
1:14:22correctly based on its actual
1:14:23incrementality. And this is kept at a
1:14:25relatively low spend in line with
1:14:26frequency. What doesn't change at this
1:14:29spend level? What stays the same? Well,
1:14:31number one is the two non-negotiables.
1:14:33Everything that we introduce, every
1:14:34extra bit of complexity needs to still
1:14:37ensure that we have data integrity so
1:14:39that we can read the data and then make
1:14:41decisions. And it all needs to have
1:14:43commercial alignment. Is this aligning
1:14:45with the products, the margin, the
1:14:46portfolio, the personas, or are we just
1:14:48adding complexity for the sake of it
1:14:50when we don't need it? Number two is the
1:14:52breakdown effect still applies.
1:14:54Suddenly, just doing breakdowns or
1:14:56looking at the ad level does not become
1:14:57more productive here than it does at a
1:14:59lower spend. It is the same thing. The
1:15:01principle still stays. Number three, the
1:15:03concept framework of launching creatives
1:15:06still remains the same. You can have
1:15:08this framework up at 300k a day in
1:15:10spend. In fact, I would recommend it.
1:15:12Normally, the big accounts that are
1:15:13spending those levels are structuring
1:15:14creative and testing it in this way. And
1:15:16number four, you want to be KPIing at
1:15:18the adset and campaign level. This
1:15:21doesn't change. The premise really is
1:15:23that you're adding structural layers
1:15:26into the account as spend increases and
1:15:28business complexity increases, but
1:15:30you're not giving away the fundamentals
1:15:32that we spent the first 30 minutes of
1:15:34the video setting in place. And then the
1:15:35question I've alluded to throughout the
1:15:37whole video is versus CBO. It's the
1:15:39wrong question. It's honestly personal
1:15:41preference. And the reason being is that
1:15:42this is just what your risk portfolio
1:15:44is. Okay? If you want more risk in the
1:15:46account, but you want better efficiency,
1:15:48you go for CBO. The reason being is it's
1:15:50going to distribute more spend to the
1:15:52highest performing ads and the highest
1:15:54performing adsets, which is good, right?
1:15:55You're going to get the highest ROI or
1:15:57return on ad spend in the account. The
1:15:58disadvantage is that all the spend is
1:16:00just going to go to the highest ads and
1:16:02so you can launch new tech, new ads, and
1:16:03all this money on creative production,
1:16:05put it into the account, and it gets no
1:16:06spend. So, if you just went and spent
1:16:08$20,000 on a new campaign shoot, and
1:16:10then you went put all that content into
1:16:12the account, it's not spending, what do
1:16:13you do? or let's say that you went and
1:16:15just spent $20,000 on a very expensive
1:16:18influencer to run a partnership ad for a
1:16:2060-day period, put it in the account,
1:16:22isn't getting spent. Now, you can put
1:16:24minimum spend caps in CBOS, of course.
1:16:27So, you can go and put minimum spend
1:16:29limits. You can also put maximum spend
1:16:30limits as well, but then you're
1:16:32effectively just running an where
1:16:35either you're just increasing the
1:16:36complexity of management because running
1:16:38min and max spend caps is just annoying.
1:16:39It's just a harder management tool
1:16:41within the account. Or you do this with
1:16:43like let's say 60% of budget and then
1:16:45you let 40% of budget get distributed by
1:16:47meta hour one. Now that's actually in my
1:16:49opinion the best structure to run. Most
1:16:51of the structures that we run these days
1:16:53is that structure which is we're running
1:16:55CBO but we're using minimum spend limits
1:16:58and maximum spend limits but we're
1:17:00allowing about 40% of total budget to
1:17:02just flow wherever it wants and 60%
1:17:05we're tightly controlling and saying you
1:17:07have to go here, you have to go. The
1:17:08reason why I'm not a proponent of this
1:17:10at like lower spend level accounts is
1:17:12that this is just like a lot of
1:17:13management and it requires a lot of
1:17:15oversight, which is fine for us when
1:17:17we're managing very large accounts with
1:17:19large ad spend with large revenue. It's
1:17:21worth your investment for someone like
1:17:22us to be able to do this kind of degree
1:17:24of management and budget segmentation.
1:17:26But if you're managing like 10k a month
1:17:27in ad spend, this is it's just kind of
1:17:29overkill. Like just run an or just
1:17:31run a CBO. Choose your risk versus
1:17:32efficiency profile and run it. So that's
1:17:34really the thought through of the
1:17:36process. So if you prefer ABS and once
1:17:37again I have seen ABOS's running with
1:17:39majority of ad spend on ad accounts
1:17:40spending a4 million a day and I've also
1:17:42seen CBOS running on ad spends quart
1:17:45million a day. So it really comes more
1:17:47so down to personal preference as to how
1:17:50you want the testing methodology rolled
1:17:52out in the account. Do you want spend
1:17:54forced through every new creative or do
1:17:56you not? Do you trust that the algorithm
1:17:57is only going to give spend to an ad if
1:17:59it's going to do well? That all comes
1:18:00down to your trust in the algorithm.
1:18:01What you have seen historically
1:18:03subjectively within the account. Has the
1:18:04algorithm not given spend to an ad
1:18:06before and then you went, "Oh, I thought
1:18:07this is a good ad. Let's put a minimum
1:18:08spend cap." You put a minimum spend cap,
1:18:10it becomes a winner. Immediately, you
1:18:12start losing your trust that the
1:18:14algorithm knows what it's doing cuz it
1:18:15wouldn't have spent on that ad if you
1:18:17didn't force it to spend and now it's
1:18:18the top performing ad in the account.
1:18:19So, it's going to come down to honestly
1:18:20your own bias and subjectivity in
1:18:22relation to how much you trust the
1:18:24platform in whether you're going to go
1:18:25or CBO. And I honestly don't think
1:18:27there's a right or a wrong answer. You
1:18:28can run whatever. They both perform
1:18:30well. The only caveat I will add is what
1:18:32I said previously, which is that if you
1:18:33have really high AOV and low conversion
1:18:37volume, I wouldn't recommend CBO. And
1:18:39the reason being is that the CBO will
1:18:40prioritize soft higher intent metrics
1:18:42like clickthrough rate and CPCs. And so
1:18:44you'll just end up with spend getting
1:18:46distributed to the best soft metric ads
1:18:48rather than to the best ads that are
1:18:50actually going to drive conversions
1:18:51within the business. There is also the
1:18:53complexity of we will move out and in of
1:18:57ABOS's and CBOS depending on seasonality
1:19:00which sounds kind of weird but because
1:19:02if you put these names aside and you
1:19:04just think high risk high ROI and then
1:19:07you think low risk slightly lower ROI
1:19:10like the ROI isn't that much lower here
1:19:11but it's a little bit lower. Let's say
1:19:1210%. Well, I want to pick this during
1:19:15Black Friday and November, December and
1:19:17peak periods where we're just trying to
1:19:18spend as much as possible in a short
1:19:20window. I don't care what creatives get
1:19:22spend. I just want to maximize
1:19:23efficiency and maximize spend. But then
1:19:25in like Jan Feb in Q1, I kind of want
1:19:28this, right? I want to set Q1 up for a
1:19:31bunch of creative testing. I want to
1:19:32test as much as possible. This is a
1:19:34great environment for testing because
1:19:35you aren't artificially inflating
1:19:37conversion rates, making ads seem like
1:19:38they're doing well when they're actually
1:19:39not. We have a relatively low risk
1:19:41profile during January. We don't want to
1:19:43increase our risk profile during this
1:19:45part of the year. And so what we will
1:19:46often do is actually rotate into a CBO
1:19:49in Q4 for some accounts and then roll it
1:19:51back to an in Q1. Now regardless of
1:19:53whether you run ABOS's or CBOS's, the ad
1:19:56set composition remains the same, which
1:19:58is that you always want a minimum of
1:20:00three ads in an ad set because it allows
1:20:02the adset to sequence across ads. If you
1:20:04only put one ad in an ad set, it can't
1:20:07sequence across anything. And so it's
1:20:08just going to serve that one ad to
1:20:09users, which ultimately isn't going to
1:20:11get people to buy because generally
1:20:12people need to see an ad 2, three, four,
1:20:14five, six times before they actually
1:20:15purchase. You want a minimum of three
1:20:17ads. You want a maximum of really
1:20:19infinity depending on spend. And so you
1:20:21can go all the way up to the maxed ad
1:20:23limit at an adset level. That's fine. We
1:20:24have accounts where we run that. We
1:20:26might have 200 creative in an adset.
1:20:28Generally, for most people watching
1:20:29this, you don't have anywhere near
1:20:31enough creative volume to support 200
1:20:33ads in an adset level. Like you may as
1:20:34well have a degree of segmentation
1:20:35there. and there's likely going to be
1:20:37different concepts in those 200 ads you
1:20:39don't want to consolidate anywhere. So,
1:20:40as a rule of thumb, you can think about
1:20:41three to 15 ads per adset. Number two is
1:20:44that you want all of your ads resonating
1:20:47with the same audience. Now, this
1:20:49doesn't mean that you need different
1:20:51stages of awareness or different types
1:20:53of ads. So, there's a really common
1:20:54question I get, which is that if I
1:20:55launch an adset with five ads in it and
1:20:58it's all under one persona, it's talking
1:20:59to moms 40 to 50 who need a rain jacket.
1:21:03Do I put every stage of awareness in
1:21:05there? Do I put my super topunnel VSSL
1:21:07style ads that are trying to sell this
1:21:09weird raincoat through a story line and
1:21:11a 5minute video, but then also just put
1:21:13a static ad that just says here's a
1:21:15raincoat 50% off. Do I put them in the
1:21:16same ads there? Because they're at very
1:21:17different stages of awareness. So, do I
1:21:19split that at the adset level? And the
1:21:20answer is you can do either. It's just
1:21:22dependent on how you want to read the
1:21:24data in the platform. And so, let me
1:21:26walk you through an exact example. So,
1:21:28option one is you split it out. You have
1:21:31one adset which is the persona, the
1:21:33raincoat, but this is very top offunnel
1:21:35creative. It's high up on the stages of
1:21:38awareness. And then you have adset two
1:21:40which is going to be very bottom of
1:21:41funnel in the stages of awareness in the
1:21:43way that this creative is designed.
1:21:44That's adset two. Option two is we just
1:21:46consolidate it. This has our top of
1:21:47funnel, it has our middle of funnel, it
1:21:49has our bottom of funnel ads all
1:21:51consolidated under the one adset. Which
1:21:52option is better? My understanding of
1:21:54meta's machine learning and the way that
1:21:56conversion data is split across the
1:21:57account generally pushes me a little bit
1:21:59more towards option. The reason being is
1:22:01that we want these ads to be able to
1:22:02easily sequence against each other,
1:22:04bring people through the journey and
1:22:05convert. However, I have seen option one
1:22:07also work. The reason why option one is
1:22:10an inferior option. The reason why I'd
1:22:12recommend most people don't do this and
1:22:14they go for option two isn't to do with
1:22:15the actual machine learning and the way
1:22:17that it works. Instead, it is to do with
1:22:19your own human buyer. This adset down
1:22:21the bottom here will have a 4x row ads.
1:22:23This adset up here will have a two. This
1:22:25will have a six. Now, if you go into
1:22:27this account on option two and you look
1:22:29at this adset and it's a 4x, your
1:22:30conclusion is this concept is doing
1:22:32well. Let's do more like this. Let's
1:22:34also scale budget. Incredible. Now,
1:22:36let's say instead you go into this ad
1:22:37account and you go, okay, we have two
1:22:39adset. This one's not doing well. This
1:22:41one's doing well. What do we do? Do we
1:22:42blend the numbers and say that overall
1:22:45the concepts at 4x, therefore do more
1:22:47top offunnel creative? Is the top
1:22:49ofunnel creative really performing well?
1:22:51Do we turn this off and just like leave
1:22:53bottom of funnel creative for this cont?
1:22:55Like what do we do off the back of it?
1:22:56Now the answer in terms of what you do
1:22:58is you should just blend the numbers,
1:22:59assume that these two adsets are working
1:23:01together to get the conversion and
1:23:02therefore do more scale it etc. Exactly
1:23:04what you would do here. But due to your
1:23:06own human bias in the way that you're
1:23:07going to read the data, you're
1:23:08inherently not going to want to do that.
1:23:10You're going to look at this 2x and
1:23:12particularly if you're working with
1:23:13clients that aren't inclined to think of
1:23:15it like this, they're going to go, "No,
1:23:16no way. Don't put more spend into a 2x.
1:23:18That's not a good idea. Put more spend
1:23:20here. Even though this is obviously
1:23:21bottom of funnel and then this traffic
1:23:23is driving into here which is converting
1:23:24and this is getting the last click
1:23:25attribution. And so this just adds a lot
1:23:27more complexity into the human
1:23:29decision-making process. And the human
1:23:31decision-making process is where this
1:23:32falls apart. And so I would recommend
1:23:34that you just put your middle, bottom,
1:23:36top of funnel ads all together under the
1:23:38one adset. Now this is an underrated
1:23:40topic when it comes to adset structure,
1:23:42which is actually naming adsets
1:23:43correctly. What you'll see in most ad
1:23:45accounts is something like this. March 7
1:23:49creatives. So as batches of creatives
1:23:51are made, they're just named, thrown in
1:23:53the adset level, and rap. The issue with
1:23:55this is that there's no ability to
1:23:56easily filter into what concept we're
1:23:59actually testing here. This also infers
1:24:00that we're just batching a bunch of
1:24:01concepts together rather than carefully
1:24:03structuring them so that we can get
1:24:05clear data insight and iterate into the
1:24:07creative team. And so as a product of
1:24:08that, when we pick up accounts like
1:24:10this, it's very annoying and complex.
1:24:12And often we'll actually go back and do
1:24:13retrospective renaming of old adsets so
1:24:16that we can at least get some kind of
1:24:18clear data structure historically so we
1:24:20can infer what best to double down on
1:24:21moving forward because if you don't have
1:24:23clear naming conventions in place of for
1:24:25example you'd want to rename this into
1:24:26the exact persona angle offer and then
1:24:29content type if you're going to group a
1:24:31content type and then date of launch so
1:24:33that we can start to bundle and KPI
1:24:35different concepts across a campaign. So
1:24:37what about when it comes to retargeting
1:24:39DPAs and existing customers? Here's
1:24:41three principles you need to keep top of
1:24:43mind. Retargeting is a capped supporting
1:24:46layer. It's absolutely not a growth
1:24:48driver. This is unfortunately a mistake
1:24:50that I see in so many ad account audits
1:24:53that we do, which is that you go in, you
1:24:54do an audience segment breakdown, you
1:24:56pull up from the bottom of the account,
1:24:58there's a little button there, and you
1:24:59can just see total spend distribution
1:25:01across the account, and you just see
1:25:02like 50% of spend going to existing
1:25:04customers and engaged audiences. like
1:25:0650% of your spend is going towards
1:25:08support spend that's not actually going
1:25:09to drive incremental new customer growth
1:25:11within the business. It's just
1:25:12supporting existing spend and it's not
1:25:14even that incremental. And so the key
1:25:16premise to keep top of mind here is that
1:25:18you want to be at all times minimizing
1:25:21retargeting spend as much as possible.
1:25:23This is the different thought process
1:25:25between cold and retargeting which is on
1:25:27cold on new customer acquisition. The
1:25:29constant question that we're asking
1:25:30ourselves internally is where can we
1:25:32spend more money? What platform? What
1:25:34channel? What campaign? more ad. Where
1:25:36can we spend more? We need to spend more
1:25:38to get more incremental lift. As long as
1:25:40it's profitable, as long as we can make
1:25:41a profitable exchange, we can put $1
1:25:43into this ad and get four out. Or we can
1:25:44put $1 into Tik Tok and we'll get four
1:25:46out. We want to keep putting more money
1:25:48there to grow new customers and grow the
1:25:50business on existing customers and
1:25:51retargeting. It is the opposite frame.
1:25:53We want to be thinking, how can we spend
1:25:55the least amount of money? Where can I
1:25:56take money out of? Are we spending too
1:25:58much retargeting on TikTok? Take it out.
1:25:59Take it out of Pinterest. take it out of
1:26:01better because often the top end of
1:26:02spend on retargeting is actually not
1:26:04incremental. We're almost always
1:26:06overspending and so we want to be in
1:26:08almost a scarcity mindset on retargeting
1:26:10and then an abundance mindset on cold
1:26:12targeting and new customer growth.
1:26:14Number two on retargeting is
1:26:16consolidation of audiences. Uh what is
1:26:18very like 2019 and you should not be
1:26:20doing is having a bunch of different
1:26:22adsets segmented by like retargeting
1:26:24different product categories based on
1:26:26what landing page they landed on or
1:26:27retargeting add to cart separate from
1:26:29initiate checkout separate from website
1:26:31visitors. That hyper segmentation and
1:26:33retargeting. All it does is increase
1:26:35CPMs. It just makes it more expensive
1:26:37for you to serve retargeting ads and it
1:26:39has almost no direct upside. If someone
1:26:41can give me a really strong reason as to
1:26:42why their ad account has six different
1:26:44adsets targeting six different degrees
1:26:46of separation of warmth in the funnel,
1:26:48like I don't know what you're doing. The
1:26:49CPM increase is never outweighed by the
1:26:52conversion rate increase. And then
1:26:54number three is really your core KPI
1:26:56other than obviously incremental revenue
1:26:58on retargeting campaigns should be
1:27:00frequency. You should just be looking at
1:27:01frequency very tightly to understand how
1:27:04many times are we retargeting existing
1:27:06customers, engaged audiences every
1:27:07single month, every single week. And is
1:27:09this intuitively or subjectively too
1:27:11high or too low? Should we be
1:27:12retargeting existing customers 30 times
1:27:14a month? For most people, probably not.
1:27:15And so intuitively, you know, you're
1:27:16probably overspent. So you can probably
1:27:17pull that down. Now, a big part of
1:27:19retargeting is DPAs, which is dynamic
1:27:22product ads. Uh, DPAs are a pretty
1:27:24critical component of most ad accounts,
1:27:26particularly if we're talking about
1:27:27fashion, particularly if we're talking
1:27:29about large cataloges or SKs within a
1:27:31product portfolio. DPAs are great
1:27:33because they're going to dynamically
1:27:34retarget a user with a catalog ad that
1:27:37is going to rep prioritize products
1:27:39based on the products that they looked
1:27:40at on the website if everything's
1:27:42working correctly and if the pixel
1:27:43worked and actually fired the correct
1:27:44data back. Amazing. Now, the bad thing
1:27:46about DPAs is that they will pretty much
1:27:49only place right before the moment of
1:27:52purchase as like a bottom offunnel most
1:27:54aware retargeting ad. And so, because of
1:27:56that, they get way more credit than they
1:27:58should actually get. The rorowaz on
1:28:00these things looks way better than it
1:28:02actually is. And so, people end up
1:28:04naively overspending on them
1:28:06substantially. And so, there's a few
1:28:08rules when it comes to DPAs. Number one
1:28:10is make sure that when you're assessing
1:28:12return on ad spend, you're assessing
1:28:13based on 7-day click or based on
1:28:16incremental attribution. If people don't
1:28:18know what I'm talking about with these
1:28:20attribution settings, just watch our
1:28:22video that was posted on the 18th of May
1:28:24called if your ad account looks good but
1:28:26profit isn't, watch this video. It's an
1:28:28hour and 2 minutes long. It goes all
1:28:30into attribution models and data
1:28:31integrity so you can better understand
1:28:32that. Number two is that your DPA
1:28:34performance is really just a direct
1:28:36reflection of your top offunnel
1:28:38investment. If you increase budget on
1:28:40top of funnel, if you go harder on top
1:28:41of funnel, DPA row goes up because you
1:28:43just capture more revenue at the bottom.
1:28:45If you decrease top of funnel, your DPA
1:28:47performance will go down. We actually
1:28:48made an entire video on this like 6
1:28:50months ago, which was called the DPA
1:28:51death spiral of fashion brand. And the
1:28:53whole idea was that I had seen six
1:28:55fashion brands in a row on six audits I
1:28:57did in the course of like 2 weeks where
1:28:58all of them were spending 80% of their
1:29:00spend on DPAs and the account was just
1:29:02declining over the past 6 months. And
1:29:04the reason being is that they saw the
1:29:06performance on DPAs. The agency just
1:29:07kept bumping budgets cuz it was good.
1:29:09And then the brand as a function of the
1:29:12agency's advice was like, "Oh, it's DPAs
1:29:14are doing so well. Do we need to make as
1:29:15much creative?" And they were like, "No,
1:29:17you don't because DPAs are holding up
1:29:18performance." So they just started
1:29:19decreasing creative production and
1:29:21creative and top ofunnel spend
1:29:23allocation within the ad account and
1:29:24then the whole business just falls apart
1:29:26because if DPAs sit at the bottom of the
1:29:28funnel and then you just stop spending
1:29:29on top obviously that is not
1:29:30sustainable. Then when we think about
1:29:32existing customer spend allocation, what
1:29:35happened which was really bad for the
1:29:37entire performance marketing and ecom
1:29:39space in my opinion was that Meta
1:29:41launched advantage plus campaigns like 2
1:29:43years ago and the first setting that
1:29:45there was on advantage plus campaigns
1:29:47was this little box that you could tick
1:29:49and it was called existing customer
1:29:51spend cap and it was a percentage input.
1:29:53So you would input your percentage that
1:29:55you wanted going towards existing
1:29:57customer. This framed everyone into the
1:29:59mindset of not only thinking about
1:30:01existing customer spend, which was good,
1:30:03but it got them framed towards, yeah,
1:30:05what percent should we put towards
1:30:06existing customers? What should our
1:30:08percentage be in terms of meta-pend
1:30:09allocation? It's a terrible way to look
1:30:11at it. It's the wrong frame entirely
1:30:13because percentage is irrelevant. Who
1:30:16cares what the percent is towards
1:30:17existing customers? I only really care
1:30:19about two things. Number one, is it
1:30:20profitable? If we spend another $10 a
1:30:22day or $1,000 a day on existing
1:30:25customers, do we make enough return
1:30:26revenue to be able to pay for that spend
1:30:28and make profit? Number two, which is
1:30:30kind of a softer metric that gives us
1:30:32more realtime visibility, is what is our
1:30:34frequency? Yes, sure, we can have a 20%
1:30:36existing customer spend, but if that
1:30:38means that our frequency on existing
1:30:39customers is a 50 every 30 days, what
1:30:42are we doing? Why are we targeting
1:30:43people 50 times a month? Vice versa, if
1:30:46our spend is 20%, but we only have a
1:30:48frequency on existing customers of one,
1:30:50we're probably substantially under
1:30:51spending. And so your existing customer
1:30:53spend as a percentage is a bad metric
1:30:55because percentages change based on how
1:30:58big the existing customer pool is and
1:31:00how much budget allocation you currently
1:31:01have to new customer acquisition. Those
1:31:03two variables will just substantially
1:31:05change the spend allocation as a
1:31:07percentage to where it's useless. So how
1:31:09do you think about it instead? You just
1:31:10think about it as a dollar value. So,
1:31:11how much dollar spend should we have per
1:31:14month to existing customers? And that
1:31:16you can calculate by taking your amount
1:31:18of customers, you times by your CPMs,
1:31:21and then you times by the amount of
1:31:23frequency that you want per month on
1:31:25those customers. And this tells you
1:31:26exactly what your monthly budget should
1:31:28be towards an existing customer
1:31:30audience. So, here's some settings that
1:31:31you should be turning off and not using.
1:31:33And number one is flexible ads. I
1:31:36explained briefly why before, which is
1:31:38the fact that you can load a bunch of
1:31:39creative into one single ad, but the
1:31:41issue is that that ad performs well or
1:31:43if it doesn't perform well, we don't
1:31:44actually know what creative was driving
1:31:46performance or not. And so the issue
1:31:47here becomes that it doesn't follow the
1:31:49number one foundation that we had for
1:31:52account structure, which is data
1:31:53integrity, which allows feedback loops.
1:31:56So we can read the data and then we can
1:31:57make decisions. On flexible ads, you
1:31:59can't read the data. So it's
1:32:00fundamentally a bad ad type. Now, the
1:32:02only reason why you would do this is to
1:32:03try to get around the page limit within
1:32:05the account. There's much better ways to
1:32:07get around the page limit. I already
1:32:08gave them to you. If you still can't do
1:32:10that, if you're going to do flexible
1:32:11ads, you need to put a bunch of creative
1:32:13that's very similar. And so that if it
1:32:14does perform well, we can kind of say
1:32:16that this type of ad does well because
1:32:18it's all similar. If you put any kind of
1:32:20large variation of asset types under a
1:32:22flexible ad, you then no longer know
1:32:24what's actually working and you're just
1:32:25introducing stuff into the account that
1:32:27doesn't facilitate a decision-making
1:32:29feedback loop. why we as an agency
1:32:31really don't like flexible ads. On top
1:32:32of that is that you don't have control
1:32:34over cropping the images and so it can
1:32:37serve the one by ones and story
1:32:39placements. It'll randomly crop the
1:32:40story placements into feed placements.
1:32:43It's generally terrible for the actual
1:32:45quality of how the ad presents on the
1:32:46feed. What Meta can also do sometimes is
1:32:48if you have a bunch of flexible ads, it
1:32:50will just string them together and
1:32:52create a carousel out of them, which
1:32:54also isn't ideal when a client hasn't
1:32:56approved a carousel. That's a string of
1:32:58random images that have been put in the
1:32:59account. They're generally not a fan.
1:33:01Next is cost caps or bid caps. I just
1:33:04really wouldn't be concerned with
1:33:06worrying about bidding strategies if
1:33:08you're spending under $200,000 a month
1:33:10in an ad account. It's just not the
1:33:12right bottleneck to be solving for
1:33:14within the account. The question that
1:33:15you should ask yourself is, will the
1:33:17business double if we just really focus
1:33:18on bidding strategies within Meta? And
1:33:20the answer normally is no because it's a
1:33:22low leverage opportunity. Now, is that
1:33:24to say that you shouldn't run cost cap?
1:33:26Not at all. Like if you know what you're
1:33:27doing, you can run cost caps on any
1:33:28spend level. More so, I'm saying
1:33:30focusing on changing and testing bidding
1:33:32strategies at lower spend levels
1:33:34generally isn't the biggest lever to be
1:33:36able to unlock growth. Number three is
1:33:38turn off one day view attribution or at
1:33:42least don't use it within the decision
1:33:44making of how you're flowing budgets at
1:33:45an adset level and what's working and
1:33:47what's not because it will substantially
1:33:48overattribute based on existing customer
1:33:50revenue. Number four is you want to
1:33:52check for all of the advantage plus
1:33:54creative optimizations. Um, particularly
1:33:57you want to go into your advertiser
1:33:59settings and go into creative sub
1:34:02dropown and then there'll be a bunch of
1:34:04automated rules that are turned on which
1:34:07will effectively flip your creative
1:34:08optimization settings back on over time
1:34:10if you don't turn it off at the ad
1:34:12account advertiser setting level. So,
1:34:14not only do you need to do this on every
1:34:15single ad that you launch, which is
1:34:17incredibly annoying, but you also need
1:34:18to do it at the advertiser setting
1:34:20level. And then number five is you
1:34:21really want to make sure that your
1:34:22audience segments are set up correctly.
1:34:25What has happened a lot and that we keep
1:34:27an eye on is that the sync of the
1:34:30audience segment from Clavio to meta
1:34:32will break periodically. And when that
1:34:34happens, your audience segment stops
1:34:36flowing through existing customers
1:34:38correctly and so you end up spending on
1:34:39existing customers thinking they're new
1:34:41and obviously that starts to erode data
1:34:43integrity within the account and you
1:34:44start spending in places that you don't
1:34:46want to. When it comes to testing
1:34:47discipline in the account, when you're
1:34:49setting up any kind of test, you want to
1:34:51answer two questions that are very
1:34:52similar to what we talked about before,
1:34:53which is what's the stat threshold? So,
1:34:56how many conversions do we need to say
1:34:58that it's statistically significant? Is
1:35:00it five? Is it 10? What is that
1:35:02threshold? And then number two is how
1:35:04quickly do we want to actually learn?
1:35:06So, do we want to spend to that
1:35:08statistical significance in a 3-day
1:35:10period, in a 7-day period, in a 30-day
1:35:12period? How are we spending to the
1:35:14threshold to then be able to move out of
1:35:16the test or say that the test was a su
1:35:19success and scale it? The way that you
1:35:20actually calculate it using these two
1:35:22questions is that your daily testing
1:35:24budget on a test is your target
1:35:26conversions times by your expected CPA.
1:35:29So what is your expected cost per
1:35:30acquisition divided by the test duration
1:35:33in days. For example, if our target
1:35:36conversions is 20, that's the threshold
1:35:38in which we'll call the test
1:35:39statistically significant. Our expected
1:35:41cost per acquisition is $100. We want
1:35:43the test to go for 14 days. Then the
1:35:46daily budget on this test is $143 per
1:35:50day. Now, this is also why the common
1:35:53feedback from Meta is wrong, which is
1:35:55that Meta says that you need 50
1:35:57conversions every 7 days to be able to
1:36:00exit the learning phase and achieve
1:36:02statistical significance. The issue is
1:36:03is that when you try to back propagate
1:36:05this into daily test budgets, it just
1:36:07blows out and becomes ridiculous for
1:36:08like 99% of avatars. So if you take this
1:36:11exact test case as an example, if we
1:36:13need to target 50 conversions and CPA is
1:36:16obviously the same and we need to do it
1:36:17in 7 days, that means daily budget here
1:36:20needs to be $714
1:36:23per day. And that's just for one concept
1:36:26adset test, which is just ridiculous cuz
1:36:28if we want to put like five tests in or
1:36:3010 tests in, suddenly you're spending
1:36:32$7,000 a day just on the testing
1:36:35campaign. If I was to compress this
1:36:37whole training into just four questions
1:36:39that you should write down and walk away
1:36:41from when you're looking to build an
1:36:42account structure is number one, does
1:36:45the structure produce data that I can
1:36:47read to make decisions? Because
1:36:48ultimately the whole point in the
1:36:49account structure is that we're
1:36:51producing readable data so that we can
1:36:53continue to iterate and make decisions
1:36:55within the business. Number two is does
1:36:56the structure match how the business
1:36:59actually makes? So is the structure
1:37:01commercially aligned to the objectives
1:37:03of the business and how the business is
1:37:05producing profit. Number three is the
1:37:07structure appropriate for the spend tier
1:37:09and creative output creative throughput
1:37:11as well as the amount that you're
1:37:12spending is going to substantially
1:37:14change the structure that can be allowed
1:37:16within the account. And then lastly, are
1:37:17the support layers, so DPA, retargeting,
1:37:20website visitors, existing customers, is
1:37:22that aligned with the actual retention
1:37:25in the business? Do we need this amount
1:37:27of spend? How much spend do we need? How
1:37:28are we structuring DPAs and retargeting?
1:37:30How much spend distribution is going
1:37:32here? This is all going to be once again
1:37:34context dependent on the actual business
1:37:36and the retention within the business.
1:37:38If the answer to any of these four
1:37:40questions is no, then the account
1:37:42structure is wrong and you need to
1:37:44restructure accordingly. So, that's the
1:37:46end of the video. Two things. Number
1:37:48one, if you're a performance marketer,
1:37:49we're always hiring. Please reach out to
1:37:51us at [email protected].
1:37:53If you're an e-commerce brand or retail
1:37:55business doing over $5 million a year in
1:37:57online revenue, click the link below.
1:37:59You can get a free audit from us where
1:38:01we'll break down how all of this
1:38:02strategy actually translates into your
1:38:04account. We'll give you the nuances of
1:38:05account structure. And if you're in
1:38:06neither of those two buckets, subscribe.

About this transcript

This page contains the full transcript of How To Structure A Meta Ads Account At Every Spend Tier In 2026 by Blue Sense Digital, generated from the public captions YouTube serves with the video. The transcript has 21,727 words across 3,138 segments, with the original timestamps preserved so you can click any line to jump to that moment in the embedded player.

What you can do with it

Use the transcript to take notes, quote the speaker, build a study guide, generate a summary with ChatGPT or Claude via the YouTube Summary tool, or export it as a timed subtitle file with YouTube to SRT. You can also re-open it in the transcriber to choose a caption track available on the video.

Free YouTube transcript tool

YouTube2Text is a free YouTube transcript generator — no signup, no daily limit. Paste any YouTube link and get the full transcript instantly, with timestamps, click-to-jump, caption language selection, AI prompts for ChatGPT, Claude, and Gemini, and exports to TXT, SRT, VTT, or Markdown.