Stop running campaigns. Start running systems.

00How to read this

This guide makes no claims of its own about what works on Meta. Every tactical statement traces to one of two operators, and is tagged accordingly. Where they disagree, you see both. Where I have extended their thinking into lead generation — which neither of them teaches directly at length — the panel is dashed and violet, and you should treat it as a hypothesis to test, not doctrine.

The ten sources

01The machine: what Andromeda actually does

Almost every mistake in a 2026 ad account comes from a wrong mental model of what happens between you pressing publish and a human seeing your ad. Get this right and the rest of the manual is obvious.

An ad is a web page competing in a search engine

Charley The first reframe. On Facebook and Instagram, every post — paid or organic — is a page on their site competing for attention in the feed. The feed is not a billboard. It is a search results page, continuously re-ranked.

“Every social media feed is basically just a Google search results page, where Andromeda is showing you content it thinks you'll engage with. Every single piece of content has a score called the estimated action rate, based on how likely people are to do things like engage and click.”

Charley T · 2026-05-11 / 2026-07-18

Your ad does not buy an impression. It earns one, by out-ranking organic content, your competitors' ads, and your own other ads, on predicted action rate. That is why the discipline in this manual looks more like SEO or content strategy than like classical media buying.

Two gates stand between your ad and a human

Charley Before your ad earns any real spend it passes two distinct systems, and they are optimising for different things.

The four metrics, and what they actually mean

Charley Charley's central diagnostic — the 4PI analysis — reads four numbers together to work out how Meta is using each ad. The full decision procedure is in §11; here is what each number means mechanically.

Why launching lots of ads actively harms you

Both This is the single most counter-intuitive claim in the manual, and both operators land on it independently of each other's framing.

New ads are always shown first to your warmest audience, because those people are most likely to engage and convert — which is exactly what a new ad needs in order to prove itself. So every new ad you launch is spending your money to compete against your own existing ads for the same warm conversions.

“The more new ads you launch, the more you invest in competition against your warmest audience. The new ads take spend and credit from your existing ones, so the older ads look worse. And you're not filling the funnel… You end up day trading the ad account.”

Charley T · 2026-05-11

“When you go in here and you put in a thousand different ads, the ads never realise how they work together and they're constantly competing with each other… Facebook doesn't know how the system goes together, so it's just shuffling these ads all over the place, and never optimising the system around it.”

Alex Becker · 2026-08-07

Charley And the platform is explicitly built to punish it:

“Meta specifically states that Andromeda was built because so many people were launching so many ads, and now it penalises that. What Meta wants is simpler — a few ads that do different jobs.”

Charley T · 2026-05-11

Creative diversity is not ad volume

Charley The defence people offer for launching fifty ads a week is creative diversity. Charley rejects the premise, citing Meta's own CMO:

“Alex Schultz, Meta's CMO, explains that when Zuckerberg talks about personalisation, he's not talking about making a different ad for every single person. He's talking about the overall experience somebody gets across the feed, across all of the touchpoints, paid and organic… Personalisation is a delivery sequence problem, not a creative volume problem.”

Charley T · 2026-05-11

Real creative diversity means a small set of ads that do different jobs in a sequence — which is exactly what §7 and §8 build.

02Why it feels broken when the data says it isn't

Meta made $196bn from ads last year, up 22%. More than ten million advertisers are on the platform. So why does your account feel worse every quarter? Charley's answer reframes the whole problem, and it determines what you should actually be optimising.

Era 4 explains your CPMs

Charley Era 4 was not defined by AI. It was defined by democratisation. Meta stopped building better tools for the biggest spenders and started building smarter tools for everyone else — Advantage+, Andromeda, automation. There are now more than 10 million advertisers, and over 80% of them spend less than $100 a day: restaurants, roofers, dentists, gyms, creators.

Those tools were not built for the brands that survived era 3. They were built to bring the next million advertisers online — and it worked. Demand grew faster than supply, which is why the auction is more expensive even though nothing about your account got worse.

ROAS is an averaging error

Charley Meta reports that the average advertiser earns well above $3 for every dollar spent, across $100bn+ in spend. Charley's objection is not that ROAS is a bad metric — it is that the average is meaningless:

“Imagine averaging the ROAS of a dentist, a restaurant, a roofing company, Nike, Amazon, and a $20 million supplement brand. You're going to get a number, but it won't tell you what good looks like, because none of those businesses make money the same way.”

Charley T · 2026-07-14

He then names the killer case: Allbirds peaked above $648/share after its 2021 IPO and hit an all-time low of $2.15. “You can chase 3× ROAS and still destroy the value of the business.”

The algorithm rewards a business model, not a niche

Charley Three companies kept winning across multiple eras, in totally different categories. What they share is not tactics.

Five generations — and where the control went

Charley Fifteen years of Meta updates read as one continuous project: taking decisions away from advertisers and making them better on their behalf. Fighting it has always been expensive.

03The offer, before you touch the ad account

Becker's entire contribution starts here, and he is deliberately extreme about it: nothing you do inside Ads Manager can rescue an offer that isn't built for cold traffic. This section comes before architecture because if you get it wrong, every later section is wasted motion.

“The big lesson I need to teach you in the video… nothing in the ad account that you do matters. You have one single goal when you start running ads. You need to make an offer so good that you can literally just run an ad, repeating the offer, and then showing the results, and it works.”

Alex Becker · 2026-07-30

Take the framing, not the literalism — Becker himself spends an entire other video on account structure eight days later. What he means is that offer quality dominates account quality by an order of magnitude, and most operators have the effort allocation backwards.

Cold traffic is a different buyer to warm traffic

Becker The most common structural error: selling on ads the way you sell on your website. Someone arriving at your site already has context, intent and some trust. Someone on Instagram has none of the three, and is one thumb-flick from something more interesting than you.

The five-part offer test

Becker An offer that works from cold ads has all five. Missing any one of them and, in his words, “it's not going to work” — regardless of vertical.

The 1-in-100 rule, and why lead gen should be high-ticket

Becker The economic argument for restructuring your offer upward. Send 100 cold people to three different price points and the conversion rates are far closer than intuition suggests.

The landing page is part of the offer

Becker One landing page for all traffic is, in his words, “really stupid”. Each buyer profile gets its own page, matched to the hook that brought them.

“I found CPA is the hot button for e-commerce marketers. So when they click that ad, the landing page talks about cost per acquisition as the main problem, and then shows them how we fix their cost per acquisition.”

Alex Becker · 2026-08-07

Two findings worth more than they look:

• Ugly beat beautiful. Becker built an elaborate animated page that scans the visitor's website and personalises the testimonials. His plain, “kind of ugly” page converts better. “In the age of beautiful AI websites, being very clear about what you do is going to get you more results.”
• Page production is no longer the constraint. He generates per-profile pages with Claude Code in minutes. If pages are cheap, there is no excuse for one-size-fits-all — which changes how many buyer profiles you can justify running.

Charley Charley arrives at the same requirement from the ad side: continuity. The voice in the ad should match the brand or the customer, the primary text should preview the landing page, and “when somebody sees that primary text and it's what they see on the landing page, the primary text that got that click is more likely to get that person to hit add to cart.”

04Systems, not campaigns

This is the conceptual core of the manual and the thing that makes everything else cohere. Both operators arrive at it; Becker names it most directly.

“Each campaign is not some series of ads all put together. It's not a bunch of ad sets sharing conversion data. What each campaign is, is a funnel to get a person from the top of the funnel to the bottom of the funnel and convert. And all the ads in the campaign — which you don't want a lot of them — are not used to just compete against each other and then get the last click. They're used to get people to the bottom of the funnel… they all work together like a unit.”

Alex Becker · 2026-08-07

“Facebook ads are a team sport. Messi is incredible, but he's not winning matches if he's on the pitch by himself… If you're not winning, improve the worst player on the team.”

Charley T · 2026-05-11

The consequence: you judge the system, never the ad

Both If the ads are a sequence, then no individual ad's ROAS means anything — it is an artifact of where that ad happened to sit in someone's journey and who got credit for the last click.

Ad fatigue is mostly operator error

Both A direct challenge to the industry's most common assumption.

“Ad fatigue's a really silly thing if you think about it. If the initial start of the ad works, why should it have to be changed every week? You're not running out of people you're showing it to, and if it's effectively reaching your target audience and bringing them down the funnel, you don't need to change it every week — because every new person goes down the funnel correctly.”

Alex Becker · 2026-08-07

Charley's version is blunter: “Ad fatigue is almost always operator error.” The mechanism is the one from §1 — you fatigue your own ads by launching competitors to them, then read the resulting decay as a law of nature. Becker has run essentially the same Hyros ad for six or seven years.

How good does each ad need to be?

Both Much less good than you think, and this is liberating.

• Charley “None of these ads have to be perfect. They just have to be good enough at their role in the system.”
• Becker “Once you get them mostly right, like 70% right, it's going to do most of the job… the entire system should work on its own even if the ads aren't perfect.”
• Becker His estimate of the ceiling on ad-level optimisation: a swing from a 4.7% to a 5.3% click-through rate. “Still meaningful, but not enough to completely break it either way. It really comes down to the concept itself.”

05Account architecture — and the one real disagreement

Charley and Becker agree on nearly everything in this manual. On how many campaigns you should run, they do not. This section gives you both structures exactly as taught, then a decision rule for choosing, then the budget constraint that overrides both.

Structure A — Andromeda One Charley

The modern version of the one-campaign method Charley says he invented five years ago. One CBO campaign. Everything lives inside it so that all spend runs through a single learning system.

Structure B — buyer-profile systems Becker

Becker does not run one campaign. He runs one campaign per buyer profile, each a self-contained funnel of 5–8 ads, each deliberately held at low spend, each pointed at its own landing page. There are no prospecting or retargeting campaigns, because both live inside every system.

The budget ladder overrides everything

Charley Before you copy any structure, the budget decides what you are allowed to build. The rule is about the learning phase, not about ambition.

Side by side

06The ad itself: the 322

One ad format carries the whole system. Charley says the three ads he shows that have each spent over a million dollars are the same format — and that after Andromeda it is the best kind of ad you can run, “and almost no one is using it properly.”

Anatomy

Charley Three creatives, two headlines, two primary texts, inside a single ad using Meta's multimedia creative workflow (the replacement for flexible ads / DCO, which Meta killed in April 2026). That is twelve combinations competing inside one post.

Why exactly three creatives when Meta lets you upload ten

Charley This is the single most-violated rule of the format.

“Remember that a 322 is already 12 ads. So if you load five or ten creatives, that could potentially be hundreds. Whatever gets lucky early is going to dominate everything — only because at your warmest audience, it was the thing people clicked on first. And this is why a lot of folks say ‘flex ads don't work, dynamic creative is stupid’ — because they overwhelm the machine.”

Charley T · 2026-07-18

With three creatives, Meta can make a real decision quickly and consolidate spend around the best one, which means all your spend is making that decision better. Either you have a winner or you kill it and move on.

Writing the copy

Charley The structural facts that dictate how you write:

The above/below-the-fold play: treat primary text as a conversation. What shows above “see more” is the hello; what sits below it is the close. If a detail would drive some people away but convert the right ones, put it below the fold deliberately. Charley used to put a free Facebook-group link below the fold and says it built communities of over half a million people.

Voice: the ad should sound like you if you're claiming authority, or like them if you want the customer to see themselves. Ads that do each job, running together, work well as a pair. His hard rule: “Do not sell to yourself. It's the single biggest mistake everybody makes.”

07The Olympic Rings: five ads that do five jobs

Charley's answer to “what should the ads in my control ad set actually be?”. Five concepts, each with a defined role, deliberately connected. This is what real creative diversity looks like in practice.

Broad retargeting, not pixel-event retargeting

Charley Rings 4 and 5 are retargeting, but not the kind most people run — and this distinction is worth real money.

“Funnel-based retargeting chases people who already said no. Most cart abandoners and checkout droppers already decided not to buy… If somebody's abandoning cart on your site, they're probably abandoning cart on a couple of your competitors' sites too.”

Charley T · 2026-04-09

His arithmetic: of 1,000 people who engaged with you, maybe 100 abandon a cart and maybe 10 of those will convert. You can spend your whole budget on those 100 — 90 of whom will never buy, and all of whom are being spammed by every competitor, in a high-bid auction — or you can reach all 1,000 interested people including the same 10, for less money, and probably get 12 or 13 sales instead of 10.

Broad retargeting also compounds: the pool grows every day from your organic reach and content without you raising budget, so it “never runs out of high-quality, ready-to-buy customers.”

08The ad sequence, slot by slot

Becker publishes his actual template — the one he cookie-cutters across every buyer profile at Hyros. It is the most concrete “what ads do I make” answer in either source, so it's reproduced here in full.

09Creative enhancements: the traffic light

Meta bundles genuinely useful delivery optimisations together with features that silently rewrite your marketing. Charley's dedicated guide sorts all eleven core enhancements into three buckets. This is the highest-value-per-minute section in the manual — most accounts are leaking money here without knowing.

Why each red is red

Why each yellow is conditional

• Music — fine on a silent product video. Not fine if you chose music or have a voiceover; the choices are dynamic and can be badly off. Charley's example: a friend's pet-urine-cleaner ad ran in India with “some of the most offensively generic AI music possible”, and wasted a lot of money.
• Image expansion — brilliant on a product on a clean white background. Destructive if your product touches the frame edge or the image is busy: the AI invents the missing pixels and you get warped product and “strange third-hand stuff in the backgrounds.”
• Text improvements — different from text generation, and easily confused. Meta can move your primary text into the headline and vice versa, rewrite sections, or change the emphasis of your message. Set up “restricted words” before you tick this box — it tells Meta which words, phrases and claims it may not generate or move.

Four traps that catch almost everyone

10Meta's native AI creative — and why it beats your prompt

Charley's most actionable near-term edge, and the one with the clearest mechanism: Meta's in-platform generation isn't just an image tool, it is a tool with access to what is already converting in your niche.

“It's not just learning from your own account. It's referencing what's already converting across your entire niche. Instead of guessing which angle is going to land, the AI already has a read on what's actually working inside your market right now. That is a wildly unfair advantage.”

Charley T · 2026-07-07

Inside the generation tab you get a “popular in your niche” section and a performance-based section sorted by return on ad spend — built from what everybody else running ads in your category is winning with. Charley's stated best practice: take one variation from each section and “essentially be able to print three 322 ads from right here alone.”

Becker's creative testing: vary the presentation, never the pitch

Becker A materially different — and complementary — use of AI video. His rule is that mass generation is the mistake; the win is sniper-placed variation of a proven winner.

11Reading the data: the 4PI analysis

Four metrics, read together, tell you what job Meta has assigned each of your ads — and therefore which ad to kill and which gap to fill. Charley calls this turning Ads Manager into a decision engine. It replaces staring at ROAS columns.

Charley The four are spend, frequency, CPM, cost per result (§1 explains what each means mechanically). You read every ad's four numbers relative to the campaign average, and the resulting signature identifies the ad's role.

What the analysis tells you to do — only two outcomes

12The profit metric — and its lead-gen translation

Charley's replacement for ROAS is the smallest change in this manual with the largest consequence. It is also the piece that needs the most careful translation for lead generation, because a lead has no immediate revenue.

Gross profit per transaction (GPT)

Charley ROAS is a ratio. It can look excellent while you make less money. GPT is the money.

Charley Only four numbers are needed: what you spent, what it cost to get a sale, what that sale was worth, and how much profit you made. “For every ad you look at, stop asking ‘what's the ROAS?' and start asking ‘how much profit did this make?'”

The one question that governs everything

Charley Every framework in this manual reduces to a single test, and he repeats it in four separate videos:

How to judge any test

Charley Not by CPA, ROAS, CTR or hook rate — those are diagnostics. Did the campaign's total profit volume go up? Total revenue minus total ad spend. If yes, you created more money to reinvest, so scale the budget. If no, the test is a loser. That is the whole judgement.

“It doesn't matter if any of your new ads are winners or if your ROAS goes from two to twenty. If profit didn't go up, it's a loss. Only the business outcome matters, because it's not about any individual ad — it's about the whole team.”

Charley T · 2026-05-04

Hunters and farmers

Charley The framing that explains why operators resist all of this. Hunters measure success by ROAS and daily CPA — did this ad work today? They live and die by short-term swings. Farmers focus on total profit and growth: build a system, protect it, and scale it with the resources the system generates for itself.

His argument for why farming compounds: one corn plant produces three or four pounds of food, and from what you don't eat you can plant ten more. Same work, but over time you go from four pounds to four hundred — because you invested in the system rather than the harvest.

13The testing loop

Creative testing has one purpose, and it is not finding winners. “The purpose of creative testing is not to find a winner. It's to build the best team.”

One variable at a time

Charley When the answer to “can I spend more?” is no, you run the scientific method, not a brainstorm. Start with the control — the system you know works, even if not well enough. Then change one variable until it does: a new hook, a new testimonial, a different objection handled.

“Don't change three things at once because you'll never know what actually fixed the problem.” And the target is modest by design: you are not looking for a perfect ad, you are looking for one small improvement that gets you back to “yes, I can spend more money” without performance breaking.

14Scaling: three stages, in order

“Why is it that every time you increase the budget, your performance tanks? Because scaling isn't one move, it's a system, and there's a specific order it has to happen in.” Most operators skip straight to stage three.

“Think of it like building a house. You don't put the roof on before the walls are up — but that's exactly what most advertisers do. They go straight to the budget before the foundation is even there.”

Charley T · 2026-07-18

Stage 1 — Efficiency (never budget)

Charley Growth does not come from spending more. Adding spend without adding profit is not leverage. The question is how do we make every dollar produce more?

Stage 2 — The readiness check

Charley “Feeling profitable and being ready to scale are two very different things. Budget amplifies what exists, so you need to know exactly what you're amplifying before you touch it.” Three gates, all must pass:

Stage 3 — Budget, using one of three methods

Charley Choose by answering three questions: how stable is the system, how confident are you in the performance, and how aggressive do you want to be?

Charley Use automated rules for all three. Linear and fractional get you to the next level, but eventually results plateau and you have to turn those rules off — which is when you move to marginal.

Scaling down without killing your winners

Charley Your CPA jumps 30% overnight and every instinct says fix everything. Three steps, in order:

15The signal layer — where lead gen is won or lost

Both operators converge hard here, and for lead generation this is the single highest-leverage section in the manual. When the machine controls targeting, the only lever left is the quality of what you feed it.

“The more control Meta takes over the mechanics, the more important your inputs become. Your data, your exclusions, your customer definitions, your measurement and your economics. The machine is going to get better and better at showing the right ad to the right person, but it can't fix a bad business.”

Charley T · 2026-06-28

Optimise for the outcome, not the event

Both The most important sentence in this section, from two directions:

• Charley “Standard events are for entry-level marketers. Optimise for custom conversion events that represent your most valuable customers.”
• Becker “I push back to people that actually attend calls and are verified by our team as qualified leads — not just the form fill, which a lot of people do… If you start optimising for people that show up for calls, you're going to get so many more qualified calls in.”

Front-loading: the predictive-value shortcut

Becker The problem with optimising toward closed business is time. A year of value data takes a year. His answer is to predict it from early behaviour and feed the prediction in immediately.

“You can look at the type of mobile phone, the type of information the customer's doing, the thing they first buy, and then predict their total value over the next year. Then they feed that total value into Meta and say optimise for that… To get that data, a year's worth of data would take a year — that's impossible. But if you're able to front-load all that data in, it gets so much better at targeting right away.”

Alex Becker · 2026-08-07

His simpler variant for lower volume: forget predicted dollar values and make it binary. A custom event called high-value customer, fired only for profiles predicted to exceed a threshold. “You only push that back as a conversion.”

Conversion goal is not attribution

Becker The reframe that makes the rest of his advice make sense — and the one most likely to feel wrong at first.

“Your conversion goals inside your account are not your attribution… They're using it as a data source. So even if Google got the last click, they make sure to go in and find the last click that was associated with Facebook and still slap it on there. That's not accurate reporting — that's why you do your reporting elsewhere.”

Alex Becker · 2026-08-07

The logic: the pixel's job is to train targeting. If Meta genuinely influenced a conversion but didn't get the last click — someone saw four Facebook ads, then searched your brand and converted through Google — then withholding that event from Meta teaches the algorithm that its ad failed. You are training it on a lie of omission. His example is his own: people arrive from his YouTube videos carrying Facebook clicks in their profile that Meta never attributes.

Customer lifecycle strategy and exclusions

Charley An ad-set-level setting — “and the fact that it lives here is the whole point. This isn't a creative tweak… it's a delivery strategy decision.” Selecting acquire new customers is only step one; the exclusions do the real work.

When to actually turn lifecycle optimisation on

Charley His spend tiers, verbatim in substance. Note how conservative he is — this is a rare case of an operator telling you not to use the new feature.

The gate that overrides the table: “If you can't measure new customers today, don't turn this on yet. Fix measurement first.”

16Worked example: an NZ life-cover lead-gen system

The scenario: a small NZ life-cover advisory. Revenue is commission on policies issued and staying in force. There is no e-commerce transaction, no AOV, no ROAS that means anything. This is the hardest case for everything above — which is why it is the useful one.

Step 1 — Build the offer before the account Becker

Run the five-part test. What most advisories actually run — "Get a free life insurance quote. We compare NZ's leading insurers." — fails four of the five.

Step 2 — Define buyer profiles by problem, not demographic Becker

Applying the §5 decision rule: a genuinely different problem, needing a genuinely different landing page.

Step 3 — The launch architecture

Step 4 — The economics, shaped correctly

Applied This is the calculation the whole manual exists to enable. All figures below are invented.

Step 5 — What to optimise for, by spend level

Applied Combining Charley's volume caution with Becker's climb-the-ladder push. The constraint is events per week: too few and the machine cannot learn.

17The 90-day build order

Everything above, sequenced. The order is derived from the sources' own dependency logic — measurement before optimisation, offer before ads, stability before budget.

18Where they disagree — and where each contradicts himself

A guide that presents two operators as one voice is lying to you. Seven genuine tensions surfaced across the ten transcripts. Four matter enough to change what you build.

Three lesser tensions, noted for completeness

• Becker vs Becker on whether the account matters. 30 July: “nothing in the ad account that you do matters.” 7 August: an entire master course on account structure. The first is deliberate rhetoric to force attention onto the offer; read together, his position is “offer dominates, but structure is still worth getting right.”
• Signal accuracy. Becker deliberately attributes conversions to Meta that Meta did not earn on last click — “that's not accurate reporting” — because the pixel is a training input. Charley stresses accurate truth and clean definitions. These reconcile if you keep two sets of books (§15), but the postures are genuinely different: Becker is willing to feed the pixel a useful fiction; Charley is not explicit about that.
• Retargeting mechanics. Charley builds explicit broad-retargeting campaigns against seed audiences in April, then folds retargeting into rings 4–5 inside one campaign by May. Becker has no retargeting layer at all — “the retargeting is done inside the system itself.” The May-onward Charley and Becker agree; the April Charley doesn't.

19What neither of them covers

The honest boundary of this manual. These are questions you will hit in a lead-gen account that ten transcripts simply do not answer — do not let the guide's confidence elsewhere imply coverage here.

20Source ledger

Every load-bearing claim in this manual, traced. Use this to check anything that felt too confident.