You're shipping twenty ads. Meta might be seeing three.
Meta's delivery engine now reads the ad, not just the audience — and it quietly groups whatever looks alike. This is the production system for the new rules: what to make, how much, how different, and how to know it's working. Written for founders who already know that good creative and good media are the same problem.
The 60-second version ↓Most accounts ship twenty "new" ads a month. Meta sees three. The ten stages below explain why — and what to do about it.
1. Meta's Andromeda update made creative the targeting. Volume, variety, and velocity now decide who sees you — measured on profit, not platform ROAS.
2. Brands that grow profitably start from one of two home bases — community-led or craft-led. The compound brands master one, then layer the other.
3. The failure mode is the Balenciaga Trap: beautiful ads that never say why to buy.
4. The engine is a seven-axis production matrix. Most "new" ads move the wrong axes — the machine literally can't tell them apart. Stage 06 shows which axes count.
5. New concept families ship in bi-weekly sprints, scaled to spend — with controlled variants, winner iterations, and launch assets running alongside as the data calls for them.
6. Stage 08 shows the seven leaks we find most often. The audit maps your account against all seven axes — before we've even spoken.
The receipts: one established ecommerce account's first 100 days under a stronger creative operating model — revenue up 73% on just 12% more spend, ROAS from 2.95 to 4.56, cost per purchase down 25%, and conversion rate up 62%, all versus the equivalent preceding period. Verified in the commercial reporting, not just Ads Manager.
It didn't take every format in this field note. It took the right expansion for its spend, its product, and its constraint.
That's the system. The playbook below is the how behind each line — ten stages, about twenty-five minutes.
Meta's Andromeda update rewrote how the platform decides who sees which ad.
Old system: you told the algorithm who to reach. It found them. Creative was a lever you pulled inside a defined audience.
New system: the algorithm reads the creative itself — the visual, the hook, the frame, the motion — and decides who it fits. Audience targeting became a signal source, not the driver. Creative became the routing mechanism.
The consequence is simple. Creative isn't part of your strategy anymore. Creative is your strategy.
Distinct concept families produced per month.
How differently they move across the axes of production.
How fast you make, test, and retire.
Below these live the reading signals. Profit first — real, blended, tracked to the dollar — as the only truth. Then CAC and revenue as the operating reads, with thumb-stop, hold rate, and CTR as early signals only. And profit gets read per ad, on each product's actual margins — two ads at the same ROAS can be a winner and a loser once margin enters. One more reason we treat platform ROAS as debug-only — too noisy post-iOS to make spend decisions on.
Brands still running the 2022 playbook — narrow audiences, low creative volume, watching platform ROAS — are eighteen months behind. Their CPMs are climbing. They're calling it a Meta problem.
It's a creative problem. And the cause is mechanical — something specific Meta's delivery system does to ads that look alike. That's Stage 06. The stages before it earn the context.
Now you know what the platform rewards. Whether your brand is built to earn it — that depends on which archetype you fit.
Every brand we've seen grow profitably starts from one of two home bases. Neither is wrong. What doesn't work is drifting between them — half a community engine, half a craft brand, fully neither.
Win on discovery and belonging — the "I've never seen this" moment combined with the tribal "this is for people like me" signal.
Founder in feed regularly, cultural moments picked up quickly, drop reveals framed as events, community imagery of who this is for, editorial-friendly and culturally referential.
Ganni, Djerf Avenue, Gymshark, Peachy Den, ARMEDANGELS.
Viral potential, tribal loyalty, low CAC inside a tight audience.
Craft evidence. There's rarely a clear reason to trust the price. Which is why community-led brands tend to get cut first when discretionary spend tightens — the buyer loves the brand but can't defend the price when money gets tight. Growth also caps when the audience saturates.
Injecting craft — materials stories, construction detail, longevity narrative. The moment a community-led brand earns "trust what you're paying for" alongside "this is for me," they stop being a moment and start being a staple.
Win on craft and aspiration — the trust foundation combined with the "upgraded me" narrative.
Cinematic product-forward, materials stories, atelier close-ups, professional context, understated aesthetic, editorial polish.
Buck Mason, With Nothing Underneath, Aesop, Toteme, Vollebak.
Pricing power, buyer loyalty, brand equity that compounds year over year.
Discovery and belonging. The ads read serious and considered — which works for buyers who already know what they want but doesn't earn attention from a cold audience scrolling. Growth caps at the "already sophisticated" segment.
Injecting discovery or belonging — founder POV, community imagery, cultural moment content. The moment a craft-led brand earns cultural energy alongside its craft foundation, it starts converting audiences that would never have found it on merit alone.
The best brands in any category — think Ganni at their peak, or Aimé Leon Dore in full stride — run both levers at volume. Community energy and craft conviction, live in the same account. That's the destination, and it's where this whole system points.
But none of them started there. Every compound brand we can name mastered one home base first, became unmistakable for it, then layered the missing lever deliberately. Both-at-once from day one isn't a strategy — it's drift, and drift is what the next stage is about.
So the question isn't "which archetype are you forever." It's "which is your home base — and have you earned the right to layer yet?"
Your choice weights the examples in the stages ahead toward your home base. If you're not sure, we'll show you both.
You've placed yourself. Now the anti-pattern that swallows brands who don't commit.
The third pattern isn't a pattern. It's a trap.
The trap is aesthetic-only content across the entire account. Beautiful. Moody. Cinematic. And nothing that gives the buyer a reason to actually buy. No clear signal about what makes the brand different, who it's for, or why it's worth the price. The ads pass the scroll test — they're pleasant to look at — and fail the wallet test. The buyer doesn't act.
Balenciaga can run this way. Loewe can. Bottega can. They've spent decades building brand equity that lets them communicate through ambiguity — the audience already knows why to buy. At their scale, aesthetic ambiguity is a form of authority.
At a premium price without that equity, aesthetic ambiguity becomes expensive art. The ads look great. The profit erodes. The founder blames the algorithm.
There's research behind why this fails. A 2012 Journal of Marketing Research study of nearly 7,000 New York Times articles found that high-arousal emotions — awe, anger, anxiety, amusement — drove sharing and reaction, while low-arousal states suppressed it, however pleasant they felt. Beautiful-but-ambiguous sits in the worst quadrant: mildly positive, entirely passive. The viewer feels nothing strongly enough to act on. And an ad nobody reacts to gives the delivery system nothing to learn from.
You can diagnose the trap by scrolling any brand's Meta ads and trying to complete the sentence: "I should buy this because ______." If you can't finish the sentence in five seconds, the buyer can't either. And the ads aren't going to convert regardless of budget, targeting, or bidding strategy.
Looks like art. Performs like a screensaver.
You know the trap. Now the framework that keeps you out of it.
Assets built for no funnel stage default to aesthetic filler. They burn spend without earning attention or conversion. The framework is simple — every asset gets built for one specific stage in the buyer's journey.
Who: cold audiences who don't know you exist.
Content job: capture attention in two seconds, trigger curiosity in the next five.
Format bias: motion-heavy, scroll-stop optimised, short.
Who: aware of you, weighing alternatives.
Content job: answer "why you" through craft evidence, founder story, or category conviction.
Format bias: mixed — narrative motion for story, statics for proof.
Who: near-decision, needs final proof — including warm traffic coming back.
Content job: compress remaining friction into commitment. For returning visitors: a fresh angle on the item they viewed, never a guilt repeat.
Format bias: reviews compilations, product demos, direct offers.
Re-engagement isn't a separate stage — it's MOF and BOF applied to people who already met you. Same psychology, warmer entry.
Missing stages leak spend. All-TOF accounts feed the top and never close. All-BOF accounts convert the already-warm and never grow. Neither is winning.
The funnel tells you which stage every asset serves. Now the assets themselves.
A running creative system has three tiers. The workhorses recur. The rotation varies. The heavy-lifts anchor. How much ships, and when, is Stage 07's job — this stage is the menu.
The always-on core. These carry the buyer's decision and anchor the funnel — every running account keeps them in play.
Six types. The spine of the account.
The variety pool — where range comes from. Cycle these through sprints against the gaps the diagnostic finds.
Twenty-six types. If your account has been recycling the same four, this list is the fix.
Serious production, months of use. Synced to drops, launches, and brand moments.
These are what most brands never produce — and what the leading accounts consistently do.
You've got the palette. Now the mechanism underneath every winning account.
Meta's delivery system builds a representation of every ad it runs — what the industry has come to call a creative fingerprint, or the Entity ID. Meta doesn't publish the spec, but the behaviour is consistent enough to build a system on: the fingerprint reads what the ad looks like, weighted heavily toward the opening seconds. Ads that look alike get treated alike — one set of learnings, one pool of reach.
Run enough accounts and the consequence is unmistakable. Headline tweaks and copy swaps buy you nothing — the fingerprint barely reads them. If it looks the same, it is the same, and an account's "twenty new ads" quietly collapse into three ideas the system already knows. No new learning. No new audiences. No new reach.
New reach is earned by genuine newness. Which raises the real operating question: how do you produce genuinely new on schedule, sprint after sprint, without descending into randomness?
Our answer is the Diversification Matrix — seven production axes. Every concept occupies one position on each.
In the Maxout Matrix, the seven axes split into two kinds — the axes that change what the ad structurally is, and the axes that change what it says to the person. The split is the point.
Five are machine-facing. Format, hook, POV, duration, and aesthetic register are what the visual fingerprint reads. Moving them is what makes an execution register as materially distinct — a genuinely new opportunity for retrieval, learning, and reach.
Two are human-facing. Copy voice and buyer lever shape how the person on the other side responds once the ad reaches them. The machine can't see them directly — but it reads the humans' reactions, so these axes steer delivery too, one step removed.
A concept can be new to the machine and stale to the human. It can be new to the human and invisible to the machine. The accounts that scale move both.
Our production rule: every new concept moves three or more axes — always including at least one machine-facing axis.
The rule defines how different a net-new concept must be — not how many concepts every account should produce. Breadth scales with spend, maturity, and the amount of clean learning the budget can support. Stage 07 sizes it.
The seventh axis is the psychological driver — what part of the buyer's brain the ad speaks to. We use an internal framework called the Motive Compass to map every ad to one of four levers:
"I've never seen this before"
"This makes me the person I want to be"
"I can trust what I'm paying for"
"This is for people like me"
Every high-performing ad plays at least one lever deliberately. The brands we see grow fastest layer two together. Discovery alone can go viral but doesn't convert reliably. Craft alone earns loyalty but doesn't grow the audience. Discovery paired with belonging, or craft paired with aspiration, satisfies more of the buyer's decision in a single ad — and reaches segments a single-lever ad leaves untouched.
The single most common failure we see is playing one lever across every asset. It caps growth invisibly.
A concept — creator to camera, direct address, standard duration, brand-polished, community-insider voice, belonging lever — is one fingerprint.
Changes only the copy voice. The visual fingerprint doesn't move.
Functionally the same creative. Little new signal, little new learning — and usually the same performance ceiling.
Faceless product-only, curiosity-gap hook, narrative duration, editorial register, educational voice, craft lever — six axes moved, four of them machine-facing.
Materially new creative. A new opportunity for retrieval and learning — a different buyer reached, for a different reason.
You understand the mechanism. Now the operating rhythm that runs it.
The right cadence depends on spend, growth goals, and account maturity. There isn't a single number — there's a band. But before the band, three words this stage uses precisely:
| Monthly ad spend | Net-new concept families | Total executions | Operating priority |
|---|---|---|---|
| £5k–£15k | 2–4 per sprint | 6–12 | Attack the largest concentration first |
| £15k–£30k | 4–8 per sprint | 12–24 | Expand formats and buyer levers |
| £30k+ | 8+ per sprint | 24+ | Broad, continuous coverage |
A sprint is each two-week cycle. These are typical starting ranges, not fixed production quotas — cadence adjusts with account signal, maturity, and production complexity.
The family count covers net-new concept families only — each fresh across three or more axes. The execution total is what those families ship as. Controlled variants on proven winners, element-level split tests, and drop or launch assets run alongside — triggered by the data, not the calendar. The sprint keeps the account genuinely new. The testing layer keeps it learning.
Ship below your sprint floor and we see the same thing every time: Meta under-reads the variety, delivery narrows, CPMs climb quietly. Push above it without a mature production stack and quality burns — assets get sloppy, brand equity erodes.
The sprint keeps concepts new. The testing layer keeps them learning — and it runs on signal, not on a calendar slot. A winner worth extracting, a fatigue flag, a drop landing: that's what triggers a test. The single biggest misuse of the testing layer is treating it as "more different things." Shipping ten unrelated assets at once feels like variety. It's actually noise. You can't attribute performance to specific elements when every asset changes every variable.
The winning discipline: one or two content types per test, multiple variants of each, specific elements varied systematically.
Split testing at the element level. Each test isn't asking "did creative work?" — it's asking "which hook worked?" "Which opening frame?" "Which copy angle?" Every test ships with clean isolation of variables. Every test returns learnings you can compound.
When a concept family proves itself, it can be worked in two different ways — and confusing them is how accounts end up "diversifying" into noise or iterating into fatigue.
Creator POV motion · question hook · raw register · belonging lever — proven in the account
Same family, one variable moved at a time. More extracted from a proven idea — this is testing, and it never counts toward the sprint.
Three or more axes moved: an adjacent family, a different audience entry point. This is diversification — it counts toward the next sprint's family count.
Content type, persona, duration, and buyer lever held constant. Hook type and aesthetic setting varied. When results come in, you know exactly which hook and which setting converted.
Same discipline. Isolated variables. Clean read.
Across two sprints and the tests the data triggered between them, an account covers eight content types with split-test data on each:
By month's end, the account has earned distinct fingerprints across eight content types and multiple axis positions per type. That's the kind of variety that opens delivery to broader audiences. Every asset was part of a structured test that produced conclusions. Nothing sprayed. Nothing unattributable.
The two most reliable ceiling predictors we've found: every asset playing the same buyer lever, and every asset in the same format. The sprint-plus-test rhythm exists to keep you off both.
You've got the production plan. Now the diagnostic that finds what it should attack.
This is the diagnostic we run on every account we map — the same one behind the audit at the end of this page. Fifteen minutes in Meta Ad Library is usually all it takes:
These show up most often in the accounts we audit:
All ads TOF (no conversion push) or all BOF (no cold-audience feeding). Missing stages leak spend regardless of budget.
Ninety percent static, or ninety percent motion. In the accounts we run, format-diverse accounts reach broader audiences — monoculture accounts hit a ceiling and stay there.
Zero real faces, or one face only. Buyers need both a human presence and third-party validation — missing either half kills the funnel.
Every ad the same length, usually six to fifteen seconds. Missing long-form kills MOF conviction. Missing snap-length kills TOF scroll capture.
Every ad plays the same psychological driver. Delivery settles on one audience and stops expanding to new ones — growth caps invisibly.
Same hook style across the account. Redundancy reads as one idea, and delivery consolidates around it.
All editorial, no raw. Or all raw, no polished. Strong accounts need the mix — polished for brand equity, raw for native-feeling delivery.
Aesthetic-only content across the account. No clear buyer lever, no answer to "why should I buy?" Diagnosed by scrolling and being unable to complete the sentence.
The clusters are the gaps, and the gaps are where spend leaks. The system in Stages 05 through 07 exists to close them — and mapping them properly against a live account is exactly what the audit at the end of this page does.
You've seen the leaks. Now what closing them looks like at scale.
The gap between the average account and the ones compounding at scale isn't creative talent. It's system. And the goal isn't maximum variety — it's enough meaningful coverage to support profitable delivery at the account's current scale. A smaller account earns it with a handful of distinct concept families. A larger account needs substantially broader coverage.
Across the accounts we audit, these are the habits that separate the leading brands:
Founder, team, or creators — a recognisable person the audience builds a relationship with. That equity compounds and can never be bought. The average brand has a real person in maybe two or three ads a quarter.
Relevance treated as velocity, not planning cycles. The average brand plans two weeks out. The leaders publish within two days of a moment landing.
Materials, construction, sourcing — filmed like short-form documentary, not filmed like advertising. Educational framing, not selling framing.
Forty-five to ninety seconds, founder-led or creator-led. Almost no brands produce this. The gap is enormous — closing it is often the single highest-leverage move an account can make.
Raw authenticity plus Meta's native trust boost when the ad runs under a real user's handle.
Reels-native reels. Story-native stories. Not square repurposed into everything.
Sourcing, factory visits, design process — treated as content pillars, not one-off pieces.
"This didn't work, here's what we learned." Almost no fashion brands do this. The ones that do build enormous trust.
Customer-of-the-week, tag reposts, real-wear grids — content that's about the buyer, not the brand.
You've got the whole system. Last thing — what to do with it.
Everything above is real and usable. The install order:
Run just that and you'll be ahead of most of your category.
But we'd be underselling what produced the numbers at the top of this page if we called the creative system the whole machine. It isn't. It's the engine — and an engine needs the rest of the car:
Sprints only learn if spend is structured to test cleanly — consolidated structure, budget discipline, kill-and-promote executed without sentiment.
Every read in this system assumes you can see real, blended profit daily. Platform dashboards can't show it. Most brands can't either.
Weekly diagnostics that catch fatigue, waste, and emerging winners before the P&L feels them.
Creators, editors, and a production stack that makes sprint volume economical without burning quality.
That stack is what we run. And the work starts before we ever speak: book the audit, drop your Meta Ad Library link in the booking form, and we'll map your active creative across all seven axes before you arrive — your three largest concentrations, the buyer levers you're missing, your sprint sizing, and what the first ninety days look like. You turn up to an account that's already been mapped. About an hour. No obligation.
Book your creative audit →