More ads is almost never the constraint. We know that because we went and checked rather than assuming: hundreds of strategy and discovery calls, ranked by what was actually stopping each brand.
Almost none of it was an ad account problem. So we don't start there.
This process wasn't designed from best practice. It was reverse engineered from hundreds of strategy and discovery calls run across the life of the agency, then ranked by how often each blocker actually turned out to be the thing in the way.
The single most common ceiling. A brand can't scale into stock it doesn't have, and can't order stock without the cash the scale would produce.
Contribution profit, acquisition MER, CAC, LTV to CAC, breakeven, COGS. In one account we found a 58% cost of goods against a sub-30% target. Nobody had looked.
Worst case we've seen: $67k of add to cart value against $15k of conversions. Traffic was never the problem.
One account was about to kill a channel that was actually working, because the reporting said it had collapsed.
Too low to support paid acquisition at any sane cost per customer, no matter how good the ads get.
One account had 35 dormant flows and roughly ten thousand contacts nobody had emailed.
Strategist, buyer, creative director and customer service desk, all at once. The fifth job never gets done.
The brand has trained its list to wait, and every good month has a sale inside it.
So we start with the numbers, then the leaks, then the stock, then the list, and only then the account. Usually by the time we get to the ads, we already know what they have to do.
We start with the profit and loss, not the ad account.
Before we're ever on a call.
The first thirty days. We don't touch the ad account in week one.
The account gets rebuilt around one idea.
At least five ads into the account every week. Seven in ten build on a winner, three in ten are new.
We buy to a number, not to a feeling.
One report, one strategy call, every month.
Most creative starts with a brainstorm. Ours starts with evidence, at volume, before anyone writes a word of script. Four inputs, every account, every cycle.
Reviews, comments, replies, unboxing captions, the questions people ask before they buy. Not a summary of them, the actual words. We're looking for the phrases real buyers use for the product, the occasion and the feeling, because those phrases out-convert anything we would write ourselves.
Which angles actually carried the account, which creative is fatiguing, what has been tested in the last quarter and what never has. Concentration risk too: if two ads are carrying half your spend, that's a business risk, not a performance win.
A persona isn't a demographic or a mood board. It's a group of people who described the same problem or the same want in their own words, often enough that it's a pattern. If we can't point at where a persona came from, it doesn't get a brief.
We map the angles already running against the ones nobody has touched, then rank the gaps by how likely they're to land on someone who has never heard of you. That ranking decides what gets briefed next, and it's the reason the batch is never a guess.
Out of that comes one thing: a ranked list of what to say next, and who to say it to. Nothing gets briefed because it's someone's turn.
Who it's for, what we're claiming to them, and what we're asking them to take. Format and hook are execution. They're never a substitute for the concept.
An account with two hundred ads and four concepts has four ads.
First. The message is the biggest lever in the account, by a distance.
Second. What they get changes conversion more than how it looks.
Third. Who we point it at, once we know what lands.
Last. The cheapest thing to change and the least likely reason an ad failed.
Most accounts run that in reverse, and spend the whole budget testing edits of a concept that was never going to work.
You no longer choose the audience in the settings, you choose it with the ad. Which makes the persona and the angle the most important decisions in the account, and the audience checkbox close to irrelevant.
Five edits of one concept give the system one signal to learn from. Five properly different concepts give it five. That's the whole reason we cap format repeats inside a batch.
One angle or one product per ad set, minimum four ads in it. Otherwise the result tells you the set worked, without telling you what in it did.
When frequency climbs and click through falls, no bid strategy saves it. Only a message the audience hasn't seen yet. That's why concept volume is a formula rather than a mood.
Watch the reach curve on any ad and it flattens fast. In the accounts we run, most of the new people an ad will ever reach have seen it inside about four days. After that you're buying frequency rather than reach: the same people, the same ad, a worse result every week it keeps running. That's why the calendar is weekly and not monthly.
About seventy percent of a week's ads build on something that already won, which sounds like repetition and isn't. Take the hook off a video that worked and rebuild it as a static, and the proven message reaches the people who scrolled past the video. Re-run the video with a new headline and it reaches nobody new. The message gets reused. The format has to change.
Competitor research earns its place, but only at the format level: how the ad is built, what happens in the first second, where the proof sits. The angle is theirs, chosen for their customer, their price and their margin. Lift it and you have an ad arguing somebody else's case to your audience.
Five new ads, minimum
Five new ads, minimum
Five new ads, minimum
Five new ads, minimum
It goes up with spend and with how fast winners are showing up. It doesn't go down.
Twelve briefs a fortnight, coded and filed. Each one is production ready before it leaves us, because a vague brief is just a decision passed to whoever is cheapest.
A hook isn't a line of text. It's the on-screen words, the audio, the frame-zero visual and the vibe, all landing together. Five variants are written for every brief and they're tested against the same body, which isolates the hook as the only variable.
The test it has to survive is the three second scan: is this for me, is this something I recognise, is this worth my time.
Six dimensions, one to ten: clarity, relevance, desire or hope, proof, call to action, and visual direction. The average has to hit seven. Anything still under six after a revision goes to a human rather than shipping.
Fashion and lifestyle runs desire-led, fifteen to thirty seconds, built on identity and belonging. Problem-solution runs thirty to sixty, built on the customer's verbatim language and real proof. We don't run a pain script on a fashion brand.
Plenty of agencies either hide it or sell it as the whole product. Both are a mistake. Used properly it doesn't replace the thinking, it removes the ceiling on how much of it we can do for you, and how fast.
The machine does the volume. The judgement stays human.
And one line we won't cross: we can generate the model, but never the garment. Fit, fabric and colour always come from your real product photography, because a reinvented product costs returns and it costs trust.
Most ads open in the middle, listing what a thing is made of, and nobody feels seen enough to stop. So we start where the customer actually is, which is usually somewhere low: nothing in the wardrobe feels right, the third cream hasn't worked, it's 3am and the room is too hot again.
Before anything is written we put the feeling in one sentence. If it doesn't fit in one sentence the concept isn't ready to test, and no amount of production will save it.
The opening line is lifted from reviews, comments and recorded calls, not invented by us. A customer recognises their own sentence long before they recognise a clever one.
The last line is what staying put costs. An ad that only shows the better life gives nobody a reason to move today, so the drop at the end does more work than the climb.
One line we hold to: we never tell your customer how they feel. It's written in the first person or as something they said themselves, never as an accusation pointed at them. That's partly Meta policy and mostly because being told how you feel by a brand is the fastest way to lose someone.
Not a taste check. A build that fails, hard, on any one of these. The thresholds are not arbitrary either: the contrast and dead space rules were set against a specific ad that lost money and failed both.
Primary text has to run the whole arc inside thirty words: where they start, where they could be, what staying put costs
Must be 4:5, 1:1 or 9:16, nothing else ships
No empty band larger than 14% of the frame
The primary message has to sit in the top 25%
Median copy contrast at least 55 of 255 luminance separation
Can't be left empty
Every product is checked against live availability before it can enter an ad
The stock check is the one clients notice. We don't run ads for things you can't ship.
Every asset is named with its angle, persona and format, and that's exactly what the ad is called in the platform. Reporting reads back at angle level forever, without anyone maintaining a spreadsheet. It's why we can tell you which message is carrying the account rather than which file.
Concept yes isn't build yes, and build yes isn't live yes. Each one is separate, explicit and recorded, and Amber is the final reviewer on every asset. Nothing reaches a live account without her sign-off on the record.
Every method we use is a written, versioned file rather than something in someone's head, and those files get reviewed against what actually happened. The way we brief an ad today is measurably better than a year ago because it was captured, not remembered.
Every client gets a formal written review each quarter. Not a check-in, a scorecard on us. Rank what matters most right now, score us out of ten on delivery, speed, communication and strategic direction, then tell us what to do more of, what to do less of, and anything you've paid for but not received.
How likely are you to continue next quarter, one to ten?
A six or below fires an internal alert the same day. There's no lock-in holding you here, so the answer has to be earned. We built a system that can't ignore it.
No lock-in theatre either. You can leave at the end of any month and every asset we made is yours in perpetuity.
A guarantee is what an agency offers when leaving is hard. Leaving isn't hard here, so there's nothing to be released from.
Not a performance figure, not a review, not a claim. Every figure traces to a dated pull or it doesn't get written.
Real and sourced, or the space stays empty until it is.
It's checked automatically, before the ad is even built.
It's on the application form, in writing, before anyone gets on a call.
We pull your account live, show you where the money is actually going, and tell you the three things we'd change in the first fortnight. No deck, no discovery process, just a straight answer.
The process above exists to make this, every week, in every account.
