The Creative Flywheel: Our Repeatable System for Manufacturing Winning Meta Ads

The thing everyone repeats is that winning ads come from creative geniuses having brilliant ideas. Hire the right wizard, brief them well, and the bangers will follow.

In reality, the best-performing accounts I see are run by people who couldn't draw a stick figure. They don't make winning ads. They run a system that produces them, week after week, whether anyone feels inspired that morning or not.

That's the reframe I want to walk you through. Winners aren't a stroke of luck you wait around for. They're an output of a process you can document, hand to a team, and run across ten accounts at once. We call ours the creative flywheel, and the whole point of a flywheel is that once it's spinning, each turn makes the next one easier.

Let me show you the exact loop, the ten steps we actually run, and the two or three places our version deliberately breaks from the popular one.

Why a flywheel beats a fountain of ideas

Most creative "strategy" is really just a fountain. Someone turns it on, ideas spray out, you pick the ones that look nice, and you hope. When the fountain runs dry you panic and turn it on harder.

A flywheel is different because it feeds itself. The ads you launch this fortnight generate the data that tells you what to make next fortnight. Your winners become the raw material for your next round of winners. Nothing starts from a blank page, ever.

Here's my take on why this matters more than any single creative trick. A clever ad is worth something once. A system that reliably finds clever ads is worth something every week for years. I'd take the boring repeatable machine over the occasional flash of genius every single time, because the machine is the thing you can actually scale and the genius isn't.

So the flywheel has a fixed shape: source, build, launch, wait, analyse, then iterate off what won and back to the top. Let me break that into the ten steps we run.

The ten steps, start to finish

1. Source from competitors first. This is where most new concepts come from, and it's the step people skip because it feels like cheating. It isn't. Go into the Meta ad library, or a paid tool if you've got one, and pull the ads your competitors have been running the longest. Longevity is the signal. If a rival has run the same concept for 300-plus days, they're not doing that for fun, it's paying for itself. One honest caution: don't study the giant in your category. If you're a ~$2m/year brand, the ~$200m brand's ads are solving a different problem to yours. Study the brand one or two rungs above you, not ten.

2. Source from your own winners second. Your account already knows things. The ads that have spent the most over the last 30 days, sitting above your return target, are telling you which angles your specific audience responds to. Those are concepts to mine, not just ads to keep running.

3. Pull the threads, not the screenshots. The mistake here is copying a single ad. What you actually want is the common thread running through the winners, the underlying angle. A clean white background, a "did you know" opener, a founder talking to camera, whatever keeps recurring. One good thread can spin out into dozens of executions. You're collecting concepts, not stealing posts.

4. Write the brief before anyone touches a camera. This is where we run our APAP briefs: Audience, Painpoint, Angle, Product. Who is this for, what's bugging them, what's the angle that lands, and how does the product resolve it. Every concept gets one before it gets made. If you can't fill out those four lines clearly, the ad isn't ready to build and no amount of nice footage will rescue it.

5. Build in variations, not one-offs. When you produce a concept, produce it as a small family. Same angle, three or four different openings or framings. You're not trying to guess the one perfect version, you're giving the algorithm a few shots at the same idea so it can find the angle that connects.

6. Launch into a testing campaign built for competition. Every fresh batch goes into one prospecting campaign, and every new batch gets its own ad set. Four to eight creatives per set is the sweet spot for us, rarely more. Over time you build up a long stack of test ad sets, and that's exactly what you want, controlled internal competition rather than ten scattered campaigns all starving each other of data.

7. Force the spend, then wait. Here's the step that makes the waiting bearable. New ads usually can't get oxygen because your proven ads hog the budget. So at the ad set level we set a minimum daily spend roughly equal to one times the target cost per acquisition. If your target CPA is ~$45, the ad set is forced to spend about that much a day. That guarantees each test gets a real shot inside a few days instead of dribbling along for three weeks. Then you genuinely do wait. No poking it, no judging it on day one.

8. Pull the minimum at the seven-day mark. This is non-negotiable, set an alarm. After seven days you go back and lift that minimum spend. If the ad's a dud, it drops back to spending almost nothing on its own. If it's a winner, it keeps spending well above the floor without you forcing it. Either way you've learned what you needed in a week, not a month. Leaving these minimums on forever is how people quietly bleed budget into losers.

9. Analyse with honest attribution. Now the real work. Sort the account by spend, highest first, and read it through a sensible attribution setting rather than the flattering in-platform default. You're hunting for one thing: high-spending ads that are sitting above your return threshold. Not the tiny-spend 8x that only ever spent $40, the ad pulling real volume at a profitable number. That's a true winner. The little high-multiple ads are nice but they can't carry an account.

10. Iterate off the winners and feed the top. Take those proven, high-spend winners and treat them exactly like you treated your competitors' ads in step one. Variants, fresh openings, the same angle on a different product, two winning concepts merged into a third. These go back in at the top of the flywheel as your next batch. And as the account grows, let the spend grow with it. The flywheel is a growth system, not a maintenance one, so don't get shy about the learning phase or a rising budget when the economics hold.

That's the full turn. Source, brief, build, launch, force, wait, pull, analyse, iterate. Then around again.

Where our version breaks from the popular one

The loop above will look familiar if you've watched the usual creator videos. Here's where we deliberately do it differently, because a few of the popular defaults quietly cost you money.

We don't chase only the highest multiple. The old instinct is to find your 8x ad and pour everything into it. The trouble is your 8x ad often can't absorb much spend before it falls over. We'd rather run a stable of concepts at different multiples, a 2.2x doing real volume sitting happily next to a 4x doing less. As long as each one is profitable, each one adds to the bank account. Profit in dollars beats a pretty multiple on a tiny spend, every time.

We brief before we build, every time. Plenty of versions of this jump straight from "make a heap of ads" to "throw them up and see". We force the APAP brief in between. It's the cheapest filter we have. A bad concept caught on paper costs nothing. The same bad concept caught after a shoot and a week of spend costs real money.

We treat the seven-day rule as a hard stop, not a vibe. The forced-minimum-then-pull mechanic only works if you actually go back and pull it. The number of accounts I audit where someone set the minimums and forgot is genuinely high, and it's a slow leak that never shows up as a single obvious problem. Discipline on the boring step is most of the edge here.

We keep losers documented, not just deleted. Every dead concept goes into a log with a line on why we think it died. A dud isn't a waste, it's a data point that narrows the next brief. Over a few months that log becomes worth more than most of the winners.

How this runs across many accounts at once

The reason I trust a flywheel over individual brilliance is that brilliance doesn't scale and a documented loop does. When the ten steps are written down as actual SOPs, a media buyer can run the same disciplined process on account number eight without reinventing anything.

The cadence we hold is steady rather than heroic. A handful of fresh concepts per account per week, every account on the same loop, quality kept high by keeping the volume sane. I'd far rather run six well-briefed concepts a week than forty rushed ones, because the win rate on the six is dramatically higher and the whole system stays legible. Forty random ads a week is just a fountain again, sprayed faster.

The flywheel is what lets a small team hold standards across a lot of accounts at the same time. Not because anyone's a genius that week. Because the system doesn't need them to be.

Where to from here

If you've been treating creative like a fountain you turn on harder whenever results dip, the flywheel is the shift worth making. The ads stop being a gamble and start being an output.

If you want a second pair of eyes on whether your loop is actually spinning or just spraying, that's a chunk of what we do in a Signal/Noise Audit. We look at your creative history, what's genuinely winning at scale versus what just looks good on a tiny budget, and where the next round of concepts should come from. Often the fix isn't more ads, it's tightening one or two steps you're already half-doing.

So here's the question I'd sit with: of the ten steps above, which one is your account quietly skipping right now?