TL;DR
Our creative framework has six stages: mine, concept, produce, test, iterate, scale. Each stage has a clear output and a clear exit rule, so nothing moves forward on gut feel. Research feeds concepts, concepts become briefs, briefs become batches, batches get tested on hook metrics first, winners get variations, losers get autopsies, and scale only happens with fresh creative waiting in the wings. It's unglamorous. That's why it works.
Frameworks usually fail at step one: they're diagrams
Most creative frameworks you've seen are circles and arrows on a slide. Research leads to ideation leads to production leads to optimization, around and around, forever. Beautiful. Useless.
The problem with diagram frameworks is they describe everything and decide nothing. There's no point where someone says "this concept is dead, kill it" or "this angle earned more budget." So everything drifts forward on vibes, and the framework becomes wall art.
Ours is a pipeline, not a diagram. Each stage has one output and one exit rule. If the output isn't met, you don't advance. Here's the whole thing.
Stage 1: Mine
Output: a list of raw angles, each tied to a source. Exit rule: no angle enters the pipeline without a source.
Mining is research, and research has a specific job: find things your competitors haven't said yet, in language your customers actually use. Sources, in order of value:
- Customer reviews and support tickets. The exact phrases customers use to describe their problem are your best hooks. Copy them verbatim.
- Ad comments (yours and competitors'). Objections, questions, skepticism. Every repeated question is a creative brief.
- Competitor ads. Not to copy, to map. What angles are they running? What's missing? The gap is your opportunity.
- Sales calls. The moment a prospect says "oh, I didn't know you did that" is an angle.
What you're not doing: brainstorming in a vacuum. "What if we did something with a drone?" is not mining. That's a hobby.
Stage 2: Concept
Output: one written hypothesis per concept. Exit rule: if you can't write the hypothesis in one sentence, it's not a concept yet.
A concept is an angle plus a bet. The format we use: "We believe [audience] will stop for [hook territory] because [insight from mining]." Example: "We believe price-sensitive skincare buyers will stop for a direct price comparison because reviews keep mentioning the $80 competitor."
One angle per concept. This is the rule everyone breaks. "It's a UGC testimonial that's also funny and also about the sale" is three concepts wearing a trench coat. Split them.
Prioritize concepts into two buckets: iterations (new hooks or formats on proven angles) and new bets (untested angles from mining). The ratio depends on account health. Winning? 70% iterations, 30% new bets. Struggling? Flip it.
Stage 3: Produce
Output: finished assets, batched. Exit rule: the batch ships complete, or it doesn't ship.
Production follows the brief (we have a whole article on writing briefs that don't get ignored). The framework-level rules:
Batch, don't drip. Produce 5-10 concepts per cycle. Single ads produced one at a time cost more per asset and arrive too late to matter.
Shoot variations at birth. For every concept, get 3-5 hook variations in the same shoot. The marginal cost is tiny. The testing value is enormous. An ad with one hook is a guess. An ad with five hooks is an experiment.
Match format to angle. Testimonial angles want faces. Comparison angles want split screens. Problem-first angles want direct address. Don't shoot everything as talking-head and wonder why the comparison angle flopped.
Stage 4: Test
Output: a read on every concept: winner, iterate, or kill. Exit rule: every concept gets a verdict within 7 days. No "let's give it more time" without a reason.
Testing structure matters less than testing discipline, but here's ours: new concepts go into a dedicated testing campaign (or ad set), broad audience, controlled budget. Same budget per concept. No favorites.
Read metrics in order:
- Thumbstop rate and 3-second views. The hook's report card. Bad here means the opener failed, full stop.
- Hold rate / ThruPlay. The body's report card. Good hook, bad hold means the middle lost them.
- CTR (link). Did enough people want more?
- CPA / ROAS. Only now. Judging a 2-day-old ad on ROAS is how you kill winners early.
Kill fast and without sentimentality. An ad that fails the hook test three times with three different hooks has a body problem. An ad that fails everywhere has a concept problem. Write down which, and move on.
Stage 5: Iterate
Output: variations of winners, autopsies of losers. Exit rule: every winner has at least 3 live variations before you call the job done.
This is the stage most advertisers skip, and it's where the money is. A winner is not a finished product. It's a template.
Winner variations: new hooks on the same body, same hook in a new format, same angle with a new creator, longer and shorter cuts. Each variation is cheap and each one extends the winner's life.
Loser autopsies: for every killed concept, record where it failed (hook, body, offer) and why you think so. This is the compound interest of the framework. Six months of autopsies is a strategy document no competitor can copy, because it's built from your data.
Feed both back into Stage 1. The mine gets richer every cycle.
Stage 6: Scale
Output: increased spend on proven winners. Exit rule: no scale without fresh creative already in testing.
Scaling is the reward for the previous five stages, and it has exactly one rule that matters: never scale into a creative vacuum. The moment you increase budget on winners without new concepts in the pipeline, you've started a countdown to fatigue. Budget goes up, frequency goes up, performance decays, and everyone acts surprised.
Practical scaling: increase budgets gradually (20-30% steps), expand to new audiences and geos where winners haven't run, and keep the testing pipeline running at the same cadence. Scale is not a stage you finish. It's a stage you maintain.
The whole thing on one page
| Stage | Output | Exit rule |
|---|---|---|
| Mine | Raw angles with sources | No source, no angle |
| Concept | One-sentence hypothesis per concept | One angle per concept |
| Produce | Batched assets with hook variations | Batch ships complete |
| Test | Verdict per concept (winner / iterate / kill) | Verdict within 7 days |
| Iterate | Winner variations + loser autopsies | 3+ live variations per winner |
| Scale | Increased spend on winners | Fresh creative already testing |
FAQ
How is this different from other creative frameworks? Most frameworks describe activities. Ours enforces decisions: every stage has an output and an exit rule. The difference between a framework and a diagram is what happens when something fails.
How long does one full cycle take? For most accounts, 2-4 weeks from mining to test verdicts. Faster if production is in-house, slower if you're briefing external creators. The cadence matters more than the speed: never let the pipeline go dry.
What if we don't have a research team for the mining stage? You don't need one. Reviews, comments, and support tickets are free and better than most paid research. One person spending half a day a month on mining beats a quarterly brainstorm every time.
How many concepts should be in testing at once? Enough that you're never waiting on creative. For most accounts spending five figures monthly, 5-10 concepts in test with 3-5 hook variations each. Scale the volume with spend, not with enthusiasm.
What do we do with concepts that are "okay" but not winners? Iterate once. Change the weakest link (usually the hook). If the second version still doesn't clear your bar, kill it. "Okay" creative is how accounts end up with 40 mediocre ads and no winners.
Written by the MetaMaxd team. We run Meta ads for brands that are tired of agency theater: no vanity metrics, no "hacks," just accounts engineered to make money. If your creative process is a Slack thread and a deadline, we should talk.
