Creative Strategy Central / Cheat Sheets

📐 Angles & Ideation

An angle is the strategic lens (the argument), not the hook (the opening line) — pick one lens per ad, spread lenses across the awareness map so Andromeda sees genuinely different ads instead of headline swaps.

Core distinctions (get these right first)

  • Angle = the argument/lens (WHAT story). Hook = the opening line (HOW you start). Concept = idea+format. Iteration = variation of a proven winner.
  • Same offer has 8-15 distinct angle territories; most accounts run only 2-3. One angle on 80% of ads = your headline finding.
  • Two testimonials saying the same thing = ONE test run twice. Changing the person you talk to = diversity; changing the headline = iteration.
  • Andromeda collapses ads >~60% similar into one Entity ID. Differ meaningfully on one major dimension or it's not a real test.
  • When testing diversity: vary the ANGLE first, hook second. 5 angles x 3 hooks = 15 testably distinct ads.
  • Same scenario + same mechanism + same angle = ONE concept, no matter the creator/hook/wording.

Pick the angle by awareness level (Schwartz)

  • Unaware → lead with curiosity, story, worldview/emotional reservoir; HIDE the product. Straddle frame is safest cold entry.
  • Problem-aware → mechanism/reframe: 'it's not X, it's Y'; name the root cause more precisely than they can.
  • Solution-aware → differentiate category vs category: 'Tried A, B, C? Here's what they all miss.'
  • Product-aware → skeptic-to-believer, proof, comparison: 'I didn't believe it — here's day 3, 14, 30.'
  • Most-aware → offer/urgency/scarcity only; give a reason to act NOW.
  • Volume rule: ~40% of ad volume should be unaware — most teams underinvest because it's harder to write.
  • Towards vs Away lever per bucket: Towards = surface opportunity/success; Away = surface threat/pain/root cause.
  • Saturated market: plain-benefit angle is dead on arrival — winner must show a NEW mechanism or sell an IDENTITY.
  • MVBC: shift the MINIMUM viable set of beliefs to get the sale; lead with your strongest proof, don't stack every argument.

High-leverage angle families (recipe each)

  • Mechanism-of-problem: 'The reason you still have [problem] isn't [common belief] — it's [specific hidden cause].'
  • Todd Brown formula: '[outcome] by [named unique mechanism] without [common pain/objection].' Use at sophistication Stage 3+.
  • Enemy/industry-lie: '[Big Industry] made billions convincing [audience] that [problem] is fixed with [common solution].'
  • Insight reframe: 'You don't have [labeled problem] — you have [reframed deeper cause].' Or reverse causation: '[A] doesn't cause [B], [B] causes [A].'
  • Contrarian: 'Stop [common advice]. Start [opposite].' / 'Everything you know about [topic] is backwards.'
  • Authority + counterintuitive: 'As a [credential] with [years], I'm telling you [counterintuitive claim].'
  • Stalker/hyper-specific pain: 'If you're [situation] AND [symptom] AND [frustration]...' — makes Meta look like it read their mind.
  • Mechanism is the single highest-leverage tag: a proven mechanism run for one persona but NOT an adjacent one = high-probability gap.

Segment = the WHO (angle multiplier)

  • Cut the market 3 ways only: Outcomes (WHAT) x Demographics (WHO) x Facets (WHY). No fourth cut.
  • Sort outcomes by WANT not TYPE (pest control: 'get them out'/'protect the house', not by pest). Aim 4-7 outcomes.
  • The gate: only split a segment if it needs a genuinely DIFFERENT ad; otherwise merge.
  • 5 facet families: Belief→the argument, Identity→vocabulary, State→the problem you lead with, Value→the promise, Occasion→the why-now.
  • Belief trap: write the CUSTOMER's belief ('my hormones are the problem'), never the marketer's ('science-backed ingredients'). Find beliefs, don't install.
  • Stack 2 facets → invent a pocket nobody targets (vegetarian x just had a baby = vegan postpartum). Stack 3+ → no volume.
  • Name 5-15 personas alliteratively by defining trait: Cortisol Carl, Melasma Mia, Postpartum Paula.
  • Say-it test: would a real customer say this segment/belief out loud? If not, it's marketer language — cut it.

Coverage map — find the gap before you write

  • Coverage = Segments x Awareness x Concepts x Angles x Styles x Hooks. See the whole grid; make intentional bets, not dart throws.
  • 25-cell matrix = 5 segments x 5 awareness. Cell states: PROVEN / TESTING / EMPTY. Most accounts cover only 3-6 of 25.
  • Concentration trap: 70% of ads in one cell = '5 ads wearing 5 shirts' — algorithm sees one signal and fatigues it together.
  • Trunk-Branch-Leaf priority: fix awareness/core-desire (trunk) before concept/angle/format (branch) before hook/copy tweaks (leaf).
  • Push a gap UP if it re-points a PROVEN mechanism at a new persona, is adjacent to a proven cell, or fills a thin cold column.
  • Push DOWN if no comparable account touches it and no proven adjacent mechanism — that's speculative, belongs in the 30% budget.
  • Reading competitors with no metrics: longevity (30-90+ days = workhorse), variation count (10-20 = scaling), impression bucket, format mix.
  • is_proven rule of thumb: days_active >=30 OR variations >=8 OR high impression bucket.
  • Also check FORMAT coverage: a wall of BOF product-hero + offer-promo with zero TOF cold formats is the most common gap.
  • Thin/new account 5-concept starter: problem-solution video, lifestyle carousel, founder-story video, long-form-static advertorial, product demo.

CASH variation engine — 1 winner → many ads

  • Break the winner into Concept, Angle, Style, Hook. Change ONE dimension at a time; each single-change = a genuinely new ad/Entity ID.
  • Keep 3 constant, swap 1: new hook (Warning Sign→Testable Proof), new angle (Story→Contrarian), new style (static→UGC video), new concept.
  • Intermediate-mechanism swap: same end mechanism, route through inflammation OR blood flow OR toxins — each intermediate is a new variation.
  • 70/30 rule: ~70% of next batch = iterations on YOUR proven cells; ~30% = brand-new angles off the gap list. Flip only in exploration mode.
  • Probabilistic tiers for a 40-ad batch: 12 direct swipes, 10 analog swipes, 8 data hypotheses, 6 CASH combos, 4 wild cards.
  • Multiply 1→20 by changing one variable: voice, color, actor, script, hook footage, CTA, or angle. #1 rule: don't copy — make it better.
  • Density check: every line = a fascination stacking Copy Blocks (Pain/Promise/Proof/Curiosity/Constraints); make each Clear, Relevant, Accurate, Visual, Expressive, Specific (CRAVES).

⚡ When Stuck

  • Stuck on what angle → identify awareness level first, then pull a formula from that Schwartz bucket (section 19 of Angle-Formulas-Master).
  • Account feels 'diverse' but stalling → check concentration; if 70% sits in one cell, you have 1 signal not 20 ads.
  • Can't find a winner → don't optimize, add breadth. Need 10-15 genuinely different creatives before patterns are real.
  • Out of fresh ideas → take one proven winner, run CASH: swap angle, then style, then hook — three new ads in minutes.
  • New/thin offer → cross-pollinate proven hooks+angles from sibling offers, start with video, cast the widest net (unaware straddle).
  • Mechanism exists but scale stalled → re-point that PROVEN mechanism at an adjacent persona who isn't getting it — highest-probability gap.