Creative BlueprintCreative Blueprint← Both skills · wildhorn-static-ads · 1-understand/ad-dna.md
Source: adn.machineroom.app (/srv/adn/index.html mind map + lab, simple.html Evolve × Crafter
view, script.html pipeline, builder.html spec, brief.html clone a winner). This file is the
operational summary for analysis work: how to code an ad, how to count what is really new, how to plan
crosses, and how the simplified (course-aligned) view maps onto it.
An ad isn't "a video". It's a combination of genes. Name every gene and you can count ads scientifically, know exactly what won, and spin a winner by swapping one piece at a time.
| Branch | Genes | Rule | Meaning |
|---|---|---|---|
| Foundation | PD Product · OF Offer · DS Destination | Change one → funnel test (not a new creative) | Constants during a creative test |
| Concept (stage 1) | AV Avatar · AN Angle · ME Mechanism · ST Story · AW Awareness | Change one → new concept (C) | What we say, to whom |
| Execution (stage 2) | HK Hook · MS Messenger · FO Format · PR Proof · SC Script structure · TO Tone | Change one → variation (V) | How we say it |
| Finish (stage 3) | SE Setting · ED Editing · SO Sound · LE Length · CT CTA · CP Ad copy | Change one → iteration (i) | The polish |
Rule: the highest branch that changed decides what you count.
AV Avatar — who exactly, by situation (trigger, cause, what they tried). Never demographics; age and sex are chosen later through the actor. AV1 Performance-driven (gym, work) · AV2 Hides it (avoids beach, shirt off, bedroom) · AV3 After a life event (40, baby, menopause) · AV4 Just got a scary number / diagnosis · AV5 After another treatment (GLP-1, pills) · AV6 Partner complains / is affected · AV7 Tried everything, now skeptical · AV8 Buying for a loved one. Cross: keep angle + mechanism, change who we talk to (words, setting, a messenger who looks like them) → open a new audience with a proven concept.
AN Angle — the exact pain or desire attacked, the entry door. AN1 Symptom (fatigue, sleep, belly, hair) · AN2 Performance (gym, sex, work) · AN3 Looks (mirror, man boobs, skin) · AN4 Couple & family · AN5 Fear for future health · AN6 Identity ("be the man I used to be") · AN7 Trigger event (40, summer, wedding) · AN8 Past failure ("nothing worked"). Cross: a mechanism won → it unlocks other pains (parasites → man boobs becomes parasites → erections, hair, focus).
ME Mechanism — why the problem exists and why this fixes it when everything else failed. ME1 Hidden cause inside the body (parasites, inflammation, cortisol) · ME2 Outside culprit (plastics, modern food, industry) · ME3 Deficiency ("you're missing X") · ME4 Absorption ("the rest never gets absorbed") · ME5 Hero ingredient (ancestral plant, clinical dose) · ME6 Timing or ritual · ME7 Synergy (the combo that unlocks it) · ME8 New discovery (recent study). Cross: the angle works but the mechanism is worn out or the market is saturated (erections: parasites → plastics → zinc deficiency).
ST Story — the big idea that wraps the mechanism. ST1 Discovery ("a researcher found…") · ST2 Enemy / "you've been lied to" · ST3 Personal transformation · ST4 Ancestral or cultural secret · ST5 Us vs them · ST6 Myth-busting ("stop doing X") · ST7 News, something new · ST8 Pure education · ST9 Tested challenge ("30 days with…"). Cross: the concept is good but the ad is fatiguing ("you've been lied to" → "I tried it for 30 days").
AW Awareness — what the prospect already knows; decides where the ad starts. AW1 Unaware · AW2 Problem-aware · AW3 Solution-aware · AW4 Product-aware · AW5 Ready to buy. (Note: the Copy Constitution numbers awareness the other way round — AW5 unaware → AW1 most aware. Always write the word, not only the code.)
HK Hook — what stops the thumb; 3 layers you can swap separately: visual, on-screen text, voice. HK1 Callout ("if you're over 40 and…") · HK2 Question · HK3 Bold contrarian claim · HK4 Number or stat · HK5 Story open ("6 months ago…") · HK6 Scroll-stopping visual · HK7 Result first (before/after) · HK8 Secret, curiosity · HK9 Forbidden ("stop doing…") · HK10 Someone's reaction. Cross: first move on any winner, the cheapest: 5–10 hooks shot in one session on the same body.
MS Messenger — who talks. MS1 Regular customer (UGC) · MS2 Expert (doctor, naturopath, coach) · MS3 Founder · MS4 A loved one (wife, daughter) · MS5 Known creator · MS6 Voice-over, no face · MS7 Older, wise character · MS8 AI avatar / animation · MS9 Interview duo. Cross: the script is good, credibility caps out (customer → doctor → his wife).
FO Format — the container. FO1 Talking head · FO2 Podcast clip · FO3 Street interview · FO4 Green screen (article/study behind) · FO5 Fake news report · FO6 Demo, tutorial · FO7 Routine, unboxing · FO8 Static image · FO9 Carousel · FO10 Native screenshot (tweet, iPhone note, SMS, Reddit) · FO11 Before/after split · FO12 Long video (VSL 2–5 min). Cross: one winning script = several free formats (talking head → static → tweet → podcast). (Compliance: "fake news report" formats imitating real media are banned for our brands — see copy-analysis.md.)
PR Proof — the main proof on screen. PR1 Testimonial · PR2 Study, scientific number · PR3 Visible demo · PR4 Before/after · PR5 Authority · PR6 Volume (50,000 customers, 4.8★) · PR7 Guarantee, zero risk · PR8 Logic (the explanation is enough). Cross: people watch but don't buy.
SC Script structure — the order. SC1 Problem → agitate → solve · SC2 Story (before → turning point → after) · SC3 List ("3 reasons") · SC4 Objection → answer · SC5 Side-by-side comparison · SC6 Education, product revealed late · SC7 Direct offer from second one. Cross: watch time drops in the middle.
TO Tone — the dominant emotion. TO1 Serious, medical · TO2 Humour · TO3 Emotional, vulnerable · TO4 Anger, outrage · TO5 Urgency, fear · TO6 Buddy, casual · TO7 Aspirational, premium. Cross: the audience is tired of one register.
PD Product (formula, form, packaging) — the "brand swipe cross": take a competitor's full structure and execution, swap their product and mechanism for ours (e.g. a Mars Men winning format with our angle + mechanism). OF Offer (price, bundle, subscription, guarantee, gift) — test when conversion, not the click, is the problem. DS Destination (product page, advertorial, quiz, VSL, listicle) — test when CTR is good and conversion isn't.
SE Setting (car, kitchen, bathroom, doctor's office, gym, bed, outdoors) · ED Editing (pace, b-roll, zooms, caption style) · SO Sound (music, voice, SFX, silence) · LE Length (15/30/60 s) · CT CTA ("tap below", "take the quiz", promo) · CP Ad copy (primary text, headline, description). Use when the ad fatigues (CPM up, CTR down): new setting + new edit on the same script = same ad, longer life.
| What changed | You count | Code | For Meta |
|---|---|---|---|
| Any Concept gene | New concept | C+1 | Truly new: can reach new people |
| Only Execution genes | New variation of the same concept | V+1 | Close: may be grouped with the original |
| Only Finish genes | Iteration (same ad) | i+1 | Near-identical: zero diversity |
| A Foundation gene | Funnel test, not a new creative | — | Test separately, creative frozen |
Ad name: BRAND-C014-V03-i2 (brand · concept · variation · iteration).
DNA string: AV2·AN3·ME1·ST2·AW2 | HK1·MS2·FO4·PR2·SC1·TO1.
The number that matters is concepts, not ads. 60 ads that differ only by hook = 1 concept, 60 variations. Meta sees them as nearly the same ad: they steal delivery from each other instead of reaching new people. A healthy account runs 3–5 live concepts at once.
Two modes, never mixed.
Vocabulary warning: in Ad DNA, "concept" = the strategy genes (who × which pain × which mechanism). In Evolve, "concept" = the idea being tested ("test us-vs-them"). Say which one you mean.
| Beat | Content |
|---|---|
| HOOK | "If you're over 40 and growing man boobs, it's not what you eat." |
| MECH | A naturopath explains: gut parasites block nutrient absorption; the body stores fat; that fat makes estrogen. |
| PROOF | Study on green screen, diagram of the gut. |
| PRODUCT | Oregano oil kills the parasites, absorption comes back, the fat melts. |
| CTA | "Link below." |
DNA: AV2 (hides it) · AN3 (looks: man boobs) · ME1 (hidden cause inside the body) · ST2 (enemy, "you've
been lied to") · AW2 (problem-aware) | HK1 callout · MS2 expert · FO4 green screen · PR2 study · SC1 PAS ·
TO1 serious. Name: OR-C001-V01-i1.
Crosses from this one winner (each row changes exactly one thing):
| Cross | Gene | New ad | Counts as |
|---|---|---|---|
| Angle | AN3 → AN2 | Same parasites → fat → estrogen, pain = erections failing | C002 |
| Angle | AN3 → AN1 | Same mechanism, pain = hair loss | C003 |
| Mechanism | ME1 → ME2 | Same man boobs, cause = plastics in water and food | C004 |
| Avatar | AV2 → AV1 | 28-year-old gym guy: "you train and your chest stays soft" | C005 |
| Story | ST2 → ST9 | "I took oregano oil for 30 days, here's what came out." | C006 |
| Hook | HK1 → HK9 | "Stop doing push-ups for your chest." Same body. | C001-V02 |
| Messenger | MS2 → MS4 | Same script told by his wife: "my husband was ashamed to take his shirt off" | C001-V03 |
| Format | FO4 → FO10 | Same text as a Reddit post screenshot | C001-V04 |
| Finish | 60 s → 30 s | Same ad, recut | C001-V01-i2 |
One winning concept → 5 new concepts + 3 variations, without reinventing anything.
The lab (on the site) computes the verdict against a reference: NEW CONCEPT (a concept gene changed; "clean cross" if exactly one), VARIATION (only execution genes; warning if > 1 changed), SAME. The combination space is huge (thousands of concepts × thousands of executions) — that is why you test branch by branch, never all at once.
Evolve's chain: Product (UMP + UMS) → Desire (13) → Sophistication (1–5) → Sub-avatar (by exclusion) → Angle (the outcome) → Ads. An ad = STRATEGY × SCRIPT × VISUAL; script and visual are independent axes.
| Family | Levers | Change → |
|---|---|---|
| STRATEGY (who, which desire, why it works) | Desire · Sub-avatar · Angle · Mechanism · Authority | NEW |
| SCRIPT (the words) | Hook (0–3 s, ×3) · Bridge (~8 s, ×3) · Hold (~32 s, ×1) | ITER, same angle |
| VISUAL (who, where, how it's cut) | Format · Actor · Location · Length · Edit | ITER, same words |
Fixed during a creative test: product, offer, landing page. Polish (not a lever): caption colour, "Shop now", badges, logo, music track.
Named moves on a winner
| Move | Keeps | Changes | When |
|---|---|---|---|
| Static ↔ video | The whole message, headline included | The medium | A winning angle must exist in both |
| Same script, new visual | Every word | Actor, location, format or edit (one) | Winner fatigues; keep hook wording ~80 % |
| Same visual, new words | The visual | Hook or bridge wording | Sniper / Evolve axis 6 |
| Answer the comment | Format, persona | The objection the hold answers (most-liked comment) | A recurring doubt blocks a pocket |
| Fill the gap | A format already winning | The angle, from another winner | Winning format A and winning angle B never combined |
| Isolate the sub-angle | Everything else | The angle cut down to one part | A winner mixes 2 ideas: which one sells? |
| Language / market | The message | Language + settings | A winner in one country |
| Other product | The whole creative | The SKU | Fits another product in the store |
| Wording cascade | Every past winner | Wording, replaced by a new golden nugget | A new review phrase works |
| Funnel stage | Angle | Length and offer (shorter, product-first, promo) | BOFU version of a TOFU winner |
Script × visual grid: one angle = a grid. Rows = script versions (hooks/bridges), columns = visual versions (actor/location/format). Along a row = same words, new visual (visual ITER); down a column = same visual, new words (script ITER). Never change both at once.
What to change, for the result you see
| What you see | The problem | Pull first | Then |
|---|---|---|---|
| Doesn't spend, bad CPA, nobody stops | The hook doesn't reach the right people | HOOK (more specific, not louder) | Format, actor |
| They stop, then drop | The middle loses them | BRIDGE | Edit pace, length |
| They watch, don't click | Not believable / no reason to act | HOLD (proof, objection, booster) | Authority, mechanism |
| They click, don't buy | The page doesn't continue the ad | LANDING / OFFER | Page adapted to the winning angle (week 3) |
| A top comment keeps coming back | A pocket blocked by an objection | Answer the comment | One ad per recurring objection |
| Winner fatigues | Same people saw it too often | Actor, location, format (same words) | Static ↔ video |
| Well-researched angle flops | Execution, not the angle | Hook, format | Drop the angle only after 20–30 creatives |
| Can't scale past a budget | Angle audience used up | ANGLE (Marksman) | Sub-avatar, language/market |
| Everyone in the market says the same | Sophistication 3–5 | SUB-AVATAR (exclusion resets) | Mechanism (new or better) |
Known defects of the detailed tabs (kept as built): demographic avatars there are an Evolve error; hook and format sit at the same "variation" level though script and visual are separate axes; "concept" means strategy there. When in doubt, follow /simple.
Since the founder rejected the "pick 6 genes + 80/20 iterations" model, ads are made by transplanting a winner. In DNA terms:
Clone testing: one ad set = one source, 3-2-2 shell (3 creatives × 2 primary texts × 2 headlines); the ad name carries source + angle; a winning clone becomes a source itself; a source that loses on 3 angles is dropped; an angle that loses on 3 sources is dropped.
Every video beat belongs to one of 7 colour-coded moments: Hook (0–3 s) · Problem · Mechanism · Product (~25–33 s) · Proof (~33–40 s) · Offer (~40–45 s) · CTA (~45–48 s). Gene → moment: AV/AN live in Hook + Problem; ME in Mechanism + Product; PR in Proof; OF in Offer; DS where the CTA sends; ST, AW, MS, FO, SC, TO, SE, ED, SO span the whole ad.
Ownership: Strategy (AV, AN, ME, ST, AW, PD, OF) → Script (HK, PR, SC, TO, CT) → Shoot (MS, FO, SE) → Edit (ED, SO, LE) → Launch (CP, DS). The script decides WHAT (words, order, who speaks, proof, visual direction); the editor decides HOW (b-roll, captions, music, SFX, pace, exports) and never changes words or claims.
For any ad under review, write: