## 1. Executive Summary

Host Joao Campos sits down with three of Brazil's top creative copywriters Bruna Campolina (owns her own operation, writes her own creatives, claims R$1M+/day revenue days), Bruna Rodrigues and Luisa Veloso (both at Grupo Six, each earning six-figure monthly commissions, with Rodrigues only ~10 months into her career). All three share the same mentor origin: Thiago Filemon, plus courses, immersions, and the audacity to cold-DM him for reviews. Their shared method is systematic, not inspired: they spy the paid market daily, extract validated patterns (big idea, invisible structure, format, angle), harvest hooks from the organic feeds of TikTok/Instagram/YouTube, and write to a documented hypothesis instead of gut feel. Senior Luisa calls copy "math, not luck"; a validated creative is treated as an asset to be mined through variations, hook-stacking, and A/B of structure vs. elements.

The episode's spine: position the creative as entertainment first; disguise the sale; let curiosity do the CTA's job; and use metrics (hook rate, hold rate, CPA, cost per initiate checkout, conversion) to form the next test batch's hypotheses. They distinguish big idea (the central message that stops the scroll), invisible structure (the order of psychological arguments), format (the visual packaging), and angle (a way of delivering the big idea). Production tricks that sound absurd but reportedly validated: no-CTA creatives, a woman performing a recipe with only sounds, a plastic banana tapped against glass, and a "contextualized clickbait" that only works when it matches the niche's identity.

## 2. Alpha / Sauce what actually moves the needle, and exactly how to run it

- **Copy is math with an invisible structure.** Each sentence is a psychological argument block that moves the reader to a conclusion (learned as "gradualization" from legend Codinha Fortes via David Deutsch's framework). The order of those blocks is the "invisible structure": Proof Benefit CTA second proof, etc. Once a big idea validates, generate variants by keeping the structure and swapping elements (scientific study testimonial demo) or keeping elements and reordering the structure.
- **Big idea is the person, benefit is the jacket.** The big idea is the single central message that makes the brain say "stay" in three seconds the user decides. In weight loss, "the celebrities' secret" works as a big idea; a curiosity benefit like "she got arrested for losing weight" opens clicks even outside the niche. Benefits, proof, and objections are the jacket they add efficiency, but the big idea is what's essential.
- **Borrow hooks wholesale from viral organic content.** Most hooks that scaled seven figures/day for them were lifted verbatim from an Instagram cut with ~24M views or a TikTik sound with absurd reach. If it stopped millions in organic, it can stop people in paid traffic. Judge organic by views AND engagement (likes/comments/ratio), then re-skin the base structure with that packaging.
- **No-CTA creatives outperform forced call-to-actions.** Creatives that end on a curiosity rather than "click here" make the person decide to click, so they arrive far more qualified. CTR isn't necessarily lower; the platform's "see more" button handles the mechanism. Analogy: a soap opera doesn't say "watch tomorrow" it ends on a cliffhanger.
- **"Contextualized clickbait" beats generic clickbait.** A fat person falling oddly qualifies exactly the overweight segment you want they identify and stop. Generic clickbait pulls unqualified cheap metrics; if you want a cheaper audience, buy it with a broad clickbait and add a qualifying step (quiz or pre-VSL) between the ad and the VSL so effective acquisition cost stays equal.
- **Stack hooks on validated bodies.** A winner gets variations by literally appending hook #2 and even hook #3 before hook #1 ("water tiktak, then banana tiktak, then the copy starts"): Facebook treats it as a new ad and stops different people. It barely needs congruence and it works their theory is you get 1 2 more seconds to grab attention.
- **Format > everything else in paid organic translation.** Bruna and Luisa independently validated the same viral format (a Kelly Clarkson moment) with completely different copy and both produced their top sellers Luisa's version alone passed 100M+ views. The format ("the packaging") is what carries the familiarity that hides the ad.
- **Metrics drive the next batch, not post-mortems.** High hook rate (3s retention) but no conversion = wrong avatar or weak credibility change the angle or add proof. High hold rate (retention to 75%) but no sales = the invisible structure is weak restructure. Every failed ad becomes a hypothesis for the next batch.

## 3. Methods & Setup Step-by-Step

1. Start every day with paid-traffic spy in the Meta Ad Library for ~1 2 hours, reading what is running now (the market is dynamic; what worked two months ago may be dead); setup: block daily library-mining time before writing anything.
2. Filter candidate ads by hard criteria: running 5+ days (before that it's a pre-scale test), 10+ copy variations on one creative, same body across many campaigns, and the strongest signal multiple different companies spending on the same copy structure; setup: build a saved checklist with those thresholds.
3. Log each promising ad with big idea, invisible structure, angle, and format in a running document or analog notebook, including the test question you want each batch to answer; setup: create a doc with columns "ad, big idea, structure, format, hypothesis".
4. Go to organic (TikTok desktop, Instagram Reels, YouTube Shorts) and hunt the formats, avatars, hook styles, and sounds the audience is already entertainment-habituated to; setup: save finds into platform collections ("avatars", "recipes", "before-after", "formats") or a blank Google Doc of links.
5. Decide the big idea for the niche e.g., "the celebrities' secret" using only themes that your audience demonstrably stops for; setup: list 2 3 candidate big ideas with organic evidence each.
6. Lay out the invisible structure for one variant, then generate variants by combining structure-order element-swaps rather than inventing from zero; setup: produce 3 big ideas 3 formats a few structures as a combinatorial matrix.
7. Write the copy last and write it like a conversation with one real person no filler words, no force because the ad must feel like content, not an ad; setup: write directly into the approved matrix cell, then re-read aloud.
8. Decide what to leave out: if adding a CTA, a loud animation, or an over-explained mechanism would break the seduction, drop it the sale happens in the viewer's imagination and expectations; setup: apply the "does this feel like an ad?" elimination pass.
9. Produce with clean editing; a first-minute VSL retention of 84% they attribute to this restraint, and heavy or flashy editing can ruin an otherwise validated creative; setup: send editors the exact copy + format brief, no improvisation.
10. Launch the batch and read metrics as questions: hook rate (3s retention) tells you if the hook stops people; hold rate (retention to ~75%) tells you if the body holds; CPA/CPC/cost-per-initiate-checkout/conversion tells you if it sells; setup: pull the metric set per creative, not just spend.
11. Kill or iterate by evidence: low retention cut before it burns a full CPA budget; high hook but no conversion audience is curious but unqualified, add proof or change the angle; high hold but no sales restructure the argument sequence; setup: route each outcome to a pre-decided action.
12. Scale a validated creative as an asset: first variation is always stacking new hooks, then visual variations (curious niche-contextual stunts), then avatar variations; setup: queue a variation ladder for every winner.
13. Stack hooks for "new ad" incremental reach: append a second and third hook in front of the validated body; setup: build hook-stack variants systematically.
14. Repeat the cycle daily their early-career months were three straight months of daily library mining, including weekends; setup: make spy-and-log a calendar habit, not a mood.

## 4. Tools & Stack what each one is for here

| Tool | Role in this operation |
|---|---|
| Meta Ads Library | Primary spy source for paid creative; zero delay (ads appear the day after launch), and you can mine "scaled" offers by expert/company |
| Deeptub | Secondary paid-spy database with a lag; still valuess how it finds good scaled assets |
| TikTok (desktop) + Instagram Reels + YouTube Shorts | Organic trend + format + hook mining; desktop TikTok used purely as a work spy tool, not for personal feed |
| TikTok Creative Center / Google Trends / Twitter trends | Early-wave detection: the "pink salt" hook sat at #1 on Trends for a long time; Trends also surfaces scaled-offer expert names and viral moments (their Kelly Clarkson play) |
| Platform collections & playlists + Google Docs | Saving and categorizing mined hooks/formats/avatars by niche ("recipes", "before/after", "avatars"), plus a blank doc for links |
| Vtub (host Joao's platform) | Videotape retention analytics: used by one operator as an ad-kill criterion (low retention don't pay the CPA to validate) and to diagnose by running a single creative |
| Notebook / hypothesis doc | Every creative's big idea, structure, and test-question logged by hand so decisions are evidence-driven, not luck-driven |

## 5. Actionable Checklist

- [ ] Block 1 2 hours of Ads Library mining into every workday.
- [ ] Mine only ads running 5+ days with 10+ copy variations or multi-company duplicate structures.
- [ ] Log every candidate: big idea, invisible structure, angle, format, and the question it answers.
- [ ] Mine organic feeds daily; save hooks, formats, avatars, and sounds into niche collections.
- [ ] Confirm a big idea with organic evidence before writing (views + engagement, not just views).
- [ ] Generate variants by combinatorial matrix (big idea format structure), never from zero.
- [ ] Write the copy last, as a conversation not a list of promises.
- [ ] Apply the elimination pass: if a CTA/animation/mechanism-dump would break seduction, cut it.
- [ ] Keep editing clean; send editors copy-briefs, not freedom.
- [ ] Read metrics as hypotheses: hook rate hook; hold rate structure; CPA/IC/conversion offer fit.
- [ ] Route each metric outcome to a pre-decided next action (kill, add proof, restructure, change angle).
- [ ] Scale winners by stacking hooks, then visual variations, then avatar variations.

## Post notes

- Guests: Bruna Campolina (@eubrunacampolina), Bruna Rodrigues (Grupo Six), Luisa Veloso (Grupo Six), host Joao Campos; episode Segredos da Escala #116.
- All revenue ("7 figures/day", 6-figure monthly commissions, ~130M views across two versions of one viral format, one creative selling ~R$100k/day in Brazil) is self-reported.