## 1. Executive Summary

Bruno Monteiro joined Empiricus in 2016 as a front-end programmer and left roughly five years later as CTO, having grown with the company from about 30 people to more than 250 while personally leading a team of around 25. Empiricus was the largest direct-response operation in Brazil at the time, built on The Agora Co.'s subscription model (founded by Bill Bonner in the 1970s) with an in-house checkout, in-house BI, and a flagship newsletter that reached around 400,000 paying subscribers before the company was acquired by BTG. His doctrine: annotated references plus full-funnel depth beat clever copy test against a written kill limit, cut what fails, and keep moving. He now runs Swiper, what the show bills as Brazil's largest swipe file, at R$39/month, grown entirely with organic traffic.

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

- **Annotate the swipe file; a raw archive is worth nothing.** Every entry must answer the question "which line turns a passive reader into a potential buyer" that annotation, not the clipping, is the asset. Bruno learned this the hard way: an unsorted drive of saved ads was unusable, and even Clayton Makepeace's 500-item file leaves a beginner asking where to start.
- **Model validated old angles and swap the variables.** The classic process he studies gave a writer about four weeks for a 1,000-word newspaper ad priced by the page, so the angle itself is already proven and almost nobody in Brazil works this way. The host's version of the play: an advertorial headlined "Atlanta housewife investigated and almost arrested for losing 35 kg" ran on native with an absurd CTR and the offer did around R$35M; his "dentist who hasn't brushed his teeth in 20 years" line was modeled from a newsletter headline about a doctor who had not drunk water in 20 years.
- **Run every funnel layer, not just the first sale.** Free newsletter at zero friction front-end from R$1 to R$19.90 mid-end bridge product back-end above R$2,000 combos (all products for a year, lifetime access); the mid-end exists so the jump into the back-end doesn't trigger refunds, and the flagship newsletter alone reached about 400,000 paying subscribers. A R$19.90 front-end only prints with huge volume and cheap traffic, and the real damage is the silence two months after a sale when nothing else is offered.
- **Write the kill limit before you launch.** Empiricus ran on attempt volume of 100 ideas, about 10 survive, about 3 are good, and 1 explodes, mirroring the options-market rule they respected: out of every 20 trades, one pays for all the others. Set your own threshold in writing, for example an offer must sell inside its first 1,000 views; when you hit it with no response, cut it and go back to the core offer.
- **Peer review every piece before it ships.** Weekly, the writer presents and the room attacks the argument; if a stranger in that room can't follow it, the market won't either. Swiper runs the same ritual live: the subscriber reads their own copy aloud out loud, not silently so boredom and bottlenecks surface, then the group audits for a missing guarantee, bonus, and payment terms.
- **Hire outside the copy market and train a step-by-step process.** Empiricus hired journalists such as Helena and other people with no direct-response background, because Agora's one-page, big-idea template turns ordinary writers into people who can sell financial letters; Copy Camp was the pipeline free training, a championship, about R$10,000 to the winner, then a job. The copy chief's filter was not framework compliance: he wanted copy that keeps you awake and makes you say "if this ad reached me, I would buy."
- **Keep data in-house and decide within the day.** Empiricus's edge was data volume for roughly its first eight to nine years it competed alone with internal BI and a wall screen in marketing showing campaigns, lead volume, and average CPL, so a route set at the start of the day could change by the end. When one cart opened, more than 1 million leads were queued behind it, and money goes out first and returns later, so a slow call is an expensive call; the base math never changes: spend less than LTV, which you cannot even run if you never measure LTV.
- **Match the message to the source before blaming the copy.** Meta, native, Google, and YouTube audiences need different arguments native is a far colder, "different planet" audience than Meta, and YouTube wants a slower, educational read native to the platform. On one VSL analysis, Bruno saw desktop retention run about 40% higher than mobile, so split by source and device before killing a funnel.

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

1. Build an annotated swipe file before you need one; setup: save every ad, letter, or page that stops you, tag it by hook, angle, and funnel stage, and write one line on why it works the note is the asset, not the clipping.
2. Mine old copy for angles, then swap the variables; setup: start with newspaper ads and letters from the 1900s 1950s (about four weeks of work on a 1,000-word unit priced by the page), keep the structure and replace the nouns "doctor who hasn't drunk water in 20 years" becomes "dentist who hasn't brushed in 20 years."
3. Map the buyer's journey from complete beginner to expert, then hang a product on each level; setup: free newsletter at zero friction R$1 R$19.90 front-end mid-end bridge back-end above R$2,000 annual and lifetime combos, and add a layer only where the customer's next problem actually appears.
4. Write down the number that kills the idea; setup: for example, an offer must sell inside its first 1,000 views or clear a stated revenue figure, plus the cut rule hit the limit with no response, kill it, return to the core offer.
5. Put every piece through peer review; setup: a weekly meeting where the writer presents and the room questions it, with a strict reading rule if a stranger in the room cannot follow the argument, the audience will not either; solo operators gather two or three friends who study copy.
6. Read your own copy aloud in a live session; setup: Swiper's format you read it out loud while others listen for boredom, then audit for a missing guarantee, bonus, and payment terms, since silent reading hides the bottlenecks.
7. Ask buyers what they do the moment they pay; setup: a manual message right after checkout "What do you do? Copy? Editing? Traffic?" with no quiz and no automation, logging answers so each unanswered need becomes the next product or a partner offer.
8. Instrument the funnel and decide the same day; setup: a real-time view of campaigns, lead volume, and average CPL (Empiricus used a wall screen in marketing), then check spend against LTV before scaling.
9. Ship a second offer before the customer goes quiet; setup: do not stop at the first sale and go silent for two months build the next layer yourself, or partner so someone else's product fills the back-end.
10. Launch the entry layer as a free experience, then ask for the money; setup: Swiper began as a free hobby library and grew by a single pay button for the translated version, and now runs at R$39/month with 7-day full access, the whole library visible before purchase, monthly or annual billing, and 100% organic distribution no paid traffic, no VSL, text copy only.
11. Split performance by source before judging the copy; setup: compare desktop against mobile retention and channel against channel (Meta vs native vs Google vs YouTube); one VSL showed about 40% higher retention on desktop.
12. Read one piece of copy daily in a calm window and do the exercise; setup: before bed or any quiet slot with notes nearby, because every Swiper entry ends with a contextual exercise such as writing a TikTok ad on a similar hook a live project today usually gives you two or three days to deliver.

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

| Tool | Role | In this episode |
|---|---|---|
| Swiper (swiper.com.br) | Annotated swipe-file library, Bruno's product | R$39/month, 7-day full access, everything readable before buying; started free as an ice-breaker; 3 years old, 100% organic, no VSL |
| VTurb | Video player for sales letters with retention data | The diagnostic for "copy problem or source problem": retention by channel and device, e.g. about 40% higher on desktop for one VSL |
| Meta Ads | Paid social source | The channel where the same funnel can flop while native wins arguments must differ by source |
| Taboola (native ads) | Placement for cold, frozen audiences | Where the host ran the adapted "arrested for losing 35 kg" angle: absurd CTR, offer around R$35M |
| ChatGPT | Funnel-gap brainstorming | A 30-minute session to list what to sell next when customers stop renewing |
| Hotmart | Third-party checkout and membership | What Bruno would move an in-house recurring-billing stack to today to cut operational weight |
| Copy Logic (book) | Source of the peer-review method | The review format Empiricus ran weekly and Swiper runs live with copy read aloud |

## 5. Actionable Checklist

- [ ] Create one swipe file today and save the next ad that stops you, with a one-line note on why it works.
- [ ] Pull three classic ads from the 1900s 1950s and rewrite each hook for your current offer.
- [ ] Draw your funnel on one page and mark what sits at free, R$1 R$19.90, mid-end, above R$2,000, and combo.
- [ ] Write your kill threshold (first 1,000 views or a revenue figure) before launching the next test.
- [ ] Book a weekly review where you present your copy and take the criticism; find two or three peers if you work alone.
- [ ] Read your next draft aloud, cut every line that bores you, then check guarantee, bonus, and payment terms.
- [ ] Message every new buyer with three questions about what they do and log the gaps as product ideas.
- [ ] Put campaigns, lead volume, and CPL on one live dashboard and check spend against LTV.
- [ ] Ship the next offer or partnership before the two-month silence after every sale.
- [ ] Read one swipe a day in a quiet window and complete the exercise attached to it.