## 1. Executive Summary

This is a roundtable, not an interview: three of the largest YouTube Ads operators in Brazil Sara Lima, Lucas Rego and Denis Pereira sit down with the VTurb host to walk through the full stack of running YouTube traffic in 2025, from creating an ad account to scaling. All three claim eight figures per month, Sara claims over R$1 million per day with YouTube as her main media, and one offer discussed in the episode reportedly did around R$70 million lifetime. The core thesis: YouTube is the most profitable media you can buy an offer doing 1.5 1.6 ROAS on Facebook pulls 2.2 2.3 on Google at scale but it is also the hardest to run, because accounts fall constantly, Google rewrites its own rules every two months, and the entire game now runs on account volume, disciplined warm-up, and never testing unproven creatives inside Google.

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

- **Never test cold creatives on Google validate on Facebook first.** Google is too fragile: a few days of testing can kill the account, so the platform is for scaling, not discovery. Sara's rule is a CPA under R$170 after 3 4 days and R$100 150 of spend on Facebook; Lucas requires 10 15 sales at a good CPA before an ad earns a YouTube slot. Facebook's expensive CPC (R$12 20 in weight loss vs R$4 5 on YouTube) is the price of finding out what works before you risk a farmed account.
- **Two-stage warm-up: "simple" then "complex."** Simple warm-up is 1 3 days of tiny spend (R$5 10/day) on something plainly white a search campaign to a pet shop or blog just to get the first approvals. Complex warm-up is where the real alpha is: run 7 15 days on the *actual* structure you will scale (your VSL, your presell, your domain), pushing ~3 ads per day for approval, duplicating up to three times a day, targeting 40 100 approved ads inside one campaign before any black offer goes live.
- **Approved-ad volume beats spend.** R$100 200 of budget means nothing; 50 100 approved ads on your own domain is what makes Google stop looking. Google re-reads structure repeatedly change a line of the page or swap the domain and the account often drops for fraud the next day. So you warm up the exact structure you intend to monetize, keep a weekly human routine (duplicate, pause, launch something new nobody real leaves an account untouched for 60 days), and let the review pass over a page it has already approved a hundred times.
- **Farm accounts, don't buy them and if you buy, buy access.** Buying peaked; recent spend now matters far more than lifetime spend, and even a fresh account with R$5,000 of recent spend can die in days. Before paying: request view-only access, filter deleted and rejected ads (even 5 rejections on a "white" account is a red flag, sexy-shop history is an instant no), and require full control of the payment profile the panel lost close to R$700,000 to a payment-profile scam, half of it possibly recoverable only through litigation. Leave the account untouched for a day, then remove the previous admin.
- **Farm at volume, with randomized process.** Sara's team creates each account inside a MultiLogin profile, browses like a normal user for 3 7 days before touching Gmail, fills the inbox to roughly 300 newsletters before opening Google Ads, and rotates 12 16 proxies so no proxy mints many accounts in one day. Denis keeps three farm frameworks in parallel bought Gmail, created Gmail, even plain Chrome on home Wi-Fi deliberately refuses to lock step timings into fixed numbers, and changes one process variable every 1 2 months even when it's working, because a repeating pattern is what gets 200 accounts burned at once. He stocks 300+ accounts and farms 200+/month so he never scales from zero.
- **Launch settings: maximize conversions first, target CPA later.** Farmed accounts arrive stuck a target-CPA campaign can sit for a week spending R$10 30/day against a R$2,000 budget. Denis now launches the first campaign on maximize conversions (it starts spending within a day, supports R$10 30k/day), then runs later campaigns on target CPA; Sara goes straight to target CPA with a R$700 1,000 bid because her complex warm-up unlocks the account faster. Typical first budget: $200 500 (R$250 most often), with three creatives as either 1 campaign / 3 ad groups / 1 ad each, or 3 campaigns of 1 ad.
- **Pre-scale has hard numbers: 10 15 conversions before you touch budget.** A new account that starts well but has few conversions cannot hold scale push it and you reverse. Sara steps budget +30% at a time up to about $10k, then doubles and triples; Lucas waits for 10 15 conversions, then doubles once or twice a day in 2 3-hour windows (two VSL cycles), always reading *today's* sales by hour, not yesterday's ROAS polluted by midnight traffic. His definition: first double = pre-scale, three doublings (8x budget) = real scaling; the moment the campaign survives two doubles you pop the champagne "fitting the campaign" is the whole skill.
- **Niches die cyclically read the window and the numbers.** ED was killed off after an adult site scandal made Google purge the niche, weight loss is currently hard because everyone flooded it, and each niche carries its own initiate-checkout benchmark: ~R$20 for weight loss, ~R$15 for ED, R$35 40 for diabetes. When a niche falls, accounts are the constraint, so operators rotate to whichever window is open instead of forcing a dying one.
- **Operations are the moat: revive accounts, and never stop the machine.** "The account will fall" is a certainty; what separates operators is a process to bring it back daily review requests through the MCC, sometimes a dozen people doing nothing else, because an account revived at R$40,000/day already carries 60 80 conversions and resumes selling immediately. Around that: a 15-minute daily traffic call, weekly creative and VSL calls (hook rate above 60% is a strong ED creative even without sales), a shared metrics spreadsheet with an ID per ad, and a Monday sprint locking copy, editing and traffic to fixed launch dates so a VSL ships in 10 days instead of a month.

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

1. Farm accounts in-house: MultiLogin profile, browse like a normal user for 3 7 days, create or buy an aged Gmail (R$5 7), then fill the inbox to ~300 newsletters before opening Google Ads; setup: 200+ accounts/month with 300+ in stock.
2. Rotate 12 16 residential proxies across profiles never many accounts per proxy per day, and give each proxy a ~3-day gap before reuse; setup: proxy log per account with standby rotation when one loads too many.
3. Keep three farm frameworks running in parallel and change one process variable every 1 2 months, even while things work; setup: ranges (3 7 days) instead of fixed step counts, randomized task order.
4. Create the ad account with a deliberately tiny budget (R$5 10/day) a high day-one budget triggers instant fraud suspension; setup: budget cap set before the first campaign exists.
5. Run simple warm-up for 1 3 days: white search campaign to a pet shop or blog, small spend, first approvals; setup: one white page per account, never the same domain across all accounts.
6. Run complex warm-up for 7 15 days directly on your real structure VSL, presell or quiz on the domain you will scale pushing ~3 ads/day and duplicating up to 3x/day until 40 100 ads are approved; setup: approved-ad count as the KPI, not days or spend.
7. Move the warmed account under a master MCC with a sub-MCC per manager and a corporate-domain email; setup: all recovery requests and reviews filed through the MCC.
8. Launch the first real campaign on maximize conversions with $200 500 and three creatives (1-3-1 or three single-ad campaigns); setup: target-CPA bid of R$700 1,000 reserved for accounts that already spend freely.
9. Feed the pipeline from Facebook: top performers adapted to YouTube (strip licensed music, subtitles, aggressive clickbait), organic look one avatar, one recipe, a viral hook; setup: two edit formats delivered from the same copy, YouTube and Facebook.
10. Judge creatives on niche benchmarks CPA plus initiate checkout not vanity: cut at $50 200 with no sale on a fresh account, give proven Facebook winners a second chance; setup: per-niche IC reference (R$15 40 range).
11. Wait for 10 15 conversions before raising anything, then pre-scale: +30% steps (Sara) or one doubling per 2 3-hour window (Lucas), reading sales by hour and skipping increases when today's numbers are bad; setup: decision windows set to two VSL cycles.
12. Declare scale only after three consecutive doublings (8x original budget); keep the white approved-ad wall intact as contingency and never touch domain or page once live; setup: suspension playbook daily review via MCC, resume the same campaign.
13. Run the operating rhythm: 15-minute daily traffic call, weekly creative call (hook rate, CPA, winners mapped to IDs) and VSL call (retention, AOV, best-of-week comparison), plus a Monday launch sprint; setup: master metrics spreadsheet + ClickUp/Notion calendar feeding the next week's copy.

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

| Tool | What it does here |
|---|---|
| MultiLogin (antidetect) | One isolated browser profile per farmed account; proxy configured inside it, never the home Wi-Fi |
| Residential proxies (Proxied, GO Contingencia-style) | Rotating IPs 12 16 shared across accounts so creation never looks robotic |
| Google Ads MCC (master + sub-MCCs) | Tree structure per manager, ~500 accounts without cascade drops, all recoveries requested through it |
| Gmail marketplaces (GG Max, ACC Market) | Aged inboxes at R$5 7 to skip the 5 7 day phone-verification struggle; a parallel farm framework |
| Google Ads campaign types | Maximize conversions to unlock fresh accounts, target CPA to run them; demand-gen is now the only campaign shell |
| VTurb | Sales-video hosting and VSL metrics (retention, play rate) discussed in the weekly VSL call |
| Metrics spreadsheet + BI/automation | One ID per ad with CPA, ROAS, CPC, IC copy and editing teams live in it; real-time profit queries via automated WhatsApp |
| Spy / format research | Feeds the biweekly creative call: which formats scale, presented by the research team to copy, editing and traffic together |

## 5. Actionable Checklist

- [ ] Stock accounts before you need them: 200+/month farmed, 300+ always sitting idle
- [ ] Run three farm frameworks at once and randomize/rewrite the process every 1 2 months
- [ ] Warm up in two stages: 1 3 days white spend, then 7 15 days on your real domain and structure
- [ ] Treat "40 100 approved ads" as the warm-up exit criterion, not a number of days
- [ ] Keep every account alive with a weekly human touch duplicate, pause, change something
- [ ] Validate every creative on Facebook first; take only winners (CPA target, 10 15 sales) to YouTube
- [ ] Adapt winners for YouTube: remove music, subtitles and aggressive clickbait, keep it organic
- [ ] Launch with a tiny budget and maximize conversions; switch to target CPA only after it spends
- [ ] Never touch the domain, page or structure of a live account
- [ ] Hold budget until 10 15 conversions, then step +30% or double in 2 3-hour windows
- [ ] Scale only after three straight doublings; watch today's hourly sales, not yesterday's ROAS
- [ ] Build the suspension playbook first MCC reviews daily, resume the same campaign, keep white ads as backup