Four Meta ads videos, distilled for FiftyStrong

None of the four speakers shows verified client data. Three run agencies, and the videos sell their services, a tool or a PDF. Treat their numbers as claims.

His order of levers. Fix these in order, and only then scale creative (1:31):

1. The offer and positioning. Say why your app beats the alternatives for this audience.
2. LTV and first-month cash. A high LTV that takes a year to collect needs funding. You want a high order value, or most of the money in month one, so that you can pay a high CPA on Meta (2:01).
3. The quiz. It gives the highest order value and LTV of any app funnel he runs (3:01).
4. Upsells. CPA rises as you scale, so upsells keep the order value ahead of it (3:30).
5. Creatives. They become the main lever once the funnel works (4:00).

Ads for apps sell a future self. An app has no physical product. So the ad must make her see the person she becomes after using it (4:00).

The format follows the age. Young audiences get memes, POVs and fast edits. Problem-aware and older audiences get direct-response formats. He says "for older audiences … video sales letters work pretty well" (6:04).

Two ad strategies (6:32):

• Sell the app. A 30–60 second video: the problem, the solution, or a comparison with alternatives. This works best when the niche is not saturated.
• Sell the mechanism. Explain why she has the problem, in a video of up to 3–4 minutes, without naming the app. The quiz then reveals the solution, fitted to her answers (7:01). His example reframes the blame: the problem is not her discipline, it has a cause she can fix (15:32).

Formats that work for his apps: fake podcasts, street interviews, an authority figure in a podcast set, skit conversations, and simple music-plus-text videos (8:00, 30:30). Each copies a format that already works in organic content.

Andromeda made winners smaller. One ad no longer spends $1M a month. Instead, 10–30 ads share that spend (8:31).

Treat each persona as its own business. Give each one its own concepts, its own quiz path and its own sales page (9:31). Map the creative work as persona × awareness level × angle × format. Then double down on the cells that work, and ask why the others failed: was it the angle or the format (10:01)? In weight loss, fully unaware audiences are very hard. Stay problem-aware or solution-aware (11:31).

Quiz craft (19:00):

• Put social proof between questions: user counts, short testimonial quotes, short videos.
• Near the end, show people whose situation is like hers who reached the goal (20:31).
• Customise the sales page to her answers. Start with two or three versions, and add more as you scale (21:01).

Sales page: a 10-minute discount timer (our coupon paywall already has one), a money-back guarantee tied to using the app, then upsells. One upsell he shows costs $44 more. An upsell bought by only 10% of buyers still raises the order value (22:01).

Find your bottleneck to more winners. Do you make too few ads? Is your win rate low? Or do all your ads reach the same audience (28:31)? He prefers 10–50 strong concepts a month to thousands of AI ads (29:32).

The thesis. "Make 15–20 creatives and let Andromeda pick" works for accounts that spend thousands a day. At $30 a day it spreads the money so thin that no ad can prove anything (0:00). In his example, one ad took about $450 of $586. Most of the others got under $10, and some got cents (5:00). Some of the ads Meta ignored turned out to be the best ones once they got spend (5:32).

His three campaigns, in order:

1. Creative test (CT). Use one campaign budget, broad targeting and 16 ads (4:31). Give each ad at least 1,000 impressions, and 2,000–3,000 at about $80 a day (6:30). Turn each ad off by hand before it passes 3,000, so that Meta has to spend on the others (7:30). The test takes about a week.
2. Landing test (LT). Use the winning ad in every ad set, with one ad set per landing page and an equal daily budget each (15:31). A page built for one persona's pain points won (17:00).
3. Scale. Run an Advantage+ sales campaign and some interest ad sets, with only the top 3–4 ads and the winning page (19:00).

How he reads the test. With equal exposure, compare CTR (all clicks), link CTR and CPM (8:00). A 6% CTR against 2% means three times the attention. One ad paid more than three times the CPM of another for the same impressions. Then check the funnel steps: add to cart, checkout and purchase (9:31). A winner is an ad that finished its impressions and produced buying behaviour. A 5× ROAS on $12 of spend is not a winner (10:31). Keep a shortlist of 2–5 (11:01).

Cut the fat. On a small budget, every weak ad competes with a strong one for the same dollars (11:32). He cut a $6M brand from 100+ ads to its top 30. Revenue rose 45% on the same spend (12:30). "The structure has to match the amount of money available" (18:31).

Scale slowly. Do not push a 16× ad to $150 a day overnight. Add about $20 a day at a time (21:30). For each new batch of about 10 ads, restart the test campaign on the side. Promote a new ad only if it beats the ads already running (22:01). Move up to volume as the account grows, never before (23:01).

The cheapest growth is after the click. Find the funnel step below your threshold (23:30). His thresholds: at least 10% of product views add to cart, and no more than half of carts drop before checkout (24:01).

1. Statics over video (0:00). Andromeda needs a large, steady supply of fresh ads, and statics are faster and cheaper to make. His rough native statics beat a polished VSL that took 2.5 months to make (2:00). Block one hour a week to make new ones.
2. The one-keyword hack (2:30). Duplicate the winning ad. Add one niche or audience word to its headline or body. He calls the word an "identity trigger". It tells Meta whom to look for, and opens audience pockets a generic ad misses. Judge it on CTR.
3. The winning-format clone (5:00). Feed the winning ad to an AI, and ask for versions in the same voice for different audiences. Do the body copy first, then the headlines, then the creative. Put them all in one campaign-budget campaign. Then move the ads you believe in but that got no spend into a separate "zombie" campaign. He says it finds about 20% more winners (6:30).
4. Don't make ads look like ads (7:00). Run content that already did well organically. Open a fresh "burner" Instagram and TikTok account, and follow the big creators in your niche. Then copy what rises in that feed (8:31).
5. Broad targeting with long, specific copy (9:02). Long copy gives Meta more words to learn who the ad is for. In his words: "Broad targeting with super specific creative is the new meta." To test it, run your best ad twice, side by side: once with interests, once with no targeting. Compare CPA after 7 days.
6. Match the ad to the landing page (11:00). Run 20–40 headlines per ad, then put the winning headline on the page. Do the same for the subheadline and the lead-in. He claims at least a 15–20% lift. Always have three split tests running.
7. Retarget with a different offer (13:00). She did not buy because the offer did not fit. Show her ads that answer her objections, proof carousels with testimonials, and other offers (14:31).
8. Track net cash, not ROAS (15:30). A lower ROAS on a much bigger spend can make far more money. Know your break-even, and scale toward it. Spend 3 hours a month checking the numbers yourself.

He splits growth into four spend stages. Each stage has one job.

Stage 1: we are here.

• Structure and bidding are not the lever yet. "Creative is the biggest lever," and psychology outlasts any platform trick (0:31).
• Meta is a feed of content, so make the ad look like what she already reads. His examples are viral news and gossip pages (2:00). "Consumption precedes conversions": more people must read the ad before more people can convert (2:31).
• His three-part hook (3:00). First, a pattern interrupt stops the thumb. Second, intrigue makes her want to know more. Third, a big specific benefit speaks only to the target audience.

His ad checklist (4:00):

Stage 2 (8:00). Run two campaigns only: testing, then scaling for the winners. Target broad. His own example is a marketer who wants only women aged 40 to 55. He says to let the creative target them, with long, specific copy (9:01). Upload your lead and customer lists as exclusions at least every 30 days. Otherwise, he says, about 40% of the budget reaches people you already have (9:30).

Stage 3 (10:30). Scale sideways, not up. Use one-keyword clones by age bracket or niche, and AI versions per demographic in one campaign-budget campaign. Add a zombie campaign for good ads that got no spend. Match the winning headline on the landing page; he claims a 30–100% lift (14:00). Use what wins in the ads in your pages and email subject lines too (15:00).

Stage 4 (16:00).

• Cheap early winners are often false positives that only reach the hottest buyers.
• Cold traffic needs story ads, not offer ads. A founder story is his go-to (17:00).
• Set a cost cap at 80% of the most you will pay, with a daily budget twice the real one (18:33).
• Judge total revenue against total ad spend across all channels. Then judge net free cash flow (20:31).
• Know your maximum customer cost at 30, 60, 90 and 180 days (22:30).

Our reading. All three put the page after the click on the same level as the ad. That supports the plan's order: fix quiz screen 1 and the paywall before adding budget. The statics-or-video split is about cost, not truth. We test with statics because they are cheap, and keep the video winners. Our best checkout ad, Beginner friendly, is a video.

The full plan, with benchmarks and kill rules, is on the playbook page. The full transcripts are in deep-research/runs/meta-ads-playbook-20261021-113254/raw/youtube/.