Pinned Tweet
We Generated $27 million+ for brands
How?
Through powerful converting creatives.
Now I have assembled all our strategies in 1 doc.
Creatives That Actually Convert
Could charge you $3 069,
But it's FREE!
Want it?
a) Follow me
b) Like and comment anything
c) RT for a special bonus
You won't regret it ;)
Selling supplements on Meta?
Your competitors' customers are writing your next ads. In any market for that matter.
Collect their best and worst reviews with dates, praise and complaints, next to their live ads from the Ad Library.
Say taste complaints went from 1 in 20 to 1 in 6 in 90 days, buyers keep praising how fast it dissolves, and none of their 30 ads mention either.
Buyers leaving them over taste are shopping now, and nobody's ads talk to them. Lead with taste, and the fast dissolve if yours does it too.
Keep it in a sheet or Airtable that grows monthly. Your AI reads it before every batch.
A review scraper collects it (Apify is a great place to start), or Claude Code builds you one. Want the prompt? Message me.
63
We tested a price rise for one of our clients on a separate page, so the page our winning creatives land on stayed untouched.
It still leaked, "You may also like" sent visitors back to the old product at the old price.
2 things decide whether a page test tells you anything.
1. Isolation.
Our rule: never split-test the page your Meta traffic already lands on. Put the test on a new URL behind a new ad, and check every widget, upsell and menu link that leads out of it.
2. The yardstick.
A $59 bundle and a $149 3-pack can't be ranked on conversion rate.
At a 3% and a 2% CVR, after product, shipping, fees and the discount, the bundle leaves about $0.90 a visitor and the 3-pack about $1.40.
So judge each page on contribution per visitor.
Before your next page test, write down what a visitor leaves you on the current page.
That's the number the new page has to beat.
1
70
Meta wants about 50 purchases in 7 days to leave learning. If you sell mid-ticket, at a $100 CPA that's $714 a day for one test.
On $30k a month nobody funds that for every test. So most tests run underfunded, look bad on day 2 and get cut before Meta has found anyone for them, or it found them, but attribution lands late and they look great 1 day after you killed them.
Size tests to read the angle instead:
target purchases × expected CPA ÷ days in the test
10 purchases at $100 over 7 days is $143 a day.
Then 2 kill rules: nothing gets judged before day 7, and on day 7 anything above 1.5x the target CPA gets cut.
Write the budget and the kill line on the ad set before launch, so nobody decides on day 2.
59
10 years ago the edge in Meta ads was media buying: bid tricks, audience stacks, dayparting. Meta took those decisions into the machine, and the "mechanical edge" went with them.
It moved to creative, as you must have heard 100+ times, and most people are terrible at it.
In the accounts we open the claim is right, with one change: the gap sits a step earlier, in what the brief is built from.
2 brands, same AI, same editors, same budget. One briefs from its customers' words, objections and refund reasons, its own ad data and what its category hasn't run yet.
The other briefs from its product page.
The first can find an angle nobody else is using.
The second says what every competitor's AI already says, and pays a higher CPA for it.
List what your last brief was built from.
If the product page is the longest item, that's where the gap starts.
Slop is your enemy, unique creativity is your friend.
5
71
Your best ad is still under break-even on lifetime CPA. Its last 7 days may already be losing money.
Lifetime numbers carry the good weeks, so a fading winner looks fine.
Say it sits at $38 lifetime against a $45 break-even, and ran at $58 the last 7 days. At $1,000 a day, 3 more weeks like that costs about $4,700 above break-even.
That's when to make the next version, while the angle still works.
Have your AI pull your top 10 spenders from the Meta API every morning, and flag any whose 7-day CPA crossed break-even while lifetime hasn't.
Bonus if you save all the data to an external database, and have AI do what its best at, pattern recognition.
52
Meta says 3.1 ROAS, but new customers in Shopify haven't moved?
Part of that ROAS may come from people who already bought from you.
In Advertising settings, add your customer list as existing customers.
Then on any sales campaign: Breakdown > By demographics > Audience segments.
Say 45% of spend reaches existing customers at 4.8, and the rest at 1.7, under break-even. Blended, that's the 3.1.
So the spend finding new buyers loses money, and the 3.1 leans on people who might have come back anyway.
Judge Meta on new-customer ROAS against break-even.
Have your AI pull it from the Meta API weekly and flag when the existing-customer share rises.
You can take it a step further, collect and connect data from Shopify, GA4, Microsoft Clarity - or your heatmapping tool of choice to the AI brain, and deliver the same calculations.
Your business might be entirely different compared what you think.
2
68
Checking Ads Manager every hour? Your AI brain can check it once a morning and report only what breaks.
Connect it to the Meta API and Shopify, and give it 3 rules:
1. Any ad rejected or limited overnight (a disapproved winner stops, and its budget shifts to weaker ads).
2. Any ad whose share of spend doubled overnight (Meta just moved budget onto it: check its CPA holds at the new spend before the day's money is gone).
3. Meta's purchases moving outside their usual share of Shopify orders over 4 weeks.
Nothing broken? It says so in one line. Rule 3 catches a tracking break before you cut ads that were working.
Want the prompt? Message me.
3
1
101
Your account had a bad day and its not the first?
Before you touch a single creative, check the one number that sits outside Ads Manager: the gap between Shopify's orders and Meta's purchases for the same days.
On most accounts it sits in a steady range, Meta claiming a bit more or a bit less than the store. What matters is whether it moved.
If Meta's count jumped while orders didn't, something is counting twice: browser and server events without a shared event ID, or a second app sending the same purchase.
If Meta's count dropped while orders held, the pixel or the click ID broke. The click ID is how Meta ties a sale back to the ad that caused it. Lose it on a redirect, a checkout on another domain or a consent banner, and the purchase arrives with no ad attached. You can't see that in the counts.
Open Events Manager and check how many of your purchase events arrive with a click ID. If that share fell, Meta is optimising on missing data.
Every other number on the screen is built on this, and so is Meta's read of who to show your ads to.
Fix that first, then look at the creatives.
What's the first number you check?
1
1
101
Everyone has the same AI.
The edge is what you feed it: your customers' words, their objections, your ad performance data, which creatives brought in customers with a higher CLV, and your competitors' live ads read against both.
In regulated categories like beauty and wellness, the line that sells hardest is often the one Meta flags, and the flags it doesn't act on can still quietly cost you in CPMs.
A skincare client ran "your dark spots keep coming back". It ran for a while, then Meta pulled it.
Read as a customer, it's a fair line. Read by Meta, it tells the viewer something is wrong with her, and that counts as negative self-perception.
The version that runs: "Dark spots don't form on the surface." Same product, told through how dark spots form instead of who has them.
The fix is a compliance database: the claims you can make, the wording your category gets flagged for, and every rejection you've had. Let AI check every script, landing page and static against it, and have a person check what it passes.
Start with your last 3 rejected ads. Read the first line of each.
1
61
Ads Manager showed a 7.17 ROAS for one of our clients. Their best month had never cleared 6. Too good to be true.
2 systems were sending the same purchase to Meta, and Meta counted both. The real number was 5.57.
Meta learns from the purchases it's told about. A double count teaches it, and you, the wrong takeaway.
2
3
703
There is one thing you can know about your customers that a competitor structurally cannot.
Not the reviews. Anyone can scrape those.
Your own list's answers. Your call transcripts. What a segment believes, and what it converts at.
Nobody can buy that. Do you use it?
1
136
Your AI and your competitor's AI read the same internet.
The difference is not the model. It is what you kept.
Directed research, saved and reused, compounds. Most brands keep nothing and start over every quarter.
What does your account still remember from last year?
123
You automated the two things that needed a human, and hand-built the one thing AI is good at.
Judgment: which belief to sell, when to scale. Yours.
Production: the fastest way to make it. The machine's.
Most brands run it backwards. Which way round is yours?
1
105
A DTC account, run fully on AI for six months.
312 ads. 289 killed under €50 of spend. CPA up 64%.
The tool did what it was told. No one in the loop could tell it it was wrong.
That is the risk. Not bad AI. No owner for the call.
Who owns it in yours?
1
80
Six numbers decide if a DTC brand can scale:
- New-customer revenue (30d)
- Ad spend (30d)
- Gross margin after fulfillment
- 60-day repeat rate
- CAC
- First-order AOV
Most founders can answer three. The other three are where the ceiling hides.
111
Reviewed 40 AI-generated ads for a brand last week.
All 40 opened with the product. Zero opened with the customer's problem in the customer's words.
AI prompted with a product description contains zero customer psychology. Nobody in the loop was qualified to notice.
1
99
Your creative isn't fatigued. Your angle is.
If the last 5 "fresh" creatives died at identical CTRs, stop briefing designers and start rereading your customer reviews.
1
2
95
Blended ROAS is an average. Averages hide the thing that kills you.
Track aMER instead: new-customer revenue / total ad spend.
Two brands at 3.5x blended can be one healthy business and one slow-motion cash crisis.
Comment "SPLIT" and I'll send you the calculator that shows exactly what your blended ROAS is hiding.
2
88





