Claude Has Officially Changed Facebook Ads Forever
A two-step Claude workflow turns a coffee brand's raw order data and customer reviews into a hero product pick and a fully hypothesis-tested ad build.
Posted
5 days ago
Duration
Format
Tutorial
educational
Views
7K
156 likes
57 · 43
Big Idea
The argument in one line.
Scaling an ad account isn't about producing more creative volume, it's about finding which product actually earns repeat customers through statistically confident data, then building every ad around one evidence-backed job instead of an unsupported claim.
Who This Is For
Read if. Skip if.
READ IF YOU ARE…
You run paid ads for an ecommerce brand with at least a few thousand orders of order history to analyze.
You've been optimizing creative around ROAS or CTR and want a data-backed way to decide which product and message to test next.
You want a repeatable process for turning reviews, surveys, and support tickets into ad concepts instead of guessing at angles.
SKIP IF…
Your business has no order history or review and survey data to feed into the process, since every step depends on real customer evidence.
You're looking for a general Claude tutorial rather than a Facebook-ads-specific workflow.
TL;DR
The full version, fast.
This tutorial walks through two free Claude skills for running Facebook ads on evidence instead of guesswork. The first finds a brand's true hero product by analyzing order history with the Wilson Score Interval, a statistical method that ranks products by confidence-adjusted repeat-purchase rate rather than raw sales volume, often surfacing a product very different from the current best-seller. The second, the 3:2:2 Ad Builder, gathers reviews, surveys, support tickets, and existing ad copy into an evidence map, flags unsupported marketing claims, assigns the ad a single funnel job (earn attention, clarify value, or convert trust), and produces three linked ad executions with a written, falsifiable hypothesis for what the test does and doesn't prove.
Free for members
Chat with this breakdown — free.
Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.
Reframes the goal from producing more ad creative volume to scaling the bank account, and previews a free two-step Claude process.
00:22 – 01:06
02 · Finding your true hero product
Introduces the Hero Product skill and the single order-history export it needs (customer ID, order date, product, revenue per order).
01:06 – 01:45
03 · Selling products vs. buying customers
Draws the steakhouse-salad-vs-filet analogy: a first purchase can be highly profitable and still fail to buy loyalty.
01:45 – 02:42
04 · The cash flow farmer mindset
Reframes the operator's core question as whether you can spend more money tomorrow, and shows the coffee brand's 13-product entry report.
02:42 – 04:31
05 · The math behind the pick: Wilson Score Interval
Explains why ranking by raw repurchase percentage is misleading, and how confidence-floor ranking surfaces Single Origin Ethiopia as the real hero product.
04:31 – 06:00
06 · Building the 3:2:2 Ad Builder
Introduces the second skill and its six weighted evidence categories, credited in part to creative strategist Dara Denney's research approach.
06:00 – 07:24
07 · Guided research: reviews and churn surveys
Feeds in 340+ product reviews and 62 churn survey responses, surfacing that price and suitability are two distinct, unresolved objections.
07:24 – 09:03
08 · Support tickets and unsupported claims
Support conversations contradict the survey data on taste and price, and the current landing page and ad copy get folded in as required evidence.
09:03 – 10:15
09 · The evidence map and customer vocabulary
The finished evidence map shows the brand's own core positioning language appears zero times across 400+ pieces of customer evidence.
10:15 – 12:35
10 · The three jobs of any ad: attention, value, trust
Every ad is assigned exactly one funnel job based on evidence, not theory, and the account is diagnosed as having a scale problem.
12:35 – 13:25
11 · Concept and audit: evidence into creative
Five candidate concepts are generated, three are eliminated with stated reasoning, and the surviving two are ranked by strength of evidence.
13:25 – 15:13
12 · The 3:2:2 build: scripts and hypotheses
Three UGC video executions (identity statement, disqualification, near-miss confession) are scripted around one shared strategic center, plus a written, falsifiable hypothesis.
15:13 – 16:23
13 · What happens if the ad loses
The specification states exactly what the test does and doesn't prove, what's in doubt if it loses, and what to revisit next if it wins.
16:23 – 16:52
14 · Are you buying customers?
Closes on the video's central question and points to the paid, kept-current skill versions and a follow-up video on the full campaign method.
Atomic Insights
Lines worth screenshotting.
Repurchase rate ranked by raw percentage alone often rewards flukes: the highest naive rate can come from the smallest sample, not the strongest product.
In one coffee brand's order data, a product outside the top 3 best-sellers had a 35% repurchase rate, 4.3 times the sampler's repeat rate, while pulling in half as many first-time buyers.
The best-selling entry product in a coffee business had an 8.15% repurchase rate, third worst of the 13 products it sold to new customers.
One brand's support tickets and post-purchase churn surveys flatly contradicted each other: 23 of 62 surveyed buyers cited being 'too fancy,' while zero of 18 support tickets ever mentioned taste or price.
A brand's core marketing language ('specialty,' 'seriously about coffee') appeared zero times across 340 reviews, 62 surveys, and 18 support tickets.
Five of nine running ads in one account were built on claims (provenance, altitude, tasting notes) that never once appeared in over 400 pieces of customer evidence.
Every ad only has one job to do: earn attention, clarify value, or convert trust, and mixing jobs in one ad is a common reason accounts stop scaling.
A test without a stated hypothesis is functionally just spending money and hoping, not testing anything.
Writing down what a test does NOT prove, before running it, stops a single winning or losing ad from being read as proof of something outside its scope.
Attribution-based performance data like ROAS can only rank last month's winners. It can't explain why an ad worked, which keeps an account remaking the same creative without learning anything new.
Takeaway
Find the product that buys you customers, then build ads on proof, not opinion.
WHAT TO LEARN
The highest-selling product and the highest-converting ad claim are often not the ones actually earning repeat customers, and both can be tested against real order data and real customer language instead of guessed at.
01Cold open: the agency incentive problem
Agencies are often incentivized to produce more creative and more dashboards, not more profit, so the volume of work an agency shows you can be disconnected from what actually grows the bank account.
The starting question worth asking of any ad account isn't which product sells the most, but which first purchase makes customers come back.
02Finding your true hero product
A hero product isn't identified by opinion. It's found by feeding raw order-history exports (customer ID, order date, product, revenue) into an analysis that traces which first purchase is associated with repeat buying.
Only one file is required to start this analysis: an order-history export with four fields, and most ecommerce platforms can produce it in under a minute.
03Selling products vs. buying customers
Selling a customer a one-time high-margin item and buying a customer with a lower-margin item that builds loyalty are fundamentally different moves, even if the first looks more profitable on paper.
Correlation between a first purchase and repeat behavior isn't proof of causation. The data only shows which first order is associated with people coming back, not that the product itself caused the loyalty.
04The cash flow farmer mindset
The single most useful business question to keep asking is whether you can spend more money tomorrow than you did today.
A high best-seller volume can mask a low repeat-purchase rate. Units sold and returning revenue can point at completely different products.
05The math behind the pick: Wilson Score Interval
Ranking products by raw repurchase percentage rewards products with tiny sample sizes that got lucky. A confidence-adjusted ranking can outrank the store's flashier but thinly-sampled numbers.
A product overlooked because of low unit volume can be the strongest acquisition channel in the business, with a confidence-adjusted repeat rate several times higher than the actual best-seller.
07Guided research: reviews and churn surveys
Customer reviews and cancellation surveys answer different questions. Reviews reveal why people stay; churn surveys reveal what nearly stopped them buying, and both are needed.
A single objection word can hide two unrelated concerns. Two different phrases can both get filed under 'price' or 'suitability' hesitation but need completely different messaging to resolve.
08Support tickets and unsupported claims
Support tickets and post-purchase surveys can directly contradict each other, and when they do, what customers actually asked before buying is usually the more trustworthy signal.
A high-performing ad built on a claim customers never actually make can still post a strong ROAS while quietly acquiring customers who never buy again.
09The evidence map and customer vocabulary
It's possible for a brand's own core positioning language to appear zero times across hundreds of pieces of customer evidence, meaning the marketing and the buyer's actual vocabulary have diverged.
When claims in current ads sit outside any reason customers actually give for buying, real ad budget is likely being spent making a case nobody asked to hear.
10The three jobs of any ad: attention, value, trust
Every ad exists to do exactly one of three jobs: earn attention, clarify value, or convert trust, chosen from evidence about the account's actual bottleneck, not from theory.
A funnel-stage complaint that doesn't reduce to either a scale problem or an efficiency problem probably isn't real; forcing every diagnosis into one of those two buckets keeps the reasoning honest.
11Concept and audit: evidence into creative
Generating multiple creative concepts and then explicitly rejecting the weaker ones, with the reasoning shown, produces more trustworthy direction than picking the first idea that sounds right.
A concept without a testable hypothesis is just spending; every creative test needs a stated 'because' that ties the format back to a specific piece of customer evidence.
12The 3:2:2 build: scripts and hypotheses
Three ad executions using different entry points can all still be testing the same single strategic idea, which is what makes the test results interpretable instead of noisy.
A specification that states what a test does NOT prove is as valuable as the hypothesis itself, since it stops a loss from being blamed on the wrong part of the campaign.
13What happens if the ad loses
Writing down which single assumption is actually in doubt before running a test keeps a losing result from being misread as proof the entire strategy failed.
A winning ad can still expose the next bottleneck. If cold-audience creative outperforms, the very next constraint is often whether the landing page it points to matches the promise the ad just made.
Glossary
Terms worth knowing.
Hero Product
The single product an ecommerce brand should build acquisition ads around, chosen for its statistically confident repeat-purchase rate rather than its raw sales volume.
Wilson Score Interval
A 1927 statistical method for estimating a confidence range around a percentage from a small sample, used here to separate reliable repeat-purchase rates from ones that only look strong due to a small sample.
3:2:2 (ad build)
A creative test structure of three video or image executions, two primary text variations, and two headlines, all built around one shared strategic message.
Ring (ad job)
The single funnel role assigned to an ad, one of earn attention, clarify value, or convert trust, chosen from customer evidence rather than a generic funnel template.
ROAS
Return on ad spend, a platform-reported ratio of ad revenue to ad cost that says nothing about whether the customers it acquired ever buy again.
Evidence map
A structured comparison of every customer-facing ad claim against the actual language customers use in reviews, surveys, and support tickets, flagging claims with no supporting evidence.
UGC testimonial
User-generated-content style ad creative shot to look like an unscripted customer video rather than produced brand content.
blunt reframe of agency incentives vs. operator incentives, works as a standalone hook→ TikTok hook↗ Tweet quote
01:40
“You have to sell the steak, not the sizzle.”
punchy metaphor that stands alone with zero context→ IG reel cold open↗ Tweet quote
02:20
“This is how you stop being a ROAS hunter and start being a cash flow farmer, sowing the seeds of infinite scale.”
reframes the whole video's thesis in one line→ newsletter pull-quote↗ Tweet quote
13:10
“A test without a hypothesis is just spending.”
tight, counterintuitive, and quotable with no setup→ TikTok hook↗ Tweet quote
16:23
“Are you selling products or are you buying customers?”
the video's closing thesis question, doubles as a cold open→ IG reel cold open↗ Tweet quote
The Script
Word for word.
Read-along
Don't just watch it. Burn it in.
See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.
17px
metaphoranalogy
YouTube is littered with ad agency owners showing you how to connect your meta ads to Claude so you can make more dashboards, more creative, and scale the volume of work. That's their job. Your job is to scale the bank account.
Today I'm going to show you the super simple two -step process none of them will ever tell you about. And you can do it right now in Claude for free. It all starts with one thing, your hero product.
Every business has one and almost nobody knows which one it is. So here's how to find yours. What I'm pasting in looks like a prompt and it is, but it's calling a skill, a file you install once and it's just there every time.
And it's free. The link is in the description. It takes 30 seconds to add.
And if you're on the free plan where skills aren't available, the whole thing runs as one big copy and paste prompt instead. Same link, same process, same output, 100 % free. Now watch what it does first.
It walks me through exactly what it needs from me. And it's super simple. Look, that's the whole ask.
If you can export your orders, you're already qualified to do this. So let's give it a file. By the way, this is a coffee business doing a couple million a year out of Washington.
So let's watch the method. Don't worry about the numbers right now. And while that runs, here's the shift you have to make before any of this matters.
You're not selling products anymore. You're buying customers. Those are completely different jobs.
Any steakhouse could sell you a salad for 20 times what it costs them. And even if it's the greatest salad in the world, you're never going back. But give somebody the best filet mignon they've ever had, and they'll bring 10 friends every chance they get.
I can't prove that the steak did that, and neither can this report. All it shows is which first order is associated with people coming back. You have to sell the steak.
not the sizzle. And this is a report nine -figure businesses pay analysts tens of thousands of dollars produce, and we just got it for free in seconds. This is the entire business on one page.
This is how you stop being a ROAS hunter and start being a cash flow farmer, sowing the seeds of infinite scale. 13 entry products and the entire customer journey. Now, three things jump off that page.
First, no surprise. The Discovery sampler is the best seller at nearly sixteen thousand. Of course, it's the sampler.
It's cheap and it's always had the best row as the most ad spend. The second point is really interesting. The best seller has just over an eight percent repurchase rate.
That's third worst in the business. The thing they sell the most is the thing basically least likely to ever get a customer to come back. And the holiday gift box isn't any better.
And then the third one, Panama Geisha, with a repeat purchase rate of over seven percent. That's the best in the business. So now I have to show you why that is a terrible hero product and the reason might be the most useful thing you learn in this video.
In 1927, a statistician named Edwin Wilson figured out a way to apply confidence to percentages, even if you don't have a lot of data. This is called the Wilson Score Interval, and it's the math behind basically every social media platform and some of the biggest businesses in the world to understand what people really want, versus vanity metrics that look sexy on a report.
If you look at the chart on confidence, all the way to the right, the highest thing, there's a blue dot. Single origin Ethiopia. It's not the best seller.
It's not the highest out of context repeat purchase rate winner. It's a third product that we weren't even looking at. The single origin Ethiopia has nearly 8 ,000 customers with a 35 % repurchase rate.
In fact, somebody who starts on Ethiopia is 4 .3 times more likely to buy again than somebody who starts the sampler. And Ethiopia already earns more returning revenue than any other product in the business from half the first -time customer that the sampler gets. So let's be honest about what you actually got.
This report describes a pattern in your receipts, but it does not explain that pattern. The report can't tell you why the winner caused anyone to come back. Maybe the product is genuinely excellent.
Maybe the people who buy it are just the kind of buyer who repurchases from anyone. The data can't separate those two. And I'm not going to pretend that it can.
But here's why I start here. All empirical data says that this is the best way to have the volume of revenue we make from non -first time buyers go up every month. And that is the only definition of success any business operator should ever care about.
That is scaling cash flow. Simply put, the most important question you ever need to ask is, can we spend more money tomorrow? This tells us the easiest way to make sure the answer to that question is always a yes.
So now that you know the best thing to sell we got to make the best ads to sell it and that Brings us to the 322 ad builder. And by the way, these tools both work very well on their own I'm just showing you the easiest way to copy and paste your way to success and for that Use both.
And like before, this is referencing a skill. And you can find the link down below to the page so you can copy and paste it like a prompt if you're on the free version of Claude. So this opens by asking one question.
What are we building this ad for? And it gives you four ways to answer. And any of them starts this process.
I'm going to hand it the hero product analysis we just did because it's objectively the single best offer this business has. But you can pick anything. All right.
Ethiopia. Let's go. Now, two things you need to know before we move any further.
First, your order history knows what happened, but has no idea why. And second, you don't need a full funnel of ads to get started. You only need a couple of ads that do one job.
The research we're about to leverage decides which job that is. And this is the foundation of how we build out the Olympic rings and all of my Andromeda One campaigns. And credit where it's due.
Dara Denny is the master of researcher Facebook ads and I've learned a lot from her and she's a good friend of mine. If you want to see how deep this rabbit hole really goes, I really suggest you check her out. But in this video, we're going to keep it copy and paste actionable.
Now the builder leverages six categories and it offers me two ways to do it. Guided or Braindump. We're going to choose Guided so you can see every step of the way.
Open kimono, full transparency. Reviews come first, and it says outright that this one carries the most weight. For the Ethiopia, we have over 340 reviews, 50 works, 200 is really strong, so let's give it the file and wait for it to do its work.
Okay, interesting here it says so this file gives me a lot on why people stay and almost nothing on what nearly stopped them buying Objections will have to come from somewhere else which makes categories two and three more valuable than usual here The surveys are next and it only wants free text column So let's give it the surveys now this company had a survey of everyone who stopped buying coffee and for the single origin Ethiopia product 49 out of 62 mentioned price or suitability and that's important like it says here.
Those are not the same objection, and they don't get solved by the same ad. So now it's time to give it support.
We're going to upload the tickets, chat logs, and all the comments under the ads. That sometimes takes a minute, but what do we got? Support is all about objections, and we have 18 tickets here.
Not a single one ever mentions taste or price, which is really interesting because this does not corroborate the too -fancy -for -me hesitation from the surveys. Nobody wrote in asking whether the coffee was too advanced for them. Now, notice what all three of these inputs have in common.
Every one of them is using somebody else's words. So now it's time for our own. So we're going to give it our landing page and every ad that we've got running.
And that category is required. In the account, there are nine single ads promoting the single origin, Ethiopia. So we give it the copy and transcript of the video ads.
we scraped the landing page. Now, side note, I used Claude to assemble all of these requests so the files are in MD. This step is vital because the gap between what you claim and what your customers say is often what keeps people in the doom cycle of constantly launching ads but getting nowhere.
It will not accept any performance data because attribution -based vanity metrics don't tell you anything about why something happened. Rank your ads by results, and all you get back is make more of last month's winner. That's the loop we're trying to get out of.
A high -performing ad built on an unsupported claim might have a great ROAS, but it's buying the customers who never buy again, which is an absolute waste of your time and money, and that's reason number 347 that one of the mantras on this channel has always been, fuck ROAS. All right, two left, and they're both optional.
Competitor material, which is... reviews of competitor products and all of their ads. So we can go to the meta library, search their page name, and it's all just sitting there.
Remember, success leaves clues, and so does failure. If you don't have a mode, you don't have a business. And then interviews, actual recorded conversations with customers.
We don't have any of those. And most people don't. And then reviews the evidence and asks us whether anything else is coming before it starts and nothing else is coming.
So Let's get to work. And now we have the first report of the builder.
This is the evidence map, and it's got a lot of really interesting stuff in it. Right off the bat, our core vocabulary shows up exactly zero times in over 400 pieces of customer evidence. And five out the nine ads we're running currently claim something different than what customers say.
This unsupported claim is massive, and the research shows we should probably never say that again. And here's a ton of good information in here, but I want to call out something important. All of this is from the Single Origin Ethiopia product.
And to be fair, Ethiopia happens to be my favorite kind of coffee. And it's easy to think that people kept buying because the coffee was great. But what they're saying is they're buying it because it became a habit.
which is one less decision they had to make this is the thing that gets customers to come back again and again and again and if you notice on this customer evidence map every insight is being mapped to a part of the funnel what job it could serve and how relevant that might be to the creative all right moving on to phase three finding the right customer doesn't fix the account if you ask the ad to do the wrong job and there are three jobs any ad can do earn attention clarify value or convert trust.
There's your funnel. Upper funnel is for scale. Lower funnel is for efficiency.
So the job gets decided first and the format comes after that. Based on the evidence, it recommends that we focus on earning attention, which solves the scale problem. But it also gives me choices.
based on my problem and I can type in whatever I want. Now, fun fact, if you tell you have a mid funnel problem, it won't accept it. Mid funnel is a myth.
Either you're trying to scale or you're trying to get more efficient and you can't make a wrong choice here because it's asking about your ad account instead of asking about theory. So we're going to go with the recommendation. OK, now we get our next report, the ring assignment.
Which job does the ad have? Now that we've agreed, let's look at the data. Now, there's a fun table that talks about all the different jobs that could be done and the decisions we've made and some of the reasoning.
And I like the section below that. What this means in practice, our current ads are filtering out our best customers, which is why we can't scale. And our competitors keep leveraging a feature nobody cares about.
And suitability is a real big gap. This ad only has one job. to earn attention and recognition.
This is going to be an upper funnel ad, but it keeps that data, the price rebuttal, the brew versatility, the bag mechanics. They're also on the map marked and collected, but not load bearing for this job. If the next ring is about clarifying value, those become really load bearing points and nothing gets gathered.
twice. This is how you complete a conversation in a funnel. This is an actual multi -stage sequential messaging strategy.
This is what builds cashflow. This is what scales a business. Now it's going to actually ask us what kind of content we can make video static or both.
I'm going to go ahead and say both. All right. Now, based on the job we're asking this ad to do, it looks like we're committing to video.
And then it's going to give us some reasons why it's making that choice versus others. And it's going to suggest the content form in this case, a UGC style testimonial. It's going to let us know what static could have done that this ad couldn't.
That brings us to the concept and audit report. All right. It built five concepts, then it removed three and it showed us exactly why.
So I could disagree with it. And I like that it gave me the one that it has the most confidence in and the second one that it's holding. And look at what each one carries, the customer evidence, the format, the core message, the objection that it's attacking, and a testable hypothesis.
Because we're using a scientific method here. A test without a hypothesis is just spending. The one that it picks goes directly at who this is for.
It disqualifies an objection, which is fucking cool. And then the last thing on this page. Nothing at all got built on provenance, altitude, tasting notes, brewing qualifications, or free shipping.
Five out of the nine ads that are running right now are built on that material. And the customers have never once mentioned any of it. Which brings us to the final report.
The 3 -2 -2 build. Three creatives, two primary texts, two headlines. Twelve combinations out of one idea.
And every one of those twelve carries the same strategic center. That's the whole point. Three executions, all video, all UGC, all under 25 seconds.
Shot on a phone, vertical, with no color grade. Let's break down these executions. Not a coffee person.
I'm not a coffee person. I don't want to be a coffee person. That's the reason this works.
We also have the disqualification. I don't weigh my coffee. I don't own a scale.
I couldn't tell you what a washed process is. And then they sell it to me. Or nearly didn't.
I almost didn't order this. It looked like the kind of coffee you were supposed to already know things about. That's an identity statement, a disqualification, and a near -miss confession.
Three ways in, one meaning. one concept, three executions. And it's giving us the setting, the script, the editing notes, and the lines that have to be set in the places where you can ad lib.
And every spoken line came out of the review file with a count sitting next to it. And then it did the thing I've been asking people to do for years. Every 322 is a test.
But if you never write down what you're testing, you can't tell me what you've learned. And most people that run 322s don't learn anything. So it wrote the specification.
The hypothesis has a because in it. If cold buyers are excluding themselves because they think they're not serious enough about coffee and 23 of 62 told us they nearly did, then a video that disqualifies expertise brings in first purchases at or below what I'm paying now because it removes the barrier before price is considered.
It names a business result, not a platform metric. And the because is the part that we've skipped on the previous ads. And then it writes down.
what this doesn't test and then it does the thing almost nobody does it writes down what happens if this loses and the list is one item long the concept is in doubt that's it the hero product isn't in doubt that came from receipts and a creative test can overturn it the evidence isn't in doubt people did say those things and a losing ad can't unsay them what is in doubt is one assumption the connoisseurship barrier that is capping cold reach.
And if it does lose, it tells me what to revisit in order all of the different parts of the ad in the order of importance, which is really important. And then it did the same. on the win if this works the landing page becomes the bottleneck it still opens on a single origin ethiopia for a specialty coffee enthusiast an ad that disqualifies expertise pointing at a page that demands it is a real problem that's a lack of continuity now all of this is a massive amount of work mapped out for free before you spend any dollar on any ad and it's giving you everything you need to know to literally copy and paste your way to success.
If you'd rather run them as skills than paste them as a prompt every time, the packaged versions are in Disruptor Academy, kept current as I improve them, which brings me back to where we started. So the only question left is the one you started with. Are you selling products or are you buying customers?
Now, if you want to know how I run my accounts, the Olympic rings, the one campaign method to get you to infinite scale, that's what this next video is all about.
The Hook
The bait, then the rug-pull.
Most Claude-plus-Meta-ads tutorials teach you to produce more creative faster. This one argues that's the wrong goal entirely, and spends the next sixteen minutes proving it on a real coffee brand's order data.
Frameworks
Named ideas worth stealing.
02:42model
Hero Product analysis (Wilson Score Interval)
Ranks entry products by the statistically defensible floor of their repeat-purchase rate rather than raw percentage or unit volume, so a lower-volume product with a large enough sample can outrank the store's best-seller.
Steal forpicking which product to build acquisition ad spend around in any ecommerce account with order history
04:31model
3:2:2 Ad Builder
3 creative executions
2 primary texts
2 headlines
Twelve ad combinations built from one idea, all sharing the same strategic center so the test results are interpretable instead of noisy.
Steal forstructuring any creative test so every variant is measuring the same underlying hypothesis
06:00list
Six weighted evidence categories
Customer reviews
Churn surveys
Support conversations
Own marketing (landing page + ads)
Competitor material
Interview transcripts
Ranks and requires different sources of customer evidence, with reviews weighted heaviest and interviews treated as optional but high-signal.
Steal forauditing any ad account's current claims against what customers actually say
10:15list
The three jobs of any ad
Earn attention
Clarify value
Convert trust
Every ad should be built to do exactly one funnel job, chosen from evidence about the account's real bottleneck (a scale problem or an efficiency problem), not a generic funnel template.
Steal fordiagnosing which stage of the funnel any single ad should target before writing copy
CTA Breakdown
How they asked for the click.
VERBAL ASK
16:23product
“If you'd rather run them as skills than paste them as a prompt every time, the packaged versions are in Disruptor Academy, kept current as I improve them.”
soft pitch delivered only after the entire free workflow has already been shown end to end, framed as a convenience upgrade rather than a hard sell
Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.
Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.
A media buyer pulls up real Ads Manager screenshots to prove one account really went from $30,000 to $1.1 million a month in Meta spend, then breaks down the three things behind it.
A developer who shipped 89 merged PRs in 24 hours breaks down Claude Fable 5.1's pricing, benchmarks and real-world coding behavior against Fable 5 and GPT-5.6 Sol.
A free Claude skill automates guest-post hunting, then the video pivots to the real pitch: Arvow's own paid automatic-backlink feature, which reportedly added 55 links in a month with zero outreach.
A reference-site-to-Claude workflow covering site mapping, AI image generation, mobile optimization, de-AI'd copy, and SEO: the seven levels between a pretty AI site and one that actually converts.
A performance ad-agency founder breaks down the 12 static formats, messaging rules, and design levers behind ads that have spent hundreds of thousands of dollars, then shows the AI workflow he uses to research and draft them.
00:13
00:29
00:44
01:22
01:34
01:47
02:12
02:25
02:50
03:03
03:16
03:28
03:41
03:54
04:06
04:17
04:28
04:57
05:09
05:17
05:35
05:45
06:13
06:25
06:38
06:51
07:03
07:16
07:43
07:48
08:06
08:19
08:29
08:44
08:57
09:22
09:35
09:48
10:00
10:13
10:23
10:38
10:51
11:16
11:29
11:42
11:55
12:06
12:20
12:32
12:57
13:08
13:23
13:48
13:55
14:13
14:26
14:33
14:51
15:04
15:20
15:42
15:49
16:37
16:39
20:48
55:06
08:58
22:56
15:35
23:52