Robby Frank: 5,700 followers, $265K a month. The sell-first playbook, his 7 ad rules, and how Claude runs the whole thing
Robby Frank has about 5,700 followers on X. Marc Lou, one of the best known indie hackers there, has over 411,000. Robby's company makes about 5 times more a month, verified through Stripe. His point is blunt: followers are builders, buyers come from somewhere else. This page breaks down what he sells, where his customers come from, and the exact moves you can copy, with a fill-in test grid and word for word emails. On October 20 he shared the rest: the 7 rules behind $180K in 60 days, and how about 80 Claude Code sessions run the business while he picks the stories.
Overview
On October 18, 2026 Robby posted two TrustMRR screenshots side by side. Marc Lou: 411.8K followers, $53,624 a month across 20 startups. Robby: 5,672 followers, $265,415 a month from one. His caption: "If your plan is 'grow an audience first, sell later', you picked the slowest way to make money on the internet."
Maybe 0.1% of his customers come from X, he says. In his words: "X brings builders, ads bring buyers."
What he sells
His biggest company, 1Lookup, checks whether phone numbers, email addresses and IP addresses are real, in real time. It bundles 34 data tools behind one API key and one shared credit balance, and customers pay for what they use, with a 7 day free trial.
Who buys it: call centers, marketing agencies and online stores. Businesses that call or email a lot of people and lose money every day on bad numbers and fake signups. It plugs into Twilio, HubSpot, Shopify and Salesforce.
TrustMRR, which reads revenue straight from Stripe, showed 1Lookup at $263,064 a month, $429,582 in the last 30 days and $5,442,882 all time, with 713 active subscriptions, founded January 2022. That is roughly $370 a month per paying account.
The move: sell something boring that a business needs every day, to businesses that already spend money on the problem.
Where his buyers come from
Not followers. "Ours come from paid ads and SEO."
He picks the channel by who is buying. "Industry decides it, every time." For consumer products: Meta ads, "by a mile." His newest company did $180K in its first 60 days, about 95% of it from Meta ads, with zero followers behind it. For business buyers: Google for people already searching, and cold email for the ones who are not.
On a small budget he uses pages that catch people already searching, like a page for every "{competitor} alternative" search, plus cold email. "Both cost time, not money."
Your first ad, his way
Asked how much you need to spend to compete with people who have big followings: "Less than you think. Start one campaign at $100 a day. An ad that sells pays for the next one, so the budget grows out of sales, never your savings."
Where and what: "Meta, and a 15-second vertical video that shows the product doing its job in the first second. Phone-shot beats polished." His best still images are "a phone photo of a sticky note or a receipt with one line on it." Ads that don't look like ads.
The most common mistake: "changing things every day. Give each ad about $250 before you judge it, and judge it by sales, never clicks." And if your price is low: "The kill line is what a customer is worth to you. At $49 lifetime almost no ad gets there, so fix the price before you touch the ads."
Show the price first
"Sell first means you show the price first." In one of his products, people who looked at the pricing page before signing up started a trial 31% of the time, against 11% for everyone else. "The people who check the price are your buyers. Build for them."
His view on feedback from people who never pay: "noise." "Your first 10 paying customers will teach you more than 10K followers ever will."
The 5 x 5 test
"Every test round gets at least 5 clearly different versions (layouts, offers, angles). Five tweaks of one page teach you nothing."
His team runs rounds of 5 audiences times 5 whole angles, and 3 of the 5 angles are free offers they deliver by hand after someone replies.
Asked whether to sell to 30 agencies at $299 a month or hundreds of founders at $59, he picked agencies: 34 of them gets you to $10K a month, while at $59 you need about 170 founders.
If you have no money
"Broke at 20 is exactly when you can't afford to wait 2 years for an audience. Sell by hand first: DMs, cold email, the communities your buyers already sit in. The first sale pays for the first $20 ad."
His advice to someone selling AI services: drop the automated "how can I help" DMs. Pick 50 agencies, find one thing each visibly struggles with (a slow site, no case studies, a dead blog) and email the owner about 80 words, no links.
The 7 rules behind $180K in 60 days
His newest company made $180K in 60 days, from August 9 to October 4, with about 95% of it from Meta ads. His Stripe chart shows it. He opened the post with: "Most founders will read this and still burn their money."
- $100 a day per campaign.
- One angle per campaign, so you know what sold.
- No sale by about $250, the ad is off.
- Never edit a winner.
- Before you kill a losing campaign, read it ad by ad and keep the one ad that sells.
- AI makes the ads and runs the A/B tests.
- Test hundreds, because most of them lose.
The first $100K took 46 days. The second took 16.
When the price can't carry the test: asked whether ads still work for a $15 product, he said price matters a lot. "If one sale can't cover about $250 of testing, ads stay off until the price can carry it. That's why most of our products are monthly subscriptions." And if yours can't: "Then fix the price before you buy ads. Raise it, add a bigger plan, or sell a yearly deal up front. Until one sale can carry the test, sell by hand: cold email people who already pay a competitor."
How Claude runs it, A to Z
Every step runs through Claude Code: about 80 sessions on one AWS server, one per job. Tracking, ads, A/B tests, SEO, support and reports each get their own. The main sessions each have one goal, a number and a date, and push toward it every run.
Claude writes the tracking code, makes the ads, builds the campaigns, reads the numbers and pauses the losers. Robby's one rule for himself: "I approve anything that launches or spends money."
Step 1: tracking that tells the truth about sales
"If your numbers are wrong, none of the 7 rules work."
- Count sales from Stripe. Browser analytics saw only about 85% of their signups.
- Put the visitor's id, the landing page and the UTM tags on the payment itself. That way 98% of their purchases join back to the visitor who made them, so they know which ad sold.
- Filter bots. The day they launched one Meta campaign, 99 "visitors" hit the homepage in 20 minutes with zero signups. They were Meta's ad review crawlers.
- Remove duplicate events before you count anything. Mixpanel's raw export repeats events.
"Claude set all of this up."
Step 2: 10 to 20 versions of every idea
Every concept gets 10 to 20 versions: the same story with a different person, place and first line. Every version is made square and vertical, so Meta can place it anywhere.
All 10 to 20 versions of one concept go in one campaign. "Targeting stays broad and Meta finds the buyers."
What a winning ad looks like, from one of his B2B companies: one person, one high-stakes moment, the problem named in the first 3 seconds, and the product as the way out. Yes, Meta ads work for B2B too.
Step 3: making the ads with AI
Claude writes the scripts and makes the videos with AI video and voice models. What they learned the hard way:
- The hook gets 3 seconds and at most 12 syllables. Any longer and the AI voice rushes it.
- Write the action as full sentences. Video models read stage directions out loud.
- Transcribe every finished video back to text and check it against the script. In one batch, every defect was in the audio.
- Check every new idea against the ads already running. One passed every review and still copied a live ad that had 15,000 impressions.
Step 4: the loop
One concept, 10 to 20 versions, $100 a day per campaign. Any ad with no sale by about $250 goes off. Before a campaign dies, Claude reads it ad by ad and keeps the one ad that sells. Then it makes 10 more versions of the winner.
The part most people miss: "Every mistake becomes a one-line rule that every session loads, so it never happens twice. We're past 400 rules."
His job in all of it: "I pick the stories. Claude does the rest."
The numbers
| Figure | Number |
|---|---|
| Robby's X followers | About 5,700 (5,672 in his screenshot) |
| Marc Lou's X followers | 411.8K |
| 1Lookup monthly revenue | $263K to $265K (TrustMRR, Stripe-verified) |
| 1Lookup all time | $5.44M |
| Paying accounts | 713 (about $370 each a month) |
| Founded | January 2022 |
| Customers from X | About 0.1% (his estimate) |
| Newest company | $180K in 60 days, ~95% Meta ads, zero followers |
| Saw the price first | 31% started a trial vs 11% |
| First $100K / second $100K | 46 days / 16 days |
| Ad rule | $100 a day per campaign, off at ~$250 with no sale |
| Versions per ad idea | 10 to 20 |
| Claude Code sessions | About 80, one per job |
| Rules learned from mistakes | 400+ |
| Signups browser analytics missed | About 15% |
| Purchases matched to the ad | 98% |
Do what he did: step by step
1. Pick buyers who already spend.
Do this: write down 3 kinds of business that already pay for something close to what you sell. If you are choosing between a cheap plan for many people and a bigger plan for businesses, do the math like Robby: how many customers does each need to reach $10K a month?
2. Put the price where people can see it.
Do this: one pricing section on your main page, above the signup. Track who looks at it. Those are the people to talk to first.
3. Sell the first one by hand.
Do this: list 50 buyers. For each, find one thing they visibly struggle with. Send one short email, no links.
The email, word for word (about 80 words):
"Hi [first name], I was looking at [company] and noticed [specific thing: your site takes 6 seconds to load / there are no case studies / the blog stopped in March]. That usually costs [type of business] [the result: leads / trust / sales]. I help [type of business] fix exactly this. Want me to send you a 3 line plan for [company], free? If it's not useful, just ignore this. [Your name]"
4. Run a 5 x 5 test, not 5 tweaks.
Do this: fill in this grid before you spend anything. 5 audiences down the side, 5 whole angles across the top. Make 3 of the angles a free thing you deliver by hand to anyone who replies.
| Audience | Angle 1: save money | Angle 2: make money | Angle 3: free audit | Angle 4: free sample | Angle 5: free 15 min fix |
|---|---|---|---|---|---|
| [Audience A] | |||||
| [Audience B] | |||||
| [Audience C] | |||||
| [Audience D] | |||||
| [Audience E] |
Send 10 emails per box. Keep the boxes that get replies. Drop the rest.
5. Build a page for every "alternative" search.
Do this: list the 5 best known tools or services in your space. Make one page each called "[Competitor] alternative": who it's for, what's different, your price, one button. People searching that phrase are already buyers.
6. Check the price can carry the test, then run the first ad.
Do this: answer one question. Does one sale bring in at least $250? If not, fix the price first: raise it, add a bigger plan, or sell a yearly deal up front. Keep selling by hand until it does.
When it does, put the first sale into one campaign at $100 a day aimed at the audience that replied most in your grid. Consumers: Meta. Businesses: Google search for the problem they type in. Before it goes live, make sure every payment carries the visitor id, landing page and UTM tags, and count sales from Stripe, not from your analytics.
The kill sheet, one row per ad:
| Ad | Angle | Spent | Sales | Decision |
|---|---|---|---|---|
| [Ad 1] | [one angle] | $ | Off at $250 with no sale. Never edit a winner. | |
| [Ad 2] | $ | |||
| [Ad 3] | $ |
When an ad sells, make 10 more versions of it: same story, new person, new place, new first line.
7. Talk to your first 10 payers, not your followers.
The question, word for word:
"Thanks for buying. Quick question: what almost stopped you from paying, and what made you do it anyway?"
8. Turn every mistake into a one-line rule.
Do this: keep one text file called rules.txt. Every time something goes wrong, add one line that would have stopped it. If you use Claude or another AI tool, load the file at the start of every session. Robby is past 400 lines.
Your first 5 lines, word for word:
"1. Count sales from Stripe, never from the analytics dashboard.
2. Any ad with no sale by $250 is turned off.
3. Never edit an ad that is selling. Copy it and change the copy.
4. Every new ad idea is checked against the ads already running.
5. Nothing launches or spends money without my yes."
Want 50 more breakdowns like this?
Real businesses with their revenue, tools, costs and how they grew, plus step-by-step guides to start your own. Free.
Get free accessMore breakdowns
Sources
- Robby Frank, posts and replies on X, October 18 to 20, 2026, including his October 20 thread on the 7 rules and how Claude runs the playbook
- TrustMRR, 1Lookup and Robby Frank founder pages (Stripe-verified revenue), checked October 19, 2026
- HackerNoon, 1Lookup profile
