A $20,000 order from the New Mexico State Fair.
That's what landed in Frey Chu's inbox from a directory so unfinished the homepage still said Lorem Ipsum.
No logo polish, identical stock photos on every listing, a WordPress site he'd basically abandoned.
And it still pulled five-figure orders, because nobody else made it easy to compare luxury restroom trailer rentals.
That accident turned into a system. Frey rebuilt the whole thing as a proper online directory business in 4 days, for under $250, using Claude Code. Greg Isenberg had him on the Startup Ideas Podcast to walk through every step, and the episode title tells you the stakes: a $273 a day online directory.
I pulled the full playbook out of the episode. Here's the video, then the numbers.
Three boring directories making real money
Frey opens the episode with a game: he shows Greg a directory, Greg guesses the traffic and the business model.
The three examples tell you everything about why this model works.
Parting.com, a funeral home directory. 61,000 monthly visitors from organic search. Over $1 million a year in revenue (some estimates say $5 million), and they raised $1.5 million back in 2021. The surprise: most of the money isn't the directory. It's Parting Pro, the cremation software they sell to the funeral homes listed on it.
A Place for Mom, senior living. 824,000 organic visitors a month. Revenue estimates run from $100 million to $500 million a year (Frey's gut says around $50 million). The model is classic lead gen: 18,000+ listings, and they take a referral fee or a cut of the first month's rent.
GasBuddy, crowdsourced gas prices. 1.1 million organic visitors a month, roughly $30K a month from ads alone, plus a premium debit card with Mastercard at $10 a month or $99 a year. The data comes from users, and the incentive is a public leaderboard that's been running since 2004. One guy, Prince33 from Gary, Indiana, hasn't missed a day reporting gas prices in over 5,800 days. His prize pool went from electric bikes in 2004 to $100 prepaid gas cards today.
Frey's framework for all three: a directory wins when it helps people save time, save money, or make money. Price transparency in a boring niche hits all of it.
The Lorem Ipsum site that proved the demand
A year and a half before this episode, Frey built PortaPottyMatch.com on WordPress.
It was bad. His words, not mine. Blank-looking listings, the same AI image repeated everywhere, placeholder text on the homepage.
He assumed no one would ever trust it. Then the inbound started:
- A film crew ordering a multi-stall trailer for a movie shoot
- Martha, looking for a reasonably priced unit with a handwashing station
- The New Mexico State Fair, with over $20,000 in portable toilet and trailer orders
People renting luxury restroom trailers spend $1,000 to $2,000 a day for weddings, corporate events and photo shoots. Nobody had built a clean way to compare options. The ugly site proved the demand, and Frey knew he was letting leads slip through.
So once he learned Claude Code, he rebuilt it properly.
4 days, under $250, 2,000 hours saved
The new luxury restroom trailer directory took Frey 4 days to build. Full cost breakdown:
- $100 for his Claude Code Max subscription
- $100 for the raw data
- $50 in Claude API credits for deep cleaning
Under $250 total. By his estimate, the data work alone would have been 2,000+ hours manually. He knows because he used to do it manually: six hours in a chair, clicking into websites one by one to check if a company rents trailers or just porta-potties.
And Frey is not a developer. He'd been doing AI coding for about six months at the time of recording. Before Claude Code, this kind of pipeline meant hiring someone to write custom Python scripts.
The output: around 1,000 pages about one hyper-specific niche. That topical depth is the unfair SEO advantage. You rank for "luxury restroom trailers in Bakersfield" first, then work your way up to Los Angeles.
The 7-step data playbook
Frey says data is where 100% of the directory builders he's watched give up. He's seen a thousand or two try. So the episode is really a masterclass in one thing: turning a junk CSV into a moat.
Step 1: Scrape raw listings with Outscraper. He pulled 71,000 rows of portable toilet suppliers from Google Maps, covering entire states. Cheapest option he's found (he's tried Apify and others).
Step 2: Clean the obvious junk with Claude Code. One prompt: go through these five CSV files, remove listings with no business name or address, permanently closed ones, and big retailers that clearly don't fit the niche. 71,000 rows became 20,000.
Step 3: Verify every website with Crawl4AI. This is the step where Frey used to give up. Crawl4AI is a free, open-source, LLM-friendly web crawler on GitHub. Frey is non-technical, so he literally pasted the repo link into Claude Code and asked it to install it. 15 minutes. Then Claude Code became the brain and Crawl4AI the engine: visit all 20,000 websites, flag only the ones that actually rent luxury restroom trailers. The run took 3 hours in the background and cut the list to 725 verified companies, each with a confidence score. Two years ago that was 1,000 hours of clicking. It's the same shift I keep seeing in every business automations stack: the real cost was always the manual checking nobody wanted to do, and that part just got automated away.
Step 4: Enrich the inventory. What does each of the 725 companies actually offer? Most searchers want 2 or 3-stall trailers (enough for a 50 to 100 person wedding). Frey's hard-won rule: ask for ONE field per pass. His first attempt asked Crawl4AI for inventory, images, amenities and prices in one go, and the quality was garbage. One thing at a time, inspect the output, fix the exceptions, re-run. It took him 2 or 3 passes to get clean data.
Step 5: Images, verified by Claude Vision. Scrape candidate images from each site using alt text and filenames, then have Claude Vision look at them and pick the best three per listing. Cost about $30 and ran roughly an hour for 700 listings. (Both Greg and Frey flag the legal gray zone here: you don't own those photos. Frey plans to get permission as companies claim their listings, and notes you don't need images to rank.)
Step 6: Amenities and filter data. First pass returned nonsense "features" like the word "and". After tightening the prompt, he got real filters: running water, climate control, ADA access. Those become the faceted search on the live site.
Step 7: Service areas. People want to know if a company will drive 100 miles to a field wedding. He split this into city, region and distance range columns, after catching companies that listed three different states.
Then the fun part: hand the cleaned columns to Claude Code, spin up a Supabase database, and build whatever frontend you want on top.
How an online directory business actually makes money
The first question under Frey's last episode was "ok but how do these monetize?"
His honest answer: it depends on the niche, and that's a feature. From the episode:
- → Lead generation (Frey's plan: sell high-value leads to trailer companies)
- → Vertical SaaS on top (Parting.com's cremation software is the real business)
- → Ads once traffic compounds (GasBuddy's ~$30K a month)
- → A membership or fintech layer (the GasBuddy debit card)
- → Affiliate (a guy named Andy in Frey's community built a tap water quality directory with zero backlinks, grew it past 40,000 visitors a month, got into Mediavine and sells water filters through Amazon)
- → A marketplace later (Marc Lou turned TrustMRR into a place to buy and sell businesses)
The directory gets you the traffic, and the business model is whatever you bolt on top. That's the same logic I wrote about in database business ideas: own a dataset people need, then pick your revenue model.
And if you want the SaaS route, the directory basically hands you the customer list. Frey's take is that a niche directory plus a micro SaaS for the businesses listed on it is one of the strongest bootstrapped paths right now.
One warning he repeats, and I respect him for it: if you need money in the next six months, don't build a directory. SEO compounds slowly. He told Greg that when he was desperate for cash he sold stuff at a garage sale instead.
Will AI search kill directories?
Greg asks the uncomfortable question: if people search with ChatGPT and Perplexity, why would a directory survive?
Frey's two answers hold up.
First, directories serve the decision stage, not the discovery stage. Nobody picks a nursing home for their parents by taking ChatGPT's first suggestion. High-stakes choices (health, money, legal, a $2,000 a day wedding rental) push people to compare every option. That's a directory visit.
Second, local intent still returns listings, and AI answers cite sources. In an AI answer you don't get 1,000 blue links, you get three. If your directory owns the data for "luxury restroom trailers near me", you're one of the three. Models can't invent 725 verified suppliers; they have to point somewhere.
His 2026 adjustment: go niche. "Senior living" is a bloodbath, but "senior living with dementia care" gets 1,000+ searches a month on Ahrefs with a fraction of the competition. Same for "ADA bathroom contractor" versus fighting Angi. The more specific the directory, the better it survives AI search, because specific questions need specific data.
FAQ
How much does it cost to start an online directory business?
Frey Chu built his luxury restroom trailer directory for under $250: $100 for a Claude Code Max subscription, $100 for scraped data from Outscraper, and $50 in Claude API credits for cleaning and image verification. The build took 4 days. Two years ago the same data work would have needed a hired developer and 2,000+ hours.
How do online directory businesses make money?
The main models are lead generation (selling inquiries to listed businesses), vertical SaaS sold to those businesses, display ads, affiliate links, memberships, and marketplaces. Parting.com makes most of its $1M+ a year from funeral home software, while GasBuddy monetizes 1.1 million monthly visitors with ads plus a $10 a month debit card.
How long before a directory makes money?
Plan for 6 to 12 months minimum. Frey is blunt about this: if you need income in under six months, a directory is the wrong project, because it depends on SEO compounding. Andy's tap water directory took about a year to pass 40,000 monthly visitors and qualify for Mediavine.
Do I need to be technical to build a directory with Claude Code?
No. Frey describes himself as non-technical and had only been AI coding for six months when he built this. He installed Crawl4AI by pasting the GitHub link into Claude Code and asking for help, wrote his prompts by talking them through with ChatGPT, and let the crawls run in the background.
Steal the rest of the playbook
This episode is one of the clearest "boring business, modern tooling" breakdowns I've seen this year.
If you like hearing exactly how founders build these things (real numbers, real prompts, real $20K orders from state fairs), that's the entire format of my podcast. I interview bootstrapped founders making $100K to $10M a year and get them to open the books.
