Most founders would rather ship ten features than get on one phone call.
Gaurav did the opposite. He and his co-founder Joe ran customer discovery calls at a pace of 20 per week, every week. Around 2,000 calls total.
The result: Fastlane, an AI short-form marketing tool that hit $69,000 a month in recurring revenue just over two months after its public launch. Over 1,000 paid users. More than $1M ARR across the whole business.
They didn't buy that growth with ads or some clever hack. They talked to an unreasonable number of people.
Pat Walls had Gaurav on Starter Story to break down the whole system. I pulled it apart so you can steal it.
From Watching a YouTube Video to $69K a Month
A year before this interview, Gaurav wasn't a founder. He was a guy in Australia watching Starter Story videos.
One video about Puff Count (Steven's vape-quitting app) hooked him. He started building his own apps the same week.
Then he hit the wall every builder hits: nobody saw them. He didn't want to burn money on ads. Around the same time he met Joe, who was stuck in the exact same place with his own SaaS.
So they went after the problem itself: marketing for solo builders.
The first version was called Cast AI, and it tried to do everything. SEO, getting cited by LLMs, Reddit engagement, short-form content, all in one tool. They started in July, shipped V1s through August and September.
Too wide. Customers kept pulling them toward one thing: short-form content.
So they cut everything else. Rebuilt around it from October. Shipped a beta in December to a 2,000-person waitlist. Then launched publicly as Fastlane.
Two months later: $69K a month, up from roughly $16K ARR in paid beta revenue.
Every one of those decisions (niching down, what to cut, what to keep) came out of customer calls.
The Rainbow Button That Booked 2,000 Calls
Here's the part I love. Getting 2,000 strangers on the phone sounds impossible.
Their mechanism was a button.
Inside the Fastlane UI sits a button with a rainbow border: "Book a call with the team if you want 7 days of extra access."
Click it, land on a Calendly, pick a slot. That's it.
Users got extra access. Gaurav and Joe got a steady stream of live interviews with real users. They sat in a room together, "harassing people with questions" as Gaurav puts it.
→ You don't need a research budget. You need an incentive worth 15 minutes of someone's time.
The 3-Phase Call System
They weren't winging these calls. Gaurav runs them in three distinct phases, and each phase asks different questions.
Phase 1: Discovery (The Mom Test)
Before building anything real, they ran pure discovery calls. Personal network, Twitter DMs, whoever they could reach. Gaurav says pay people for their time if you have to.
The rules come from The Mom Test, the book by Rob Fitzpatrick. Never ask "would you use this?" or "do you like my idea?" People lie to be nice.
Ask instead:
- How are you solving this problem right now?
- Walk me through the last time you solved it.
- How much time or money did it cost you?
- What happens if you do nothing about it?
When the same painful answer keeps showing up across calls, that's your signal to build.
Phase 2: Usability (Sit Down and Shut Up)
Once the MVP existed, the calls changed. Gaurav calls this stage "fighting for customers one-to-one."
They crawled Reddit threads and posted lead magnets across social media. The moment someone joined the waitlist, they got an email inviting them to a call. That follow-up email is textbook lead nurturing, and it did more than any ad they could have bought.
On the call: send the link, ask them to share their screen, then stay quiet.
If a user hunts for the login button and can't find it, don't help. The friction IS the data. Jumping in to say "just press this" destroys the very thing you're there to learn.
Those silent calls exposed so much friction that they reshaped the entire MVP.
Phase 3: Customer Success (Find the Power Users)
After launch, the question changes again: is this actually helping anyone?
They tracked almost everything in PostHog, found the power users, and got them on calls. Same Mom Test style, but now digging into where the value showed up.
The answer was clear: views, conversions, app installs from short-form content. So they doubled down on exactly that.
The UI Pivot the Calls Paid For
The best proof this system works is Fastlane's interface.
It looks like Tinder.
In the middle: a video Fastlane generated for your product. On the left: the trending post in your niche it was modeled on (one example on the call had nearly 3 million views). Swipe left to reject. Swipe right, hit schedule, and it posts to TikTok, Instagram, or YouTube inside 5 minutes, captions written by AI.
Nobody whiteboards their way to a Tinder UI for a B2B marketing tool.
You only get the confidence to ship something that weird after watching hundreds of users struggle with a normal interface. In hindsight it looks obvious. It took months of silent usability calls to earn it.
I see the same pattern in founders I interview: the unlock is rarely a new channel, it's finally understanding the customer they already had. It's the same lesson behind the $1.5M app built for a market nobody sees: be willing to let your first idea die.
2,000 Calls Plugged Into Claude
Then they did something very 2026 with all those calls.
Every call had an AI note taker in it. Notes went into Notion, structured around their question framework. Hundreds and hundreds of records.
Then they plugged the whole database into Claude and built what they call Fastlane Customer Intelligence: an internal dashboard that centralizes everything about their customers.
What it can do:
- Segment users by why they signed up. Signups who picked "just curious" retain far worse than the ones who picked "I need marketing now".
- List every paying customer by business type, subscription length, and company.
- Score every customer with a "customer love score" so the roadmap gets indexed on what the happiest users are asking for.
- Surface who's actually winning (who got a million views on TikTok, who got 200K).
The kicker: Joe, the non-technical co-founder, built it himself. Claude Code (they run the 20X Max plan), a few MCPs pointed at their data, plain English instructions. No engineers involved.
The rest of the stack is boring on purpose: Convex for the backend, Vercel for the frontend, Railway for the content microservice, Clerk for auth, Resend for email, Axiom for logs, OpenAI and Claude models for copy.
How to Run This Playbook Yourself
You don't need 2,000 calls to start. You need this loop:
- Put a call incentive inside your product. Extra trial days, extra credits, whatever costs you nothing and is worth something to them. Wire it to Calendly.
- Run Mom Test questions only. Past behavior, real costs, current workarounds. Zero "would you use this?".
- Do silent usability calls. Screen share, mouth shut, notes on every hesitation.
- After launch, only interview power users. Ask where the value showed up, then cut everything else.
- Store every call somewhere structured. Even a Notion table beats memory. Pipe it into Claude when it gets big.
Twenty calls a week is 4 a day. That's 2 focused hours. Most founders spend more than that tweaking a landing page nobody visits.
And if getting the first users to even show up is your bottleneck, I broke that down separately in how to get your first 100 SaaS customers.
FAQ
How many customer discovery calls should a founder do?
Gaurav's answer is 20 a week, forever. You don't need that to start. Keep going until you stop hearing new answers: when call 15 sounds exactly like call 14, you've found your pattern. Most founders I talk to quit after a handful, which is exactly why this works as an edge.
What questions do you ask on a customer discovery call?
Ask about past behavior, never hypotheticals. How are you solving this today? Walk me through the last time it came up. What did it cost you in time or money? What happens if you do nothing? Skip "would you pay for this?" entirely. People say yes to your face and never enter a card number.
How do you get customers to agree to a call?
Trade value for time. Fastlane's rainbow button offered 7 days of extra access and booked 2,000 calls through a plain Calendly link. Pre-product, use your network and Twitter DMs, and pay people if you have to. Paying for honest answers is the cheapest market research that exists.
Did talking to customers really grow Fastlane to $69K MRR?
Not directly, and that's the point. The calls didn't sell subscriptions. They told Gaurav and Joe what to build (short-form only, not five tools in one), what UI worked (the Tinder swipe), and where customers hang out. As Pat put it on the episode, talking to customers is the mechanism for figuring out where to market.
Steal the Systems, Not Just the Story
Gaurav's "secret" costs nothing and almost nobody does it. That's the entire arbitrage.
Every week on the Profitable Founder Podcast I get bootstrapped founders to open up their systems like this: real numbers, real playbooks, no fluff.
