Finding Product Market Fit 3 Times: Andrew Antos on the 30 & 100 Rule

Andrew Antos spent two and a half years building a legal AI tool that 'kind of worked' — and learned that incremental growth is often a polite way of saying it's not working. After three distinct product-market fit moments, he's distilled a repeatable framework any founder can apply.

Key Facts

• PMF validation number: Talk to 30 people to validate or invalidate any new idea (Andrew Antos)
• True PMF threshold: 100 paying customers before the sales motion reveals itself (Andrew Antos)
• First PMF moment speed: $350k ARR in under 6 weeks after launching a new product at a customer event (Andrew Antos)
• Experiment success rate: 9 out of 10 experiments don't work — the 1 that works carries everything else (Andrew Antos)
• Marketing channel focus: Zero Google ads; entire marketing budget concentrated on events as the single working channel (Andrew Antos)

Why 'Kind of Working' Is the Most Dangerous Place for a Startup

A product that grows slowly but consistently can trap founders for years. If you're not a top-priority problem for enough senior buyers, growth will cap naturally — no amount of sales optimization will fix a market-size or urgency problem.

Andrew Antos spent two and a half years building an NDA review tool for legal teams. It reached nearly $800k ARR. By most measures, that's something worth defending. But he and his co-founder were deeply ambitious, and the math was clear: they were growing 20–30% per year toward a ceiling, not compounding toward something big.

The realization wasn't dramatic. It was a pattern they noticed: lawyers kept asking 'why would I care about AI?' That buyer resistance pointed to a structural problem. As Antos put it, 'People just didn't want this at a high enough rate. It wasn't that thing.' The product worked. The market just didn't have enough urgency.

This is what Antos calls a market problem versus a product problem — and it's a critical distinction. Founders in this position often assume sales is broken. They hire a new rep, rewrite the deck, A/B test the demo. But as Antos observed, you can optimize a conversion rate from 40% to 50%, not from 10% to 50%. That leap requires a different market or a different problem.

"You need to be number one problem for lots of very senior people if you're selling into enterprise, so you can grow fast enough, so you have high enough deal size and you have lots of deal velocity." — Andrew Antos

"For a lot of people, they would just call it a good lifestyle business. But both my co-founder and I are extremely ambitious, so we were just like, this is clearly not what we want to do." — Andrew Antos

• Lifestyle-business growth (e.g., $800k → $1M → $1.3M) signals a market ceiling, not a sales problem.
• Buyer resistance ('why would I care?') is a market signal, not a messaging problem.
• You need to be the #1 problem for enough senior buyers to drive enterprise deal velocity.
• Two and a half years is long enough to know — run more experiments, don't just optimize.

The 30 and 100 Rule: A Framework for Validating Any Idea

Talk to 30 people to validate or kill a new idea — 10 is too easy to manipulate with friendly or hostile samples, and 50 is too slow. Then reach 100 paying customers before trusting your PMF read. At 100, the motion, the ICP, and the expansion patterns become visible.

When Antos pivoted from legal to finance teams, he didn't guess his way into it. He built a hypothesis document in writing — 'putting things in writing really matters, because once you see it, you start iterating through it' — and then ran 30 conversations before committing resources to build.

The 30-conversation threshold is specific for a reason. Ten friendlies or ten skeptics can send you in completely the wrong direction. Thirty gives you a distribution wide enough to see genuine signal. And it's achievable in a couple of weeks, which means you're not burning months on a dead end.

The 100-customer rule applies after you've launched. Antos reached 8 customers and $350k ARR in six months during his second PMF — encouraging, but not sufficient. It was only after the third product launched and raced to similar numbers in six weeks that the comparison made the signal undeniable. 'You truly find the core of the product market fit once you sell a hundred customers,' he said. 'Around a hundred, you actually have a hundred goes at it and then the motion kind of reveals itself.'

"You want to talk to roughly thirty people to validate an idea or invalidate an idea. Ten is not enough because you can always find ten friendlies or ten unfriendlies. Fifty is too many. But you can easily have thirty conversations in a couple of weeks." — Andrew Antos

"You truly find the core of the product market fit once you sell a hundred customers. It doesn't mean everybody's super happy and everybody's at a huge deal size but around a hundred, you actually have a hundred goes at it and then the motion kind of reveals itself." — Andrew Antos

How to Spot True PMF: Qualitative vs. Quantitative Signals

Early PMF is qualitative — people at a launch event want to talk about one thing, ask how to buy it, and suggest pricing on the spot. Later PMF is quantitative — usage frequency, deal size consistency, customer expansion patterns, and sentiment trends all move in the same direction.

Antos launched his third product — a tool that turns recorded Zoom calls about business processes into structured documentation — as a side widget at a 500-person customer event. He expected the audience to care about the complex features he'd spent a year building. Instead, every conversation circled back to the widget. 'Every single person, every single discussion was like, oh, that little thing that you're calling the architect is really, really, really what I want.'

That qualitative signal was enough to start selling. In the first six weeks, he ran 12–14 deals, each with a different pricing model, because he was iterating through pricing in real time. This messiness is a feature, not a bug — it's evidence of genuine demand meeting an early product.

As the company scaled past 100 customers, signals shifted to quantitative. Antos built a 'Customer 360' database combining call recordings, usage data, support tickets, and pipeline progression. Agents run on top of this data daily, surfacing sentiment trends by industry, flagging when customers describe problems differently, and tracking whether a specific feature (like their MCP connector) is hitting adoption targets. The rule: 'What would success look like?' gets defined before an experiment launches, not after.

"Once you become a little bit bigger, once you launch more product, it becomes tricky and you need to be very quantitative about it. What's the startup metric that would give us confidence to basically resource it on an ongoing basis?" — Andrew Antos

"Industries where the customer is saying the same words effectively again and again are usually industries where you can grow faster than when you see a lot of entropy." — Andrew Antos

Running 4–5 Experiments Per Quarter to Find What Works Exponentially

Antos runs 4–5 time-boxed bets per quarter, each with a pre-defined success metric and a single owner. The mindset shift: 9 out of 10 things won't work. The goal isn't incremental improvement across all 10 — it's finding the 1 that works exponentially and resourcing it heavily.

The company doesn't plan beyond 90-day cycles. Each quarter, the extended executive team reviews customer data, identifies patterns, and selects 4–5 hypotheses — some product-shaped, some go-to-market-shaped. Each gets a named owner and a specific metric. For their MCP connector launch, the target was 100 active users in 90 days, defined by frequency and depth of use. That clarity makes the call easy: resource it more or kill it.

Antos applies the same logic to marketing channels. His team runs zero Google ads. Their entire marketing budget goes to events — their own and third-party. When asked if he worries about single-channel dependency, his answer was direct: 'No, because the channel is working super, super, super well. I would worry if it was not working.'

The broader principle is a rejection of the 'cover all bases' playbook that works for large companies but kills small ones. Doing 20 incremental things that everyone else also does doesn't compound. Finding one thing that distributes exponentially — like a pop-up data art gallery in Palo Alto commissioned to launch a research report — is what creates asymmetric growth. Most of these bets fail. The ones that hit, hit hard.

"My mindset has completely shifted. I think it's nine out of ten things don't work and they might look like they're working incrementally but that really means that it's not working. But the one out of ten things works so well that it carries everything else." — Andrew Antos

"You need to continuously find completely novel ways to address the market to get that outsized distribution and value benefit." — Andrew Antos

• 90-day planning cycles only — no longer roadmaps at early and growth stages.
• Each experiment needs one owner and one pre-defined success metric.
• Evaluate ramp rate early — you rarely need 90 days to see if something is working.
• Single-channel focus beats omni-channel mediocrity when the channel is working.

Product-Market Founder Fit: Why the Right Product for the Wrong Founder Still Fails

Antos identified a second, underappreciated PMF dimension: whether the founders are the right people to build this specific business. His finance AI product grew 2–2.5x annually, but the delivery model required professional services expertise the team didn't have — and struggled to develop.

The finance automation business was growing. It reached $10M ARR. But Antos noticed something that data alone wouldn't surface: other people ran professional services implementations with 'elegance and lightness,' while his team struggled with it every single day. That gap wasn't closeable through hiring or process — it was a founder-fit problem.

When the new product launched and immediately attracted buyers without any of the implementation friction, the contrast was stark. The team sold it faster, enjoyed building it more, and customers adopted it without the heavy lift. That combination — faster growth plus founder energy alignment — made the pivot decision easy despite having $10M ARR in the legacy product.

Antos resolved the transition by selling the existing business to a professional services partner that was already implementing it for some customers. The partner absorbed the entire support team and continued serving existing customers without interruption. The framework he applied: optimize first for customers, then for the team — in that order.

"You sometimes saw people who are really great at professional services and they just blaze through it with elegance and lightness. And we were just struggling with it so much." — Andrew Antos

"The combination of there's this thing that is just growing so quickly, and we might not be the right people for this particular type of business — that made it very easy to say we need to focus on this new product." — Andrew Antos

Three Product Iterations: What Changed Each Time