产品市场契合有四个层级：First Round 给 B2B 创始人的实战框架

Todd Jackson is a partner at First Round Capital who previously led product on Gmail, worked on Facebook's newsfeed, and was VP of product and design at Dropbox. He says he became a seed-stage investor because he loves helping pre-product-market-fit founders get there, over and over again. Host Lenny Rachitsky frames the conversation as an exclusive peek into a new First Round program built on a year of studying the firm's own portfolio.

Jackson's argument for a framework is that most advice on the topic is vague ("you'll know it when you see it"), while other startup topics like hiring a first salesperson have specific, tactical guidance. He points to Rahul from Superhuman and his approach and Lenny's own benchmarking work as popular precisely because they brought some science to something treated as art. First Round's "Paths to Product-Market Fit" series of founder interviews turned up a consistent set of patterns, and those became the basis for the framework.

The framework is aimed at early, sales-led B2B founders in roughly their first six to nine months. Jackson says consumer and bottom-up products involve more alchemy and taste, while enterprise sales-led businesses have more repeatable science. He is candid that nothing guarantees success; the aim is to increase the odds and cut luck. The related free program, Product-Market Fit Method, takes about ten hours a week of work founders would be doing anyway, and First Round takes no equity.

First Round defines extreme product-market fit as widespread demand for a product that satisfies a critical need and can be delivered repeatably and efficiently to each customer. Jackson says efficiency is the part most people leave out. He cites WeWork and Casper as products with demand and satisfaction but broken economics, and repeats partner Brett Berson's vending-machine thought experiment: a machine that turns a dollar into $100 would have huge demand and happy customers, but isn't a viable business.

He also introduces the "marginal customer": the next customer should get easier to acquire and serve as fit strengthens. The four levels are nascent, developing, strong and extreme, and for the best enterprise companies the full journey takes roughly four to six years. At each level there are trade-offs, since spending on marketing raises demand but hurts efficiency, and heavy automation can improve efficiency while harming satisfaction.

At level one, a company is typically pre-seed or seed with fewer than ten people, and the job is to find three to five customers with an important and urgent problem and give them a satisfying solution. Satisfaction comes first, demand second, efficiency last. Jackson's example is Vanta, founded by Christina Cacioppo after earlier ideas (a smart speaker for meeting summaries, and something around dropshipping) didn't work.

Cacioppo talked to security engineers and CTOs and heard the same complaint again and again: security questionnaires and compliance audits were grungy manual work. Early design partners such as Segment, Front and Figma were trying to sell to large enterprises and were held back by lacking a SOC 2. She offered to do it for them, did the work manually, and they closed the deal, which Jackson sees as a product that satisfies a promise and unlocks revenue.

Typical level-one benchmarks are $0 to $500K in ARR, demand that comes mostly from your network, and perhaps 20 warm intros to land a single customer. Efficiency metrics like burn multiple or gross margin don't apply yet. Yellow flags include customers who wouldn't be disappointed if the product vanished, five customers who each want a different key feature (more consulting than product), and low or fading usage.

Jackson also relays Persona founder Rick Song's advice not to get "friend-zoned" by customers. Song sat early customers down one-on-one and asked whether the product was a necessity, how painful it would be if it disappeared, and whether they would switch to a competitor charging half as much.

Lattice began in 2015 as an OKR tool sold to heads of HR, but customers used it for a quarter and then dropped it. Jack Altman kept the persona, because he knew those buyers well, and pivoted to performance management, which was coming back into favor. He sold his first five to ten customers with Figma mock-ups before building much of anything, and Jackson says the response was "off the charts".

Jackson says the four Ps are the best framework he has for thinking about pivots. Lattice kept the persona but changed the rest, Vanta changed all four, and Plaid started as a consumer budgeting app, then licensed its bank-connection code to developers at fintech companies, which was a flip of the persona, problem and promise with the product largely retained.

Jackson himself prefers to sell before he builds, so he focuses on persona, problem and promise and lets the product exist to satisfy them. Lenny adds that those three are much cheaper to change than rebuilding a product. Both note that level one rarely feels like roaring success, and Lenny recalls Retool's David Hsu saying every early customer felt like the last one he would ever get.

Level two, developing product-market fit, is about going from five to 25 satisfied customers. Jackson says willpower can get you to five or ten customers, but beyond that the product and a scalable channel must do the heavy lifting. Benchmarks include up to about 20 people, $500K to $5M in ARR, roughly 10% conversion from first call to close without a warm intro, a magic number of 0.5 to 0.75, regretted churn of 10 to 20%, and NRR of at least 100%.

Looker, founded by Lloyd Tabb, spent a long time at level one because the product can't be sold with mock-ups: people need to see their own data modeled. Tabb spent 20 to 40 hours per prospect before they were customers, later called a forward-deployed process, and once it was repeatable the company moved quickly from five to 25 customers with a 75% close rate and no churn. Demand channels such as SDRs, AWS Redshift partner marketing and customer events came afterwards.

Ironclad's founder Jason Boehmig first called the product an AI legal assistant in 2015 and found it nearly impossible to sell. A one-line email asking "Are you a CLM?" led him to learn about contract lifecycle management, and he and his co-founder built a demo on the train to the meeting. They won against 10 to 12 established vendors. The lesson he drew was that expanding an existing category beat creating a new one, and Jackson agrees that category creation is hard.

Jackson lists the yellow flags for level two: you've been there 12 to 18 months, regretted churn is above 20%, sales cycles drag, deals are lost late in the funnel, or you can't hit your target price. He warns that polite customer phrases like "we don't have the budget" or "let's talk next year" really mean no. Near the top of level two, people should start to associate your company with a specific problem, such as Vanta with SOC 2.

He quotes Jack Altman as saying that most founders do a 10% pivot when they need a 200% pivot. Jackson's reading is that it's psychologically hard to walk away from a plateauing business with real customers, but that founders must be willing to let go and re-nail the four Ps. Level two is the second most common place to get stuck, after level one.

On sales cycles, Jackson says long ones are acceptable with large contracts of $100K or more, which can take three to six months, but a slow cycle combined with a low contract value is the "quadrant of death". When he is asked what usually needs to change, he says it's typically that the problem isn't painful enough or the promise isn't valuable enough, assuming the persona is reasonable.

Level three, strong product-market fit, is where the familiar adages apply, like fish jumping into the boat. Typical benchmarks are 30 to 100 people, about Series B, a cracked scalable channel with at least 10% of inbound coming from referrals or word of mouth, around 100 customers, and $5M to $25M in ARR. Efficiency now comes into focus: gross margin above 60%, burn multiple of one to three, regretted churn under 10% and NRR above 110%.

Even then, founders rarely feel certain. Lenny recalls Retool's David Hsu feeling that even customer number 100 might be the last, and Databricks' Ali Ghodsi saying he wasn't sure about fit even at $100M. Level-three problems include a leaky bucket, slowing growth, a saturating channel, more competition, and spending too much to grow.

At level four, a company is likely a unicorn with more than 100 people, more than $25M in ARR, gross margin above 80%, burn multiple under one and NRR above 120%. The challenge becomes expanding the total addressable market, typically through new markets or new products, as in Verkada's move from cameras to alarms and badge readers, Stripe's Radar and Atlas, or Square's Cash App. Jackson points out that every new product needs product-market fit again.

In First Round's experience, more than half and probably 60 to 70% of companies get stuck at level one or two, leaving about 30% reaching level three or four. Jackson says that moving even a few more companies past that line would matter greatly to founders and the wider ecosystem. Lenny adds Casey Winters' point that customer expectations keep rising, so fit is a never-ending battle.

For timelines, Jackson suggests taking 12 to 18 months at level one because the choice of persona and problem is the most important decision a founder makes. He quotes partner Josh Kopelman that founders spend 99% of their time building and 1% picking the market, problem and customer. Then level two might take about a year, level three one to two years, and the climb from $25M to $100M a couple more.

Going from five to 25 customers in a year is, for Jackson, almost always a sign of strong fit. If a company has gone four to five years without feeling pull from the market, he says the odds of a sudden breakthrough are low, and that there's "no shame" in returning money or seeking a soft landing, because everyone knows most startups don't make it.

On persona, Jackson says a market is best thought of as a collection of people with money to spend, and that great founders like Lattice's Altman, Plaid's Perret and Looker's Tabb knew their customers deeply, even texting and meeting them on weekends. He then previews the program's second session, dollar-driven discovery, noting that customers are polite and good at describing problems but poor at predicting their own behavior.

He demonstrates by interviewing Lenny about his top goals and what's hard about them, and then pitching a service that would find 500 great guests. Lenny says he'd pay a lot for it, which Jackson calls a "wow statement". He looks for those or for demonstrated behavior like asking to meet again or requesting the deck, and he says a reaction like "interesting" is a polite no. He also warns about the "happy ears" trap of hearing only what supports your idea.

To test ability to pay, he asks whether the customer is already looking for or building something, where the budget would come from, and how purchases get approved. The best customer has tried and failed to build a solution themselves. For willingness to pay, he recommends Madhavan Ramanujam's three questions: a fair price, an expensive price and a prohibitively expensive price, noting that the expensive one is often what people will really pay for a good product.

Fidelity matters too: Lattice could be sold with mock-ups, Looker needed a demo on real data, and Vanta needed the work actually done. Jackson's rule of thumb for having talked to enough people is when you can predict 70 to 80% of what the next one will say. Founders interested in the program can apply at pmf.firstround.com by May 7.