Jump to a section
Let's start with two pieces of advice that flatly refuse to get along.
In a 2022 video, three Y Combinator partners said the most honest feedback you'll get on your product arrives in the few seconds after you say the price out loud. Ouch, but fair.
Meanwhile, Figma's CEO told Lenny Rachitsky that Figma launched with no paid plans because the team expected the product to spread faster that way. It opened a closed beta in December 2015 and didn't charge until 2017.
Both companies did fine.
That's the problem.
If you're a solo founder staring at a half-built pricing page, "it worked for them" is available for either choice. Very helpful, history 😅
This is the fifth article in the series, and the question has been quietly hiding in all four earlier ones. Sneaky. We saw that likes and followers aren't buyers, that a pricing-page click sits below a payment on the evidence ladder, that a signup isn't the same as a user who cares, and that a Product Hunt launch can produce 400 signups and one paying customer. Pricing is where those lessons stop being theory and turn into an actual number on an actual page.
The moment you pick a number, or pick zero, you decide what kind of evidence you'll collect from then on.
A quick word on sources (I promise it's short) 🔍
I sorted what I used into four buckets: filings and company posts (the most checkable), founder interviews and retrospectives (self-reported, nobody audited them), investor and advisor opinions (YC partners, Pear VC, SaaStr), and my own analysis, which I flag when it shows up. Anything invented for illustration is labeled as invented.
Two honest gaps, because pretending otherwise would be silly.
You'll see "freemium converts 2 to 5 percent" repeated all over the internet, but I couldn't open an original dataset behind it, so I stuck to company filings and interviews.
And I looked for a controlled study comparing feedback from paying users and free users at startups. I didn't find one. Everything below on that question is experience, not experiment.
Why most of us default to free (it's not a silly instinct)
Free deserves a fair hearing first. It's not a silly instinct.
The YC partners listed the reasons they hear most often:
- Founders optimize for feedback over revenue, because getting users feels easier when nobody pays.
- They believe pricing is a one-shot decision with no redos.
- And when they sell to companies, they fear a price will make a hard sale harder.
The partners' answer to the redo worry is practical: you can change prices whenever you like, existing users can be grandfathered in, and today's users will be a small slice of next year's. One partner recalled Drew Houston saying Dropbox's early price was a reasonable-sounding number picked without deep analysis, that it may have cost them a lot of revenue, and that it was still fine.
There's also a pull that has nothing to do with strategy. Jason Fried of 37signals puts it bluntly: free sells. People want to try something with no strings attached, even when the price is two dollars. And there's research behind the feeling that zero is a special number. In experiments by Shampanier, Mazar and Ariely, people chose the free option far more often than the price gap alone would predict, even when a better product was available for a small charge. Those were choices between small items like chocolates, so treat it as a clue about human behavior, not a pricing law for software.
My read: for people who build things, there's one more reason.
Free is the only model with zero setup. No billing code. No refund policy. No tax questions. No awkward conversation.
On a Friday night, that's a very persuasive argument 😅
It's just not a strategic one.
What a price tells you that a signup can't
The strongest case for charging early is about information, not money.
The YC partners' version: you can learn from free users, but you learn a lot more from people who pay. They also warn that "customers" who pay nothing can mislead you for years. Their example is a founder who says Apple is a customer, and when asked what Apple pays, the answer is nothing. They just get to talk to Apple. A free pilot with a big company, they argue, proves little, because nobody on the other side had to spend budget or political capital.
Superhuman took this to an extreme. Rahul Vohra said in a 2020 interview that no one got the product for free, investors included, so everyone could feel the cost and have an honest conversation about whether the exchange felt good.
True Link, a card company for older adults, went against the grain. Its competitors gave cards away and collected waitlists of 50,000 or more. True Link charged $10 a month, launched with a waitlist of about 50 (44 if you take out family and friends), and had around 1,000 customers after its first year. Founder Kai Stinchcombe argues that free-trial signups hadn't really bought anything, and that charging forced the team to find customers who would sustain the business. Ads and press produced weak interest but not paid signups. Word of mouth did.
Read both stories with the usual caution. They're self-reported, both companies survived, and True Link's claim that it broke even before its peers is the founder's own. (Survivors tend to tell good stories.)
But the mechanism is worth keeping, and it's the part I find most convincing: a price doesn't only measure your audience, it selects it. Everything you build afterward, from onboarding to support to marketing, gets shaped by whoever showed up. Stinchcombe says it would have been a different company in every respect.
A price doesn't only measure your audience, it selects it.
The Product Hunt article showed the mirror image: a launch with 400+ signups and one payer, where the point at which people dropped off told the founder what to fix. Free signups gave attention. The price gave a diagnosis.
Paying isn't the same as fitting
Careful, though. Money is a filter, not a verdict.
Back to Superhuman. Vohra says that when the team ran the survey asking how disappointed people would be to lose the product, only 22% said "very disappointed." That was a company charging $30 a month from the start. Charging got him honest conversations. It didn't settle whether the product was something people needed. (We covered that survey in Article 3.)
Plausible adds another wrinkle. It's often remembered as paid from the start, but its own timeline says the public beta launched in January 2019 and paid plans arrived in May, when it had 60 active beta users and some chose to stay and pay. The month ended at $64 in monthly recurring revenue. Yes, sixty-four dollars. After that it took 324 days to reach $400. Growth came later, once blog posts took off on Hacker News and the positioning against Google Analytics sharpened.
So early payments can mean "this offer works for a few people at this price." Whether it's repeatable, or whether a stranger would pay without your help, is a separate test.
When free is the smarter move
The other side has better examples than the "free is for people who are afraid" framing allows.
Figma
Figma didn't charge until 2017. In Lenny's study, Dylan Field says the team thought it would spread faster without a price, and that eventually not paying became the barrier for companies adopting it. Customers started asking why Figma wasn't charging, because they wanted to be sure it wouldn't go away. Free did its job, then began to hurt once buyers needed something to buy.
Segment
Segment launched with a free plan, added pricing pages about a month later, and didn't build real self-service billing until roughly nine months after launch. Its CTO says the goal was to nail onboarding and value first, and to learn the price by surveying users.
Basecamp
Basecamp is my favorite cautionary tale, because the founders did everything carefully. They launched with a free one-project plan. In 2015 they removed it and ran a six-month experiment that said the change paid for itself. Years later, Jason Fried and David Heinemeier Hansson said fewer people were seeing the product, which hurts a business that grows by word of mouth. In a March 2025 episode they explained why a free plan came back. They report that weekly signups more than doubled and that paid conversion was running a few percent ahead, though possibly not by a statistically significant margin.
They also say the goal isn't to lift the paid rate this month. It's to get enough people to see Basecamp over the next several years. A six-month test can't see a five-year effect.
Slack
Slack is the textbook case. Its S-1 says it attracts new organizations mainly through self-service, and that free organizations often upgrade as they use it more deeply, likely because of the 10,000-message limit and single sign-on. About 10% of fiscal 2018 revenue and about 8% of fiscal 2019 revenue came from organizations that had been on the free plan and converted that year. Freemium can be central to growth while most revenue comes from somewhere else.
So what do these stories have in common? In my read, free had a job. The YC partners say much the same: free is fine when there's a structured plan to get paid, meaning a constraint people want to get around and the discipline to track whether free users convert and how fast. Their listed exceptions are open-core products like GitLab, freemium with a real limit like Slack, and ad-supported consumer products that need enormous scale. Jason Lemkin of SaaStr adds that freemium doesn't work as a business model without a mass-scale product, and a revenue-operations lead in his roundup says it needs a huge market, no-touch onboarding and an obvious first "wow."
My shorthand for what a free plan can be for: it spreads the product (users bring others), it exposes it (people who'd never have heard of you try it), or it feeds a limit that paying users want lifted. If yours does none of those, it isn't a free plan. It's a gift.
What the examples show (and what they can't)
| Company | Early approach | What the sources report | Source type |
|---|---|---|---|
| Superhuman | $30/month from the first user, no free tier, investors paid too | Early survey: 22% "very disappointed" | Founder interview, 2020 |
| Plausible | Free public beta (Jan 2019), paid from May 2019, 30-day trial, open source | $64 MRR at paid launch; 324 days to $400; $1M ARR by June 2022; 33.5% trial-to-paid for April and May 2020 cohorts | Company posts |
| True Link | $10/month per card while rivals gave cards away | About 1,000 customers in year one vs rivals' 50,000+ waitlists; cards division profitable by 2017 | Founder retrospective, 2017 |
| Atlassian (JIRA) | Paid from the start; free only for non-profits, academics, open source | An early user's recollection, not Atlassian's own account | Secondhand, via Lenny |
| Figma | No paid plan until 2017 | Customers began asking to pay; free became a barrier | CEO quote, via Lenny |
| Slack | Free plan plus self-serve paid plans from launch | About 8% of fiscal 2019 revenue from free orgs that converted that year | SEC filing |
| Evernote | Full-featured free version; most users meant to stay free | About 2% premium in 2010; aim to keep conversion within narrow bounds | CEO interview, 2010 |
| Basecamp | Free one-project plan at launch, removed 2015, back in 2025 | Signups more than doubled after the return; paid conversion maybe ~3% ahead, significance unclear | Founder podcast, 2025 |
Two caveats before drawing anything from this. First, the sample. Lenny Rachitsky studied 25 of the fastest-growing B2B companies. Seven had no paid plan at launch, nearly all offered a free plan from the outset, and every one eventually built a sales team. Those are winners picked after the fact. Second, I couldn't find a public retrospective from a startup that clearly blames its death on charging too early or staying free too long. That doesn't mean it never happens. Failed founders rarely write up their pricing decisions.
Notice what the table doesn't show: nothing in it proves the pricing choice made the company. Each founder was trying to answer a different question.
What each option costs 🧾
Every option comes with a bill. Here's how it shows up in the evidence.
Charging early costs you speed and exposure. True Link had 44 non-family waitlisters against rivals' tens of thousands. Plausible took 324 days to reach $400. Basecamp's lost exposure only showed up over years.
It can also cost you the right price. Shopify charged only a percentage of each sale for its first year, and its CEO told Lenny it totally didn't work; switching to monthly plans in 2007 or 2008 saved the company. YC would say prices can always change, and the YC partners are right about the number. My read is that the shape of the price (per seat, flat, percentage of sales) is harder to change than the figure.
Free costs money you may not be tracking. Jason Fried says the main cost is support: a paying customer who emails three times still leaves you ahead, a free one who emails twice means you've lost, though he says he'd rather not look at it that way.
He also says that when HEY ran in the cloud, each customer cost them about $30 a year to operate, which is partly why HEY isn't free. He adds that Basecamp can afford generosity because its margins are fat, and that you shouldn't copy it at 4% margins. If you build on paid AI APIs, that logic bites harder: when every free user triggers a model call you pay for, free isn't free. (My analysis.)
Here's invented math, not a benchmark. Say each free user costs $0.50 a month in support time and compute. Two thousand free users cost $1,000 a month. Now say 3% convert, near the real ratios below, at $15 a month. That's 60 payers and $900. Your free tier costs more than it earns until conversion, price or cost moves. Which of those you can actually move is the whole question.
Free costs you signal. Phil Libin said in 2010 that about two percent of Evernote's users were premium. In the same interview he said the goal isn't to push conversion up but to keep it within fairly narrow bounds, because a very high rate would mean the free version isn't good enough.
Dropbox's 10-K shows 18.08 million paying users at the end of 2025, and its press releases cite more than 700 million registered users. That's roughly 2.6%, but it isn't a conversion rate: "paying users" counts licenses (a 50-person team counts as 50) and "registered" counts every account ever opened. Treat it as a sense of scale. Both companies built big businesses on a small share of users. Neither can tell you your share will be big enough.
The YC partners also describe the quieter failure: founders who give the product away, say they'll see how it goes, and keep adding features without ever charging.
It changes with the business: B2B, consumer, open source
B2B
The YC case is that companies spend money on software constantly, so a price isn't the obstacle founders imagine, and not charging can read as "will these people be around to support us?" Pear VC's Pepe Agell goes further: pilots should never be free. His formula is a pilot priced at a 20 to 50 percent discount on the expected contract, about a month long and never past three, with the pilot fee credited toward the annual deal if the customer signs.
Luke Sophinos of CourseKey says in the SaaStr roundup that mid-market and enterprise pilots need some dollar commitment or they mostly don't work, and Lemkin agrees that unpaid pilots almost never do, while admitting he once closed Google after a year-long unpaid one because the customer genuinely put resources behind it.
Even the pro-paid piece has a counterexample. The CEO of ResiQuant says in it that free proofs of concept were the right call early, to validate fit with design partners, and paid ones came once the company needed a repeatable motion.
For a solo founder, nobody is signing an enterprise pilot for your $19 tool. My translation: a paid beta, a founding-member price, or a pre-order with clear refund terms and honest notes on what's built. Also, B2B and free aren't opposites.
In Lenny's data the free tier was often for individual users, while founders sold to the buying organization directly, and founder-led sales was how most of those companies got going. The user can be free while the organization pays.
Consumer and network products
YC's exception is ad-supported consumer businesses that need a giant base. They also warn about hiding behind it. Airbnb could have called itself a social network for travel and skipped payments, the way Couchsurfing did, and instead worked out how money would move. If you go free for scale, be honest that you've chosen to build a different kind of company.
Open source
The code can be free while the business isn't. YC points to GitLab's open-core model. Plausible publishes all its code and lets people self-host, and sells a hosted version with a 30-day trial. Its own numbers: about six months of donations produced six gifts of $5, while the paid cloud product grew from $400 to more than $8,500 in monthly revenue. Its conclusion is that a managed service on top of open code seems a better model than donations for some projects.
It also shows price as positioning. Plausible kept prices affordable because Google Analytics is free and people weren't used to paying for analytics. Vohra priced Superhuman toward the top of its bracket to signal quality, and says positioning should come before pricing. Both charged. They just said different things with the number.
The menu is bigger than two options
| Option | What it can test | What it can't | Seen in this article |
|---|---|---|---|
| Free forever, no plan | Reach and curiosity | Willingness to pay | YC's ad-supported exception |
| Freemium (limit plus paid tier) | Spread, and which limit people hit | Whether free users would pay for the same thing | Slack, Evernote, Basecamp |
| Free trial | Whether value shows up before the clock runs out | Long-term use | Plausible's 30-day trial |
| Free beta, then paid | Rough edges and early demand | Price reaction, until you charge | Plausible, Figma |
| Paid from the first user | Price reaction, honest conversations | Breadth of exposure | Superhuman, True Link |
| Paid pilot / founding-member price | Commitment from specific buyers | Whether strangers will self-serve | Pear VC's formula |
(That table is my summary, not a taxonomy from any one source.) Plausible's 30-day trial with no card needed is worth a note. It reports a 33.5% trial-to-paid rate for its April and May 2020 cohorts, a period when a viral post was sending it a lot of highly interested visitors. One product, one niche, one moment. Don't copy the number. Do notice that a trial asks for a decision on a clock.
How to decide: a scorecard 📋
This is a framework I put together for this article. It isn't an established model, and there are no universal thresholds.
| Question | Leans toward charging early | Leans toward free, with a plan |
|---|---|---|
| What's your biggest unknown? | Will anyone pay? | Will it work, or will it spread? |
| Who pays? | A company or professional you can name and reach | Individual users now, an organization later |
| Does use create exposure? | No: invites, shares and public pages don't exist | Yes: using it shows it to others |
| What does a free user cost? | Support time or API calls you can't ignore | Close to nothing |
| How fast does a stranger get value? | Needs setup or hand-holding | Minutes, no help (per the SaaStr point above) |
| How big is the audience? | A small niche you can list | Large enough that a few percent is real money |
| Is the product dependable today? | Yes, it earns a price | Not yet: free beta, with a date |
| Do you have a plan for free? | Not applicable | A limit, a metric and a deadline written down |
Signs it's time to charge:
- People ask to pay (Figma's customers did).
- Free users keep hitting your limit.
- The same people return.
- You've onboarded a few without hand-holding.
- The free tier costs more than you expected.
- The buyer needs an invoice and some evidence you'll still exist next year.
Signs to stay free for now:
- Nobody has reached the core value yet.
- The product breaks often.
- You can't say what a paid plan would include.
- Exposure is your bottleneck and the product spreads through use.
- You haven't built billing yet. Segment waited about nine months for that.
Whichever you pick, put a date on it.
Two invented examples to show the scorecard at work. A solo developer builds a $12/month browser extension that costs $0.30 per active user per month in API calls, used by individuals who never invite anyone. Most rows lean toward charging, so a short trial and a price from the start beat a free tier. Now a design tool where every shared link shows the product to a teammate and a free user costs almost nothing. That one leans toward a free plan with a limit, a conversion metric and a review date.Two invented examples of the scorecard at work
Two weeks of testing before you commit 🧪
- Put a real price on the page. A real checkout or payment link, with honest wording about what exists and what doesn't. (A pricing-page click without a payment is weak evidence, as Article 2 noted.)
- Show it to strangers who already have the problem, not friends. People who described the problem publicly or in conversation count. Compliments don't.
- Ask five of them for a specific commitment: a paid pilot, a founding-member price, or a pre-order with refund terms.
- Run the four-question price survey on everyone else. It's called the Van Westendorp survey. Both Segment and Superhuman used it. The questions:
- At what price would this be so expensive you wouldn't buy it?
- So cheap you'd doubt its quality?
- Expensive, but you'd still consider it?
- A bargain?
Vohra's caution is not to stare only at the bargain number. Positioning comes first.
- Before you start, write down what result would change your mind. If nothing would, you're not testing, you're decorating.
Mistakes to avoid when pricing shows up 🙈
Staying free out of fear. The YC partners' worry is founders who avoid charging because they dread the reaction, or because they think there are no redos. There are redos.
Free "for now" with no date or metric. "We'll see how it goes" isn't a plan. A limit, a conversion number and a review date are.
Counting a logo as a customer. A famous name using your free pilot is a conversation, not revenue.
Judging a pricing change on too short a window. Basecamp's six-month test looked right. The cost showed up over years. Decide in advance how long you'll watch, and which numbers matter beyond this quarter.
Treating a payment as proof of fit. Someone paying you tells you the offer landed. Whether they'd miss the product is a separate question, and Article 3's survey is one way to ask it.
Surprising existing users. If you change pricing, grandfather the people who trusted you early. YC points out that early users on old plans barely matter to your future bottom line, and being generous to them costs very little.
So, should you charge from day one?
Mostly, ask for something sooner than feels comfortable, but not for the reason the slogans give.
If you can name a buyer and reach them, charge before you build more. It can be small. A paid pilot, a founding-member price or a trial with a card on file all count, because they give you the one thing a signup can't: a real reaction to a real number.
If the product spreads through use, costs almost nothing per free user, and you have a limit, a metric and a date for the free tier, free can be the right call. Slack, Figma and Basecamp are all evidence for that. They're also evidence that free has to be doing a job, and that the job changes as the company does.
If you can't say which of those two situations you're in, that's the thing to find out first. Neither choice proves product-market fit. A payment proves someone valued this offer at this price. Usage proves someone came back. Fit shows up in whether they'd miss it.
And if you pick wrong? That's allowed. You can change a price. Just be honest with people when you do, and be kind to the ones who showed up early.
Earlier in this series
Follow The Startup Café for the next article
The Startup Café: Real Startup Lessons. Beyond the Hype.
Top comments (0)