Rahul Vohra 谈 Superhuman：口碑才是增长，产品市场契合度可以量化

Vohra says the most pivotal moment of his career before Superhuman was selling his previous company, Rapportive, to LinkedIn. Rapportive was a Gmail extension that showed who people were, where they worked and their recent tweets, and it had attracted a large share of LinkedIn's daily active users. The real value of the acquisition, he says, was working for LinkedIn's head of growth, Elliot Shmukler, who had scaled the network from 25 million to over 250 million members.

In their first one-on-one, Shmukler told Vohra that no app has ever sustained a viral factor above one. Even Facebook in its heyday managed about 0.7, and LinkedIn's address-book import feature, which was notorious, had a lifetime factor around 0.4, considered good. Viral mechanics always level off, so the true secret is word of mouth: the virality you cannot measure, when one user spontaneously tells another.

Lenny Rachitsky adds that real spread is often one person with a large audience, not exponential person-to-person sharing. Vohra agrees that Superhuman has whales who invite dozens or hundreds of people, helped by a free-month referral incentive. That incentive does not work inside companies, which is why he argues for genuinely multiplayer products and describes Superhuman 2.0 as a move to make email collaborative.

To make people want to share, Vohra bakes growth ingredients into the company values: create delight, deliver remarkable quality and build the extraordinary. Remarkable products, he argues, are another route to word of mouth.

Rachitsky notes that Product Hunt's CEO had tweeted that Superhuman's product velocity had kicked into another gear, and asks what changed. Vohra says users had earlier told them the company had slowed. He separates unavoidable slowdowns from avoidable ones, and classifies all work as either solution deepening (improving the product for existing users) or market widening (making it available to more users).

Early on, Superhuman supported only Gmail on the web, so every engineering dollar went into solution deepening and users loved it. Growth then required an iOS app, macOS, Windows, Android and Office 365 support on multiple platforms. That work made the product available everywhere and built a technology moat that lets Superhuman sell into the enterprise, but it made the perceived pace of improvement drop.

The avoidable part was management structure. Following conventional advice, Vohra had hired VPs and ended up with eight or nine direct reports, which consumed time on hiring, goals and accountability conversations. Tracking his time, he found that only 6 to 7% of his week went to product, design, technology and marketing, the things he can be world-class at.

He realized that at a certain scale a CEO can define the role. He hired a president, cut his direct reports from eight to two, and raised his time on product, design, technology and marketing to roughly 60 to 70% of his week. Rachitsky highlights that the fix was org design, and that perceived slowness can simply mean the company is building things the vocal user does not need.

Vohra explains his time tracking, which he calls the Switch Log. A calendar shows what you planned, but only the trail of work shows what you did. Each time he changes task he sends a Slack message to his assistant with "TS:" and a few words, which lets him do what feels right for as long as it feels right, still turning up for meetings and one-on-ones, and then chart the week.

The log showed he spent about 5% of his time on recruiting when it should be much higher, and 6 to 7% on product areas where he is strongest. He treats distractions that bubble up as signals worth attending to, or writes them in a large A3 sketchbook with a 4H pencil. He does not fear context switching but acknowledges the roughly 21 minutes it can take to recover from an interruption.

Vohra practices transcendental meditation: 20 minutes of meditation plus 10 of rest, in the morning and again around 3 pm, which nothing on his calendar may override. After four or five years he notes better focus, creativity and expressiveness. His advice is to learn from a coach one-on-one rather than apps or books, and he recommends Laurent Valasek of the Peak Leadership Institute in the Bay Area.

The president took over operations and the management of the executive team, including hiring, goals and accountability. Vohra also describes the role as a strategic thought partner on questions such as how far to go with multiplayer, how aggressively to lean into AI, gross margins, a second product and where to hire. He compares it to a grown-up co-founder, since his original co-founders left long ago.

Vohra says the investor Ed Sim wrote the first three checks into Superhuman, after having backed Rapportive. His first pitch was a single slide of a Gmail screenshot with most of it scribbled out. When Sim offered to wire 750K, Vohra, with no employees and no spending plans, took 250K at a time, three times over several months.

On attention to detail, Vohra describes the font. He printed his interface in about 15 families and let them sit on a desk to marinate, then rejected them all. He wanted a typeface that is gorgeous, can carry messages from party invitations to bad news, speeds reading, and makes email addresses look good.

He settled on Adelle Sans from Type Together, which is beautiful, upbeat yet serious, and unusually narrow, allowing the optimal line length of roughly 90 to 120 characters even in small windows. Its at symbol sits on the same baseline as the other letters, which makes addresses look natural. Superhuman worked with a type designer on a few glyphs, and the team had only 10 to 15 users at the time.

Rachitsky asks how to balance polish against launching early. Vohra says it depends on market structure. In a marketplace with strong network effects, such as Lyft or Uber in their heyday, every minute matters and you should work extremely hard. For a mission-critical tool like email, shipping something half-baked is dangerous.

He illustrates with a founder using Mailbox, the swipe-to-archive app later acquired by Dropbox. After a Y Combinator demo day, the founder's follow-up emails never appeared in Gmail's sent folder, and investors later received duplicates once the queued mails were sent. Vohra says such errors make you look foolish in front of customers or candidates, so email must simply work.

On choosing what to dial up to eleven, Vohra drew on the book Positioning: The Battle for Your Mind. In his first year he interviewed hundreds of potential customers, and many said Gmail and Outlook were too slow. Speed was unique and available because almost no software had been sold on it since Chrome, it reinforced a strategy of competing with incumbents who struggle with speed, and it passed his cocktail party test of how users pitch the product.

In the early years Superhuman required a one-to-one concierge onboarding before anyone could use it, with up to about 20 people doing it at peak. Today most customers use self-service, though the onboarding experience still exists. Vohra says stopping was right because some personality types refuse one-on-one sessions, and the top of the funnel eventually becomes too wide.

Building self-service was terrifying, he says, because the company's DNA was built around manual onboarding, but they eventually made it world-class. The benefits of concierge onboarding were that engagement, retention, product-market fit score and virality beat industry benchmarks, and that an early cohort of superfans kickstarts a brand.

It also let the team avoid building first-time-user flows and growth loops when money was easy to raise, so engineers could focus on product-market fit. He saw competitors spend almost half their engineering dollars on self-service flows for products that never found fit. Founders should do it only if it makes money, as Superhuman's did, and know when to stop.

Vohra lists the core ideas of his well-known First Round essay: you can measure product-market fit, optimize it, systematically increase it, and even have an algorithm write your roadmap. A second lesson is that to reach fit you must deliberately not act on many early users' feedback, even while listening intensely, because the product cannot be for everybody.

The measure is to ask how people would feel if they could no longer use the product: very disappointed, somewhat disappointed or not disappointed. Sean Ellis found that companies struggling to grow almost always scored under 40% very disappointed, and fast growers scored above it. Vohra says this predicts success much better than net promoter score.

To raise the number, ignore the not-disappointed group, who are a lost cause, and avoid over-acting on the already-loving. Focus on the somewhat disappointed users, and split them into those for whom the main benefit resonates and those for whom it does not. Disregard the latter politely, since even a fully built wish list would pull the product in another direction.

Each planning cycle, Vohra suggests spending half the time doubling down on what fans love and half overcoming objections from the relevant somewhat disappointed users. Superhuman now runs the engine on separate components such as sales or collaboration, and it can use email or in-product surveys, although changing the survey method resets your baseline.

Vohra argues that business software should be made the way games are, because people then find it fun and tell friends. Gamification, adding points, levels or badges, fails because rewards destroy intrinsic motivation. He cites a 1970s Stanford study where children who expected a certificate spent 8% of their time drawing afterward, versus 17% for those with no reward.

He was a professional game designer before becoming a founder. Superhuman uses five areas, goals, emotions, toys, controls and flow, and principles such as making fun toys and combining them into games. Games built from toys are fun at both levels, as football is built from a ball.

Superhuman's favorite toy is the time auto-completer for snoozing and reminders. It interprets input like 2D, 3H or 1MO, handles odd strings and time zones such as 8 am in Tokyo, and even lets you snooze an email until never. Vohra suggests asking whether features invite playful exploration, are fun without a goal and produce moments of pleasant surprise.

Vohra says to figure out positioning before pricing. Superhuman positions itself as the best email tool for high-performing teams and individuals, and it can compare before-and-after data: users get through email twice as fast, reply one to two days sooner and save four or more hours a week.

For pricing, inspired by Monetizing Innovation by Madhavan Ramanujam, the team used the Van Westendorp price sensitivity method with about a hundred early users, asking four questions about too expensive, too cheap, getting expensive and a bargain. Most startups set price near the bargain point, but Superhuman's best-in-category position fits the point where it starts to feel expensive.

The median answer there was $30 a month. As a gut check, a billion-dollar valuation at roughly ten times ARR would need about $100 million, or around 300,000 subscribers at $30, without counting price rises or new products. The team judged that achievable and proceeded.

Vohra lists Superhuman's AI features: Write with AI, which matches your voice and tone and is used on average 37 times per user per week; pre-computed auto summaries; instant replies; and Ask AI, which replaces old keyword search. A recent launch added auto labels, auto reminders, auto drafts and workflows that can automate things like replying to job candidates. The share of AI-written emails sent through Superhuman grew fourfold over 2024.

His biggest surprise was how unpredictable user love was. Write with AI sounded like a commodity feature but was hugely popular, while features he expected to be heavily used got less use than hoped. He jokes that what he predicts will work, people use less, and what he builds with little conviction, they love.

On the move into the enterprise, Vohra says buyers often come from Outlook, where users expect a strong calendar, external recipient warnings and sensitivity labels. IT needs controls such as Microsoft Intune support, and workplace groups want analytics. He mentions a big-three strategy consulting firm with thousands of users, which ran its own study and found Superhuman saved partners about 3.3 hours per person per week.

Vohra describes a thinking tool he learned from Reid Hoffman at LinkedIn: the single decisive reason, or SDR. For important decisions, you should be able to name one reason that alone supports the choice. People often rely on a pile of weak reasons, and a pros-and-cons list can bias you simply because of the effort of listing the cons, especially in risk-averse group settings.

When someone brings him a decision, he asks for the SDR. Not finding one does not mean the decision is wrong, only that the person has not yet worked out why they want it. If only one reason were true and the others were not, would they still do it? Sometimes yes and sometimes no, and the exercise can expose a weak case or a strong reason on the other side.

Vohra adds that multiple low-quality reasons rarely add up to a high-quality one, and that every decision has an opportunity cost, since building one feature means not building another. He notes that his own SDR for SDR is exactly that point.