0-to-1 Teardown Series · Issue 01
The problem
Superhuman had real usage and genuinely positive sentiment, but growth had stalled in the way that's hardest to diagnose: nothing was obviously broken. People who used it liked it. It just wasn't spreading. That's the failure mode "build more features" can't fix, because the product isn't failing anyone in particular — it's failing to be essential to anyone in particular.
Founder Rahul Vohra's team ran Sean Ellis's product-market-fit survey — the single question "how would you feel if you could no longer use this product?" — expecting a strong result. It came back well under the 40% "very disappointed" benchmark Ellis's research treats as a PMF signal.
What they measured
Instead of treating that number as one score to chase upward, they segmented it — by the same respondents' answers to two follow-up questions: what type of person do you think would benefit most from this product, and what's the main benefit you get from it. That turned one weak average into a clear split: a smaller group who were genuinely delighted, sitting inside a much larger group who were lukewarm.
Under 40% were "very disappointed" overall. Two follow-up questions were enough to find the real segment hiding inside that number.
The lukewarm majority wasn't wrong about the product — they just weren't the product's real audience. The delighted minority described the same tool in a completely different register: not "a nice email client," but something closer to a competitive advantage in how fast they could operate.
What changed
The team rebuilt the roadmap and the onboarding around that delighted segment instead of the average user — leaning into speed and keyboard-driven workflows for people who treat inbox velocity as a genuine edge, rather than smoothing the product into a more generic "nice email app for everyone."
Distribution followed the same logic: rather than opening self-serve signup to capture the widest possible top of funnel, access stayed invite-gated with white-glove, one-on-one onboarding calls for years — deliberately trading top-line signup volume for a much higher density of the segment they now knew they were building for.
What I'd flag
The instructive move here isn't "run a PMF survey" — plenty of teams do that and get a number they don't know what to do with. It's that they refused to average the result. A mediocre score hiding a genuinely excited minority is a different problem than an evenly mediocre product, and the fix — narrow the target user, then rebuild around them — only becomes obvious once you segment instead of summarize.
Drawn from Rahul Vohra's account of Superhuman's PMF process, published on First Round Review, and widely referenced since in product and growth writing. This is my own reading of the publicly documented case, not new reporting.