How Superhuman built product velocity, PMF, and a $30/month email product

Executive overview

Most companies slow down as they scale — Superhuman slowed down too, then deliberately reversed it. The culprit was conventional org design: too many direct reports pulling the CEO away from product, design, and marketing. The fix was structural: hire a president, cut direct reports from eight to two, and reclaim 60–70% of time for high-leverage work.

Superhuman's broader approach is built on contrarian moves: manual concierge onboarding at scale, ignoring most early user feedback, positioning around a single attribute (speed), and applying game design principles to a productivity tool.

The fastest path to growth is making something so remarkable that users can't help but tell others — then structuring the org so the CEO spends most of their time making it more remarkable.

Why Superhuman's velocity stalled — and how they fixed it

• Two distinct slowdown types: unavoidable (market widening across Gmail, Outlook, iOS, Android, Mac, Windows) and avoidable (org design).
• Market widening doesn't show up as visible features for existing users — it looks like nothing is shipping.
• Supporting every platform combination took years; it's now a technology moat almost no other email app can match.
• Avoidable slowdown came from eight direct reports consuming CEO time on hiring, OKRs, and accountability.
• CEO was spending only 6–7% of time on product, design, technology, and marketing.
• Hired a president to own executive team management and corporate strategy; CEO went to two direct reports.
• CEO time on high-leverage work jumped to 60–70%.

The switch log: tracking time actually spent

• Calendars show what you planned to do; only a work trail shows what you actually did.
• Method: Slack DM an EA (or Slackbot) with "TS: [task]" every time you switch tasks.
• Do whatever feels right as long as it feels right — attend to what bubbles up, rather than fighting it.
• At week's end, graph where time actually went.
• Surfaced the insight that recruiting and product were severely under-resourced in practice.

Manual onboarding: when to do it and when to stop

• Superhuman required one-to-one concierge onboarding for every new user for years; peak headcount doing this was ~20 people.
• Benefits: excellent engagement, retention, NPS, and virality metrics; kickstarts brand and word of mouth.
• In a capital-rich environment, human activation frees engineering to focus entirely on PMF and solution deepening — not self-service flows.
• Some competitors spent nearly half their engineering on self-service for products that never found PMF.
• Stop when top-of-funnel becomes wide enough that certain personality types won't tolerate a mandatory onboarding — and when you've built a world-class self-service alternative.

The product market fit engine

• PMF can be measured, optimised, and increased algorithmically.
• Core metric: "How would you feel if you could no longer use this product?" — track % answering "very disappointed."
• Below 40% very disappointed: growth is hard. Above 40%: companies tend to grow fastest.
• Don't act much on feedback from "very disappointed" users — they already love it.
• Don't act at all on "not disappointed" users — too far from loving it.
• Focus on "somewhat disappointed" users whose main benefit resonates — something small is holding them back.
• Split somewhat-disappointed users: those for whom the primary value prop resonates vs. those for whom it doesn't. Disregard the latter.
• Roadmap principle: half the time doubling down on what superfans love; half the time removing blockers for the right somewhat-disappointed segment.
• Still applied to individual subproducts (e.g., Superhuman for Sales) and new product bets.

Virality and word of mouth

• No product sustains a viral factor above 1.0 for any real period — even Facebook peaked around 0.7, for roughly a year.
• Even LinkedIn's famous address-book import had a lifetime viral factor of ~0.4 (considered good).
• True virality is unmeasured: one user spontaneously telling another.
• "Whales" exist — single users who invite 50–100 people — but the underlying driver is a product worth sharing.
• Collaborative/multiplayer design creates structural word-of-mouth; Superhuman 2.0 made email multiplayer.

Positioning and pricing

• Positioning comes before pricing — always.
• Spent the first year interviewing hundreds of potential customers; "email is too slow" emerged consistently.
• Speed was unique, available, and reinforced the product strategy (incumbents structurally can't match it).
• Cocktail party test: users pitched Superhuman as "dude, you have to use it — it's really fucking fast."
• Used Van Westendorp price sensitivity meter with ~100 early users; four price questions mapped to four psychological thresholds.
• Oriented around question three (starts to feel expensive but you buy it anyway) rather than question four (bargain price).
• Median answer for question three: $30/month — that became the price.
• Gut-checked against venture scale: 300,000 subscribers at $30 = $100M ARR at 10x = $1B valuation. Answered yes.

Game design vs. gamification

• Gamification (points, badges, rewards) actively undermines intrinsic motivation — Stanford study showed expected rewards halved children's drawing time.
• Game design works because it builds intrinsic motivation; gamification fails when the underlying experience wasn't designed like a game.
• Superhuman's five game design areas: goals, emotions, toys, controls, and flow.
• Principle: make fun toys first, then combine into games. The best games are built from toys.
• Example toy: the time autocompletor (triggered with H) — gibberish inputs work, time zone math is automatic, "snooze until never" delights users.
• A toy passes if it indulges playful exploration, is fun without a goal, and elicits pleasant surprise.

Typography and attention to detail

• Evaluated ~15 font families; printed and let them "marinate" on a desk to capture visceral and considered responses.
• Four criteria: gorgeous in itself; can convey any sentiment without overpowering it; maximises reading speed and comprehension; makes email addresses look natural.
• Chose Adèle Sans (TypeTogether): narrow characters allow optimal line length (~90–120 chars) even on small windows; unique treatment of the @ symbol aligns it to the text baseline.
• Worked with a type designer to refine specific glyphs. Had ~10–15 users at the time.
• Fixed typographical measure (line length) to optimal reading speed — eliminating the wide-line problem of default Gmail.

AI features and what surprised Superhuman

• Write with AI: matches the user's own voice and tone from prior sent emails — usage reached 37 times per user per week, far beyond expectations.
• Auto Summarize: one-line pre-computed summary above every conversation, updates as new emails arrive.
• Instant Reply: inbox wakes up with a draft reply on every email.
• Ask AI: natural language queries across the full email archive — tasks that took 20–30 minutes now take under 5 seconds.
• Newer: auto labels (prompt-defined), auto reminders (never drop the ball), auto drafts, and workflows (multi-step automations that run without the user in the inbox).
• Biggest surprise: unpredictability of user love. Features expected to be commodities (Write with AI) became the most-used; features expected to be hits underperformed.
• AI-written emails sent via Superhuman grew 4x across 2024.

Enterprise motion

• Outlook users have different expectations: expect email and calendar integrated, not separate.
• Invested heavily in calendar as a result.
• Required enterprise-specific features: external recipient indicators, sensitivity labels, Microsoft Intune support for mobile device management.
• Sales is multi-threaded: users, IT, and workplace management groups (analytics on team efficiency) are distinct stakeholders.
• First major strategy consulting firm (big three) took a year to pilot then accelerate — final unlock was compliant mobile app.

Single decisive reason (SDR)

• Decision-making tool from Reid Hoffman: for any important decision, isolate one reason that alone supports it.
• A collection of weak reasons rarely adds up to a strong reason — and reveals the decision hasn't been fully thought through.
• When someone brings a multi-reason decision: ask "if only one of those were true and you'd still do it, which is it?"
• SDR also forces opportunity cost thinking: building this for weak reasons means not building something else for a strong reason.
• Applied in group settings especially, where consensus-seeking tends toward weak-reason accumulation.