How this growth loop worked
A user who already relied on Superhuman got a one-on-one onboarding session, then invited colleagues onto a waitlist. Each invitee waited, then sat through their own concierge session where a specialist tuned the setup and taught the keyboard workflow. That start turned invitees into paying, fluent users, who invited their own peers. One cycle's output, a converted power user, became the next cycle's input.
An illustrative K near 0.25 means each user brought in about a quarter of a new user, not a self-replacing chain. That loop never sustains itself. It cut the blended cost of every paid or founder-led signup: a share of new users arrived pre-qualified through a trusted colleague, warmed by scarcity and a personal invite. The referral engine amplified other channels instead of replacing them.
If your loop stays below 1, stop chasing overnight virality and treat referrals as a discount on acquisition. Spend the effort where it converts: a concierge welcome that makes new users fluent enough to brag. The slow, high-touch path Superhuman chose was the point, not a compromise. Fewer, more fluent users refer better than a flood of half-activated ones, and the cycle time you tolerate sets how fast it pays back.