Twitter's growth team did not ask new users to 'engage more.' They found one number: follow at least 30 accounts, and a new user became a durable, long-term active user, according to Mattermark's account of growth lead Josh Elman's work. Twitter then rebuilt onboarding to march new signups toward 30 follows as fast as possible, front-loading suggested accounts instead of leaving discovery to chance.
That is the pattern worth stealing, not the number itself. Every company that has publicly cracked activation found a single, countable action correlated with retention, then re-engineered the first days of the user's life around hitting it. Facebook's growth team under Chamath Palihapitiya found that users who added 7 friends within 10 days were dramatically more likely to stick around, and made 'get any individual to 7 friends in 10 days' the team's singular focus on the way to a billion users, per TechCrunch's 2013 account. Slack founder Stewart Butterfield described a different threshold: teams that exchanged 2,000 messages had 'really tried' Slack, and 93% of teams that crossed that line stayed paying customers, according to First Round Review's 2015 interview.
The number is not obvious. Find it in your own data.
None of these thresholds were guessed. Twitter, Facebook, and Slack each found their number by looking backward: pull the users who stayed a year later, pull the users who churned in month one, and find the single action count that discriminates between them. That is retro cohort work, not a brainstorm. If your team is debating whether the magic number is 3 projects created or 5 teammates invited, you already have the data to settle it: split existing cohorts by that action count and check which threshold predicts month-6 retention.
Most companies skip this step and design onboarding around what feels intuitive instead. The cost shows up in the aggregate numbers. Across 62 B2B SaaS companies it analyzed, Userpilot found an average user activation rate of just 37.5% (median 37%), meaning most signups never reach the product's core value moment at all, per its 2024 User Activation Rate Benchmark Report. That gap is not a top-of-funnel problem. Those users already signed up. They are inside the product and still leaving without ever hitting the moment that would have made them stay.
Superhuman shows the harder, more valuable version
Finding the number is step one. Superhuman CEO Rahul Vohra went a step further and used it to decide who to build for. He adopted Sean Ellis's product-market-fit survey, asking users 'how would you feel if you could no longer use this product?', and treated 40% answering 'very disappointed' as the benchmark separating startups that can scale from those that will struggle, a threshold drawn from data across hundreds of venture-backed companies, per First Round Review's account. Superhuman was below that line. Instead of building broadly for everyone who signed up, Vohra's team segmented respondents and built specifically for the users who already said 'very disappointed,' on the theory that features aimed at the on-the-fence majority dilute the product for the segment that already loves it. Over several quarters, that focus raised Superhuman's very-disappointed score from 22% to 58%, per the same First Round account.
The combination of these examples is the actual framework: find the number that separates your retained users from your churned ones, then decide whether you are building onboarding to rush everyone toward it, or building the product itself for the segment already closest to it. Twitter, Facebook, and Slack solved the first problem. Superhuman solved the second. Most teams solve neither, which lines up with a 37.5% average activation rate across the category Userpilot measured.
What to do this week
- Pull your last 12 months of signups and split them by a single candidate action count (invites sent, projects created, messages exchanged). Check which threshold best predicts who is still active at month 3.
- If you cannot find a clean threshold, run Sean Ellis's 'very disappointed' survey against your active user base this week. You need enough responses to read a percentage, not a perfect sample.
- Once you have a number, audit your onboarding flow for everything that delays a new user from reaching it. Cut steps that do not serve that one number.
- Segment your PMF survey results and look at what the 'very disappointed' respondents have in common. Build your next roadmap quarter for them specifically, not for the median signup.
The founders who solved activation did not run more onboarding experiments. They found the one number that already separated their stayers from their leavers, then built everything backward from it.