Tip
Bangaly Kaba talks about Instagram’s connections pivot in 2017: “Early Instagram was built where every follow was created equal. If I followed you or Kim Kardashian, all follows were treated equally important. The average person would come on, follow a bunch of celebrities, then when they went to make their first post a few months later, none of their friends were following them. They were posting into an echo chamber. People would stop using the product because they felt bad. We had to convince Kevin and Mikey that it wasn’t right to prioritize celebrities to everybody because the regular person wasn’t having a great experience. The connections pivot changed everything—literally angle-changed the retention on Instagram. Our retention doubled over the course of a year and a half.”
Turns out AI product growth works the same way.
Your AI writing tool is growing fast—10,000 signups/week. Your growth team optimized onboarding: new users immediately see showcase features (AI that writes like Shakespeare, generates viral hooks, creates 10 variations instantly). Activation rate is incredible—80% of users try a feature on day 1. Your VP of Growth wants to double down: more showcase features, more viral examples, more celebrity AI personas.
Younger growth leaders optimize the metric in front of them. Activation is up, so amplify what’s working. They haven’t run enough growth loops to see the retention curve months later. They’re celebrating activation wins without watching what happens at month 3.
You’ve seen this pattern before. In 2015, your SaaS product had incredible demo-to-trial conversion by showcasing enterprise features. Three months later, SMB users churned—they never needed those features and felt overwhelmed. In 2018, you optimized onboarding around power-user workflows. Great activation for technical users, terrible retention for business users who just wanted simple reports.
You know the trap: optimizing early activation for impressive features often hurts long-term retention for regular users. So before you double down on celebrity AI personas, you check retention cohorts by signup source. The data shows exactly what you suspected: users who followed celebrity AI personas in onboarding have 40% lower 90-day retention than users who were guided to invite real colleagues first.
The problem is clear: showcasing impressive AI capabilities gets people excited to sign up, but regular users posting AI-generated content to an empty network feel like they’re shouting into the void. They need their actual colleagues using the tool, seeing their work, giving feedback. You pivot the onboarding: remove celebrity showcases, focus entirely on “invite 3 colleagues who need to see your drafts.” Activation drops 20%. Retention doubles.
This judgment—knowing that what drives activation often undermines retention—comes from watching enough growth loops play out over 12-18 months. Junior growth leaders optimize one metric. You’ve learned to ask “what happens to this cohort 6 months from now?” That pattern recognition only comes from watching enough cohorts mature.
Context
Bangaly Kaba led Instagram growth during hypergrowth (440M to 1B+ users). The connections pivot in 2017 was counterintuitive—stop showing celebrities to new users even though it drove follows.
The insight: celebrity follows without friend connections led to posting into empty networks, causing churn months later. Instagram’s retention literally doubled after prioritizing friend connections over celebrity follows.
For experienced executives evaluating AI product growth strategy, this pattern recognition is critical—you’ve watched enough growth loops mature to know early activation metrics can mislead. That wisdom comes from seeing 12-18 month cohort curves repeatedly.