Skip to content
Launch GitLab Knowledge Graph

[PRODUCT] Increase new user D2 retention from 40% to 55%

Business Goal

New user retention is our #1 (closed) growth lever.

Current funnel:

  • Signups: 10,000/week
  • D2 retention: 40% (4,000 return)
  • D7 retention: 22% (2,200 retained)
  • D30 retention: 12% (1,200 paying customers)

Target funnel (with better recs):

  • Signups: 10,000/week
  • D2 retention: 55% (5,500 return) ← +1,500 users
  • D7 retention: 30% (3,000 retained)
  • D30 retention: 18% (1,800 paying customers) ← +600 customers

Revenue impact:

  • 600 new customers/month × $15 ARPU = +$9K MRR
  • Annual impact: +$108K ARR

Root Cause Analysis

Why new users churn:

  1. Cold-start problem (62% of churned users): "Nothing relevant in my feed"
  2. Generic recommendations (28%): "Everyone sees the same trending stuff"
  3. Poor onboarding (10%): "Didn't understand how to use it"

Solution: Hybrid Recommendations

Current (CF only):

  • New user → No history → Random recommendations → Bad experience → Churn

Proposed (Hybrid):

  • New user → Trending + onboarding preferences → Good first experience → Retention
  • 1-10 interactions → 70% trending, 30% CF → Smooth transition
  • 10+ interactions → 30% trending, 70% CF → Full personalization

Engineering Work

Depends on:

Timeline:

  • A/B testing (#12): 3 weeks
  • Hybrid model (#13): 2 weeks
  • A/B test hybrid: 2 weeks
  • Full rollout: 1 week
  • Total: 8 weeks

Metrics to Track

Engagement Metrics

  • Session time (first session): 3min → 6min
  • Items viewed (first session): 8 → 15
  • Items clicked (first session): 1.2 → 3.0

Retention Metrics

  • D1 retention: 65% → 75%
  • D2 retention: 40% → 55% ← PRIMARY
  • D7 retention: 22% → 30%

Quality Metrics

  • NPS (new users): +12 → +35
  • "Relevant content" survey: 42% → 70%

A/B Test Design

Control: Current CF (cold-start = random) Treatment: Hybrid model (cold-start = trending)

Sample size: 10K new users per variant Duration: 14 days Success criteria: D2 retention +10% (absolute), p<0.05

Acceptance Criteria

  • A/B test shows +10% D2 retention
  • Statistical significance achieved (p<0.05)
  • No negative impact on D7/D30 retention
  • Engagement metrics improve
  • Full rollout to 100% traffic

Owner: @ben (Product) Eng Lead: @bill_staples Eng: @dmitry @stanhu

cc @jean_gabriel (A/B testing)