Data analytics — visualized
59.6%
Churn rate (30-day window)
60.6%
Low-engagement users (1–5 sessions)
49.6
Mean sessions (skewed by outliers)
1
Median sessions — the real story
Session distribution across all users
Right-skewed distribution — the vast majority of users had 1–10 sessions, while a small cohort of power users drove the inflated mean. SD = 127.27.
Low engagement (1–5 sessions) — 60.6%
Active users (6–100 sessions)
Power users (100+ sessions)
Median of 1 vs. mean of 49.6 — a gap that reveals the data cleaning challenge: outliers required careful handling to avoid misrepresenting typical user behavior.
Engagement tier breakdown
Users segmented into three behavioral groups. The low-engagement majority (60.6%) was the primary churn driver.
Low (1–5 sessions) 60.6%
Active (6–100) 30.4%
Power (100+) 9%
Churn vs. retained — 30-day window
Nearly 3 in 5 users were classified as churned (inactive for 30+ days). Cross-referenced against engagement tier to identify which segments were most at-risk.
Churned 59.6%
Retained 40.4%
Churn defined as no platform activity in the prior 30 days. Calculated from backend first-seen / last-seen timestamps after cleaning irregular export fields.
Tools used
Python
Pandas
Data cleaning & wrangling
Statistical analysis
Survey design
User interviews