Churn Analysis V2 (Corrected for Timeline Bias)

Date: 2026-04-26

Goal: True comparison of Paid Drop-offs (Churned) vs. Retained Paid users.

Methodology Changes (per Jeff's guidelines):

Excluded all system events: Removed keep-alive, auth, listGroupMembers, and non-genuine actions.
Corrected Timeline Bias: Measured ONLY the first 60 days of activity for both churned and retained users. This prevents long-time retained users from looking artificially more engaged simply because they've had more months to rack up events.
Focuses strictly on value-driving actions: Chat, Documents, Agents, Uploads, and Exports.

Cohort Summary

Cohort	Users Count	Avg Active Days (First 60d)	Avg Events (First 60d)
Churned	27	12.6 days	139.3 events
Retained	40	18.8 days	241.7 events

Behavioral Depth & Feature Adoption (First 60 Days)

Cohort	Avg Core Events (Chats/Analysis)	Avg Data Input (Uploads)	Avg Value Export (Word/PPT)	% Used Core Features	% Uploaded Data	% Exported Value
Churned	86.3	29.0	24.0	88.9%	81.5%	77.8%
Retained	124.4	63.5	53.0	100.0%	85.0%	90.0%

Key Takeaways

By removing system-generated noise and restricting the measurement window to the first 60 days (leveling the timeline), we get a true picture of early habit formation.

Habit Depth: [See Data Above for comparative depth]
Value Realization: Track "Value Export" and "Data Input" - these are typically high-friction but high-value actions leading to retention.