Churn & Reactivation Analysis
Churn & Reactivation Analysis (Churn) quantifies the duration distribution of user churn and the extent of reactivation, answering "how long have users been gone, how many have come back, and what triggered their return". It corresponds to the backend AnalyticsService.Churn.
1. What Questions It Answers
- What does the "silence duration" distribution of churned users look like? (how many churned 30-60 / 60-90 / 90+ days ago)
- How many churned users have recently returned? What is the reactivation rate?
- Which event triggered a user's return? (e.g. a campaign, a feature)
2. Key Metrics
| Metric | Meaning |
|---|---|
| Churn duration buckets | Churned user counts per tier: 30_60d / 60_90d / 90_plus |
| Churned users | churned_users (last active more than churn_days ago) |
| Reactivated users | reactivated_users (previously churned, active again within the last reactivation_days) |
| Reactivation rate | reactivation_rate = reactivated / churned |
| Top reactivation triggers | What returning users most often do after coming back (triggers) |
3. Backend API
gRPC: AnalyticsService.Churn
| Field | Type | Description |
|---|---|---|
time_range | TimeRange | Time range (end_ms is "today"; start_ms starts the reactivation observation window) |
app_id | uint32 (optional) | Filter by app |
churn_days | uint32 (optional) | Days defining churn (last active more than N days ago = churned), default 30 |
reactivation_days | uint32 (optional) | Reactivation window (active again within the last N days = reactivated), default 7 |
Response ChurnResponse: churn_buckets[] (churn duration distribution), churned_users, reactivated_users, reactivation_rate, triggers[] (ReactivationTrigger contains event_name / count / percentage).
HTTP (admin proxy)
POST /admin/v1/analytics/churn
Content-Type: application/json
{
"timeRange": { "startMs": 1718169600000, "endMs": 1718774399000 },
"churnDays": 30,
"reactivationDays": 7
}
4. Typical Scenarios
Assessing churn depth
Look at churn_buckets: a high 90_plus share means many users have been gone a long time (hard to win back); a high 30_60d share means recent churn (still recoverable).
Win-back campaign effectiveness
Compare reactivated_users and reactivation_rate before and after a campaign to gauge its actual effect.
Reactivation touchpoint analysis
Look at triggers: the events returning users most often trigger after coming back are usually the features/content that truly drew them in, and can guide follow-up operations.
5. Notes and Caveats
- Churn threshold:
churn_daysdefaults to 30 days; product rhythms differ (games may use 14 days, SaaS 60 days) — adjust to your business. - Reactivation window:
reactivation_daysdefaults to 7 days; a longer window counts more returns but may include "accidental opens". - Relation to Lifecycle: this model focuses on "churn depth + reactivation", while User Lifecycle gives a more macro stage distribution — the two complement each other.
- Depends on activity records: churn/reactivation classification is based on last active time and requires events to be reported continuously.