Churn Intervention Routing

This drill takes the churn risk scores produced by churn-signal-extraction and routes each at-risk user to the appropriate intervention. Interventions are tiered by risk severity and personalized around the user's primary churn signal.

Input

• At-risk user cohorts in PostHog (from churn-signal-extraction)
• Churn risk scores and primary signals stored in Attio
• Intercom configured for in-app messaging
• Loops configured for triggered email sequences
• n8n instance for workflow orchestration

Steps

1. Build the routing workflow in n8n

Using n8n-triggers, create a workflow triggered daily after the churn scoring pipeline completes. The workflow queries Attio for all users with churn_risk_tier in [medium, high, critical] who have not already received an intervention in the last 14 days.

Using n8n-workflow-basics, implement the routing logic:

IF risk_tier = "critical" (76-100):
  -> Create Attio task for account owner with user's signal data
  -> Send Intercom targeted message: "Your account manager will reach out"
  -> Log intervention: type=personal_outreach, risk_score, primary_signal

IF risk_tier = "high" (51-75):
  -> Enroll in Loops re-engagement sequence personalized to primary_signal
  -> If primary_signal = "activity_decay": send "We noticed you've been less active" email
  -> If primary_signal = "feature_abandonment": send "Here's what's new in [feature]" email
  -> If primary_signal = "support_escalation": send "We want to make sure [product] is working for you" email
  -> Log intervention: type=email_sequence, risk_score, primary_signal

IF risk_tier = "medium" (26-50):
  -> Trigger Intercom in-app message highlighting an unused feature relevant to their signal
  -> If primary_signal = "engagement_narrowing": show tooltip for a feature they haven't tried
  -> If primary_signal = "login_gap": show "Welcome back" message with recent product updates
  -> Log intervention: type=in_app_message, risk_score, primary_signal

2. Create intervention templates

For Intercom (in-app messages):

Using intercom-in-app-messages, create message templates for each risk tier and signal combination. Use PostHog feature flags via posthog-feature-flags to control which users see which messages. Target by the PostHog cohort (medium-risk, high-risk).

For Loops (email sequences):

Using loops-sequences, create 3-email sequences per primary signal:

• Email 1 (Day 0): Acknowledge the signal, provide value (tutorial, update, feature highlight)
• Email 2 (Day 3): Social proof — how similar users got value from the product
• Email 3 (Day 7): Direct ask — "Is there something we can help with?" with a calendar booking link

3. Implement intervention cooldowns

No user should receive more than 1 intervention per 14-day window. Track intervention history in Attio using attio-notes:

{
  "type": "churn_intervention",
  "date": "2026-03-30",
  "risk_score": 68,
  "risk_tier": "high",
  "primary_signal": "activity_decay",
  "intervention_type": "email_sequence",
  "sequence_id": "reengagement-activity-decay"
}

Before routing a user, check Attio for any intervention note in the last 14 days. If found, skip.

4. Track intervention outcomes

Using posthog-cohorts, create outcome cohorts:

• Saved: Was at-risk, received intervention, activity increased within 14 days
• Declined: Was at-risk, received intervention, activity continued declining
• Churned despite intervention: Was at-risk, received intervention, cancelled subscription

Log outcomes back to Attio and PostHog for model calibration.

Output

• Users routed to appropriate interventions based on risk tier and primary signal
• Intervention history logged in Attio for cooldown tracking
• Outcome tracking configured for measuring intervention effectiveness
• Save rate metric: (saved users / total interventions) calculated weekly

Triggers

• Run daily, 1 hour after churn scoring pipeline completes
• Respect 14-day cooldown per user
• Pause all interventions if save rate drops below 5% for 2 consecutive weeks (signals the model or interventions need recalibration)