Detect Customer Escalation Risk Early

Business Context

NimbusDesk is a B2B customer support SaaS platform serving 4,200 paying accounts. Over the last two quarters, formal customer escalations to account managers rose from 3.1% to 5.4% of accounts per month, and 38% of escalated accounts either downgraded or churned within 90 days.

Metric Scenario

The VP of Customer Success wants an early-warning dashboard that flags escalation risk before a customer submits a complaint or requests executive attention. Recent review shows that among accounts that escalated last month, median weekly active seats fell 18% in the prior 30 days, unresolved P1/P2 tickets rose from 1.2 to 3.8 per account, median first-response SLA breaches increased from 4% to 11%, and NPS dropped from 41 to 24 in the prior quarter. Product leadership wants to know which signals are most predictive and which should be monitored daily vs weekly.

Requirements

1. Define a primary customer escalation risk rate metric and 4-6 leading indicators you would monitor.
2. Explain which signals are leading vs lagging, and why.
3. Propose how you would segment and decompose the risk metric to identify the highest-risk customer groups.
4. Recommend threshold-based alert logic for Customer Success Managers.
5. Identify guardrails to avoid over-alerting on noisy accounts.

Data Available

• account_health_daily: account_id, date, active_seats, licensed_seats, feature_adoption_count, admin_logins, usage_minutes
• support_tickets: ticket_id, account_id, created_at, severity, status, first_response_minutes, resolution_hours, csat_score
• nps_responses: account_id, response_date, nps_score, respondent_role
• billing_accounts: account_id, plan_tier, arr, renewal_date, downgrade_flag, churn_flag
• crm_escalations: account_id, escalation_date, escalation_type, owner_team, outcome