/churn-analysis

name: churn-analysis
description: When the user needs to identify at-risk accounts, understand why customers are leaving, reduce churn rate, build health scores, design save plays, or create win-back campaigns.
related: [feedback-synthesis, onboarding-flow, email-marketing]
reads: [startup-context]

Churn Analysis

When to Use

Activate when a founder needs to identify at-risk accounts before they churn, diagnose churn drivers, build a customer health scoring system, design cancellation or save flows, recover failed payments, or re-engage lost customers. This includes prompts like "our churn is too high," "which customers are about to leave," "why are customers canceling," "build a customer health score," "set up dunning emails," or "create a win-back campaign." Especially relevant for seed/Series A teams managing customers manually without dedicated CS platforms like Gainsight or ChurnZero.

Context Required

• **From startup-context:** business model (B2B/B2C, subscription/usage-based), current churn rate (logo and revenue), customer segments, pricing tiers, contract terms, product usage data availability, and current retention tooling.
• **From the user:** available data sources (support tickets, Slack channels, NPS scores, usage logs, email logs, billing data), what "healthy" customer behavior looks like, any historical churn patterns, whether churn is primarily voluntary or involuntary, and the specific churn problem to solve.

Work with whatever data is available. Early-stage companies often lack formal CS systems — the skill works with support inboxes, Slack history, and spreadsheets.

Workflow

1. **Intake and baseline** — Gather all available customer data: customer lists, support tickets, Slack/communication history, NPS scores, usage data, email logs, and billing records. Establish what "healthy" looks like and identify any known churn patterns. 2. **Extract signals** — Analyze four signal categories across every account: support signals, communication signals, usage signals, and commercial signals (see framework below). 3. **Score risk** — Build a composite risk score (0-100) for each account using weighted signal categories. Higher score means higher risk. 4. **Generate save plays** — For high-risk accounts, produce specific interventions: root cause hypothesis, recommended actions, talk tracks for the CS conversation, and escalation triggers. 5. **Build the weekly scorecard** — Compile into a weekly risk report with account-by-account analysis, MRR at risk, trend data, signal distribution, and recommended focus areas. 6. **Design interventions** — For each churn driver identified, design the appropriate intervention: product fix, CS outreach, cancel flow save offer, dunning sequence, or win-back campaign.

Output Format

A churn risk report tailored to the specific request. This may include: 1. **Weekly risk scorecard** — Every account scored and tiered with signal breakdown 2. **MRR at risk summary** — Total revenue exposure by risk tier 3. **Save play briefs** — For each red/orange account: root cause, recommended action, talk track, escalation trigger 4. **Intervention designs** — Cancel flows, dunning sequences, or win-back campaigns as needed 5. **Trend analysis** — Signal distribution changes over time

Frameworks & Best Practices

Signal Extraction Categories

Analyze every account across these four signal types:

**Support signals:** Ticket volume spikes, unresolved tickets, escalation language ("frustrated," "unacceptable," "cancel"), response time degradation, repeat issues on the same topic.

**Communication signals:** Silent accounts (no contact in 30+ days), frequency decline, sentiment shifts in Slack/email, champion disengagement (the main contact goes quiet), new stakeholder asking basic questions (signals champion departure).

**Usage signals:** Login frequency drops, feature abandonment (stopped using features they previously used regularly), shallow usage (logging in but not completing core workflows), no growth in usage over time, export/data download spikes (preparing to migrate).

**Commercial signals:** Discount requests, downgrade inquiries, payment failures, renewal proximity with no expansion discussion, competitor mentions in any channel.

Risk Scoring Model

Build a composite score (0-100) by weighting individual signals:

| Signal Severity | Points | Examples | |----------------|--------|----------| | **Critical** | 25 | Explicit cancel request, competitor migration started, champion left | | **High** | 15 | Usage dropped 50%+, 3+ unresolved escalations, payment failed twice | | **Medium** | 8 | Login frequency declining, support sentiment negative, downgrade inquiry | | **Low** | 3 | Slight usage dip, delayed renewal conversation, single missed payment |

Multiple signals compound. An account with two high signals (30 points) and three medium signals (24 points) scores 54 — solidly in the Orange tier.

Risk Tiers and Response Timelines

| Tier | Score | Timeline | Action | |------|-------|----------|--------| | **Red** | 70-100 | Action this week | Executive outreach, save offer prepared, root cause identified | | **Orange** | 40-69 | Action within 2 weeks | CS outreach, intervention plan, monitor daily | | **Yellow** | 20-39 | Monitor within 30 days | Check-in scheduled, watch for signal escalation | | **Green** | 0-19 | Routine check-in | Quarterly review, expansion opportunity assessment |

The Churn Driver Taxonomy

Categorize every churn event into one of these buckets: 1. **Value gap** — Product does not solve the problem well enough 2. **Onboarding failure** — Customer never reached the aha moment (churn in first 30-60 days) 3. **Support failure** — Bad experience getting help 4. **Price sensitivity** — Too expensive relative to perceived value 5. **Champion departure** — Internal champion left the customer's company 6. **Business change** — Customer's needs changed (acquisition, pivot, shutdown) 7. **In