Churn Autopsy

Churn Autopsy

Dissect a client departure with forensic rigor and produce churn-autopsy.md: a historical record and retention playbook that turns a loss into organizational learning.

Contents

• references/inputs.md -- required and optional data to gather before analysis
• references/analysis-framework.md -- the six-phase forensic method
• references/root-cause-taxonomy.md -- root cause categories and subcategories
• references/output-template.md -- full churn-autopsy.md report structure
• references/standards.md -- objectivity, rigor, sensitivity, incomplete-data, and operating rules

Workflow

1. Collect. Gather all available inputs per references/inputs.md. Read provided files directly; query connected CRM and analytics MCP tools for account, usage, and support data. Note any gaps. 2. Organize. Build the full chronological timeline from first touch to cancellation before attempting analysis. 3. Analyze. Apply the six phases in references/analysis-framework.md systematically, without skipping: baseline, timeline of decline, root cause classification, missed warning signs, counterfactual analysis, lessons learned. 4. Classify. Assign one primary root cause and any contributing causes using references/root-cause-taxonomy.md. 5. Draft. Write the report in the structure defined in references/output-template.md. 6. Challenge. Review every conclusion against references/standards.md; attempt to disprove each finding before keeping it. 7. Finalize. Produce churn-autopsy.md with all sections complete in the current working directory or a user-specified location.

Analysis Framework

Apply all six phases systematically. Do not skip phases.

Phase 1: Establish the Baseline

Before analyzing what went wrong, establish what "right" looked like for this account.

1. Account Profile Construction

• Reconstruct the full account timeline from first touch to final cancellation
• Document the original value proposition sold and expected outcomes
• Identify all stakeholders: economic buyer, champion, end users, detractors
• Map the organizational chart as it existed at key moments
• Calculate total lifetime value realized vs projected at time of sale

2. Health Score Reconstruction

• If a health score model exists, reconstruct the score trajectory over time
• If no health score exists, build a retrospective one using available data
• Identify the "peak health" moment and the inflection point where decline began
• Benchmark this account's trajectory against retained accounts of similar size/segment

3. Expectation vs Reality Mapping

• Document what was promised during the sales process
• Document what was actually delivered and when
• Identify any gaps between promise and delivery
• Note whether the client's business context changed in ways that affected fit

Phase 2: Timeline of Decline

Construct a detailed chronological narrative of the account's deterioration. This is the core forensic work.

1. Signal Detection Timeline Build a month-by-month (or week-by-week for shorter accounts) timeline that includes:

• Usage metrics with trend arrows
• Support ticket volume and severity
• Engagement touchpoints (meetings, emails, calls)
• NPS/CSAT data points
• Stakeholder changes (departures, role changes, new hires)
• Contract events (renewals, expansions, contractions)
• Product releases and their relevance to this account
• Competitive mentions or evaluation signals
• Internal team changes (CSM transitions, rep turnover)

2. Inflection Point Identification For each significant negative shift in the timeline:

• What changed and when exactly
• Was the change sudden or gradual
• What was the proximate cause
• What was the underlying cause
• Was this signal visible at the time with existing monitoring
• If visible, was it acted upon and how
• If not visible, what monitoring would have caught it

3. Point of No Return Analysis Identify the moment when churn became inevitable:

• When did the client mentally check out
• What was the final straw vs the accumulated weight
• Was there a window where intervention could have changed the outcome
• How long before the actual cancellation was the decision effectively made

Phase 3: Root Cause Classification

Every churn has a primary root cause and typically two to four contributing factors. Classify using the taxonomy in root-cause-taxonomy.md. A churn event may have one primary cause and multiple secondary causes.

For each contributing factor beyond the primary cause:

• How much weight did this factor carry (percentage of influence)
• Was this factor independently sufficient to cause churn
• How did this factor interact with the primary cause to accelerate the outcome
• Was this factor preventable, partially preventable, or unpreventable

Phase 4: Missed Warning Signs Audit

This is the most operationally valuable section. For every warning sign that was present but not acted upon:

1. Signal Inventory Catalog every signal that, in retrospect, indicated risk:

• Declining usage metrics (specify which metrics and the decline rate)
• Increasing support ticket volume or severity
• Decreasing engagement with CSM outreach
• Negative NPS/CSAT trends
• Delayed responses to emails or meeting requests
• Reduced attendance at QBRs or trainings
• Stakeholder departures without introductions to successors
• Requests for data exports or API documentation for migration
• Social media activity mentioning competitors
• Reduction in purchased seats or licenses
• Late or disputed invoices
• Absence from user community or events they previously attended

2. Detection Gap Analysis For each signal:

• Was this signal captured in any existing monitoring system
• If captured, was it surfaced to the right person at the right time
• If surfaced, was it triaged with appropriate urgency
• If triaged, was the response effective
• Where in the detection-to-response chain did the process fail

3. Early Warning Score Rate the overall early warning system performance for this account:

• How many months before cancellation could churn have been predicted
• What was the earliest detectable signal
• What combination of signals should have triggered a red alert
• Was the account flagged as at-risk at any point, and if so, when

Phase 5: Counterfactual Analysis

What could have been done differently at each critical juncture.

1. Intervention Windows Identify each moment where a different action could have changed the outcome:

• The specific moment in time
• What actually happened (or did not happen)
• What should have happened instead
• The estimated probability that the intervention would have saved the account
• The resources that would have been required (executive time, engineering effort, financial concessions)
• Why the intervention did not happen (lack of awareness, resource constraints, process gaps, prioritization decisions)

2. Save Attempt Evaluation If any save attempts were made:

• When was the save attempt initiated relative to the point of no return
• What was offered (discounts, roadmap commitments, executive attention, additional services)
• Why did the save attempt fail
• Was the save attempt too little, too late, or misdirected
• What save approach would have had the highest probability of success

3. Hindsight Playbook Write the specific playbook that, if executed from the first warning sign, would have maximized the probability of retention:

• Step-by-step actions with timing
• Stakeholders who needed to be involved
• Resources that needed to be allocated
• Escalation triggers that should have fired
• Executive engagement that should have occurred

Phase 6: Lessons Learned and Systemic Recommendations

Transform this individual loss into organizational improvement.

1. Process Failures Identify breakdowns in existing processes:

• Onboarding process gaps that set a poor foundation
• Health monitoring gaps that missed early signals
• Escalation process gaps that delayed response
• Handoff process gaps (sales to CS, CSM transitions)
• QBR/EBR process gaps that missed the real conversation
• Renewal process gaps that started engagement too late

2. Systemic Patterns Connect this churn to broader patterns:

• Is this the same root cause as other recent churns
• Does this represent a segment-wide risk (similar accounts at risk)
• Does this reveal a product gap affecting multiple accounts
• Does this reveal a competitive threat that is broader than one account
• Are there process failures here that are likely occurring undetected elsewhere

3. Specific Recommendations For each recommendation:

• What needs to change (process, product, people, tooling)
• Who owns the change
• Priority level (critical, high, medium)
• Expected impact on retention if implemented
• Implementation complexity and timeline
• How to measure whether the change is working

4. At-Risk Account Identification Based on the patterns found in this autopsy:

• List current accounts that show similar warning signs
• Prioritize by ARR and similarity to the churned account
• Recommend immediate actions for each identified at-risk account
• Define the monitoring triggers that should be implemented to catch this pattern earlier

Input Requirements

Gather as much of the following as possible before beginning analysis. Not every input will be available for every churn event. Work with what is provided and explicitly note gaps in the data.

Required Inputs

1. Client History -- Account overview, contract dates, ARR/MRR, expansion/contraction history, renewal dates, original deal context, buyer personas involved, and any account plan documentation.

2. Engagement Data -- Login frequency, feature adoption metrics, NPS/CSAT scores over time, QBR attendance and notes, executive sponsor interactions, champion engagement levels, response times to outreach, and community/event participation.

3. Support Tickets -- Full ticket history with severity levels, resolution times, escalation paths, recurring issues, CSAT on individual tickets, and any unresolved items at time of churn.

4. Usage Logs -- Product usage trends over the full lifecycle, feature-by-feature adoption, API call volumes, seat utilization vs purchased, peak vs trough usage periods, and any sudden drops or changes in usage patterns.

5. Exit Feedback -- Cancellation reason given (if any), exit survey responses, final call/meeting notes, any correspondence from the client explaining their decision, competitor mentions, and social media or review site commentary.

Optional But Valuable Inputs

• CRM notes and activity logs from all reps who touched the account
• Internal Slack/email threads about the account
• Billing history including late payments, disputes, credits issued
• Onboarding documentation and time-to-value metrics
• Competitive intelligence about what the client switched to
• Organizational changes at the client (layoffs, M&A, leadership changes)
• Marketing engagement history (email opens, webinar attendance, content downloads)
• Product roadmap requests the client made and their disposition

Output Template

Generate a file named churn-autopsy.md in the current working directory (or a user-specified location) using the following structure.

# Churn Autopsy: [Client Name]

**Date of Analysis**: [Date]
**Analyst**: Churn Autopsy Skill
**Account Owner**: [CSM/AM Name]
**Final ARR**: [Amount]
**Lifetime Value Realized**: [Amount]
**Client Tenure**: [Duration]
**Churn Effective Date**: [Date]
**Primary Root Cause**: [Category > Subcategory]

---

## Executive Summary

[3-5 paragraph summary covering: who the client was, what happened, why it happened,
what was missed, and the single most important lesson. This should be readable by a
C-level executive in under 2 minutes.]

---

## Account Overview

### Client Profile
[Company description, size, industry, use case, original deal context]

### Contract History
| Date | Event | ARR Impact | Notes |
|------|-------|------------|-------|
| ...  | ...   | ...        | ...   |

### Stakeholder Map
| Name | Role | Relationship | Status at Churn |
|------|------|-------------|-----------------|
| ...  | ...  | ...         | ...             |

### Value Proposition
[What was sold vs what was delivered vs what was needed]

---

## Timeline of Decline

### Visual Timeline
[Month-by-month or week-by-week chronological account of key events, metrics,
and signals organized in a clear timeline format]

### Inflection Points
[Detailed analysis of each major negative shift]

### Point of No Return
[When and why the churn became inevitable]

---

## Root Cause Analysis

### Primary Cause: [Category > Subcategory]
[Detailed explanation with supporting evidence]

### Contributing Factors
| Factor | Category | Weight | Preventable | Evidence |
|--------|----------|--------|-------------|----------|
| ...    | ...      | ...    | ...         | ...      |

### Causal Chain
[How the primary cause and contributing factors interacted to produce the outcome]

---

## Missed Warning Signs

### Signal Inventory
| Signal | First Appeared | Severity | Detected | Acted Upon | Outcome |
|--------|---------------|----------|----------|------------|---------|
| ...    | ...           | ...      | ...      | ...        | ...     |

### Detection Gap Analysis
[Where the monitoring and response systems failed]

### Early Warning Assessment
[How far in advance this churn could have been predicted and what signals
should have triggered intervention]

---

## Counterfactual Analysis

### Intervention Windows
| Window | Date | What Happened | What Should Have Happened | Save Probability |
|--------|------|---------------|--------------------------|-----------------|
| ...    | ...  | ...           | ...                      | ...             |

### Save Attempt Evaluation
[Analysis of any save attempts made]

### Hindsight Playbook
[Step-by-step retention plan that should have been executed]

---

## Lessons Learned

### Process Failures
[Specific breakdowns in existing processes]

### Systemic Patterns
[Connections to broader organizational patterns]

### Recommendations
| # | Recommendation | Owner | Priority | Impact | Complexity | Timeline |
|---|---------------|-------|----------|--------|------------|----------|
| 1 | ...           | ...   | ...      | ...    | ...        | ...      |

---

## At-Risk Account Alert

### Similar Accounts
[Current accounts showing similar patterns]

### Immediate Actions Required
[Specific actions for each at-risk account]

### Monitoring Triggers to Implement
[New automated alerts and thresholds based on this autopsy]

---

## Appendix

### Data Sources Used
[List of all data sources analyzed]

### Data Gaps
[What information was unavailable and how it limited the analysis]

### Methodology Notes
[Any assumptions, estimation methods, or analytical decisions made]

Root Cause Taxonomy

Classify every churn against this taxonomy. Assign one primary root cause (Category > Subcategory) and zero or more contributing causes.

1. Product Gap

The product failed to meet the client's needs.

• Missing Feature: A capability the client needed was never built
• Broken Feature: A capability existed but did not work reliably
• Outgrown Product: The client's needs evolved beyond what the product could deliver
• Poor UX: The product was too difficult to use for the client's team
• Performance Issues: Speed, reliability, or scalability problems
• Integration Failure: Could not connect with the client's critical systems
• Security/Compliance Gap: Could not meet required security or regulatory standards

2. Relationship Failure

The human side of the partnership broke down.

• CSM Neglect: Insufficient proactive engagement from the success team
• CSM Churn: Too many CSM transitions destabilized the relationship
• Trust Erosion: Broken promises, missed commitments, or perceived dishonesty
• Executive Misalignment: Lack of executive-to-executive relationship
• Responsiveness Failure: Too slow to respond to issues or requests
• Cultural Mismatch: Communication styles or values did not align
• Onboarding Failure: Poor initial experience set a negative trajectory

3. Budget/Financial

Economic factors drove the decision.

• Budget Cuts: Client reduced spending across the board
• ROI Not Proven: Client could not justify the cost with measurable results
• Price Sensitivity: A less expensive alternative was available
• Consolidation: Client consolidated vendors to reduce complexity and cost
• Procurement Change: New procurement leadership or policies changed buying criteria
• Economic Downturn: Macroeconomic conditions forced spending reductions

4. Competitor Win

A competitor displaced the product.

• Feature Superiority: Competitor offered capabilities not available
• Price Undercut: Competitor offered similar value at lower cost
• Bundle Deal: Competitor included the capability in a larger platform purchase
• Relationship Leverage: Competitor had stronger executive relationships
• Market Narrative: Competitor won the thought leadership or analyst positioning
• Incumbent Advantage: Client returned to a prior vendor they already knew

5. Champion Loss

The internal advocate for the product departed or lost influence.

• Champion Departed: The primary advocate left the organization
• Champion Demoted: The advocate lost decision-making authority
• Champion Fatigued: The advocate grew tired of fighting internal battles for the product
• No Backup Champion: Only one person championed the product with no succession plan
• New Leadership: New executives brought their own vendor preferences
• Reorganization: Structural changes eliminated or marginalized the champion's team

6. Strategic/Organizational Change

The client's business direction shifted.

• M&A Activity: The client was acquired or merged
• Pivot: The client changed business strategy or model
• Downsizing: The client significantly reduced operations
• In-House Build: The client decided to build the capability internally
• Regulatory Change: New regulations changed the client's requirements
• Market Exit: The client exited the market segment where the product was used

Analysis Standards

Hold the autopsy to these standards throughout drafting and review.

Objectivity Requirements

• Do not assign blame to individuals. Focus on process and systemic failures.
• Present evidence for every conclusion. Label any speculation as such.
• Acknowledge uncertainty. Use confidence levels when making causal claims.
• Consider alternative explanations for each finding before settling on a conclusion.
• Distinguish between what was knowable at the time and what is only clear in hindsight.

Rigor Requirements

• Trace every claim to a specific data point, document, or testimony.
• Keep timelines precise. Use exact dates when available, approximate ranges when not.
• Quantify wherever possible. "Usage declined" is not acceptable; "Usage declined 47% over 3 months from X to Y" is.
• Compare to benchmarks. A 10% usage decline means nothing without context about what is normal.
• Test conclusions by attempting to disprove them before including them in the report.

Sensitivity Requirements

• When the autopsy reveals individual performance issues, frame them as systemic enablement failures rather than personal shortcomings.
• When the client shared feedback in confidence, note where information should be anonymized or restricted.
• When the autopsy reveals uncomfortable truths about product quality or company promises, include them anyway. The purpose of the autopsy is learning, not comfort.
• When the churn was genuinely unpreventable (rare but possible), say so clearly rather than manufacturing preventability.

Handling Incomplete Data

Not every churn will have complete data. When data is missing:

1. State explicitly what is missing in the Data Gaps section of the appendix. 2. Estimate with ranges rather than presenting single-point guesses. Label all estimates clearly. 3. Note how the gap affects conclusions. If a key data source is missing, state which conclusions are weakened. 4. Recommend how to close the gap for future analyses. If exit interviews are not being conducted, recommend implementing them. 5. Never fabricate data or invent details to fill gaps. Incomplete analysis with honest gaps is far more valuable than complete analysis built on assumptions.

Operating Principles

• Never use emojis in the output document or in any communications.
• Be direct and specific. Vague recommendations like "improve communication" are worthless. Specify exactly what communication, with whom, at what frequency, about what topics.
• Treat the autopsy as a learning document, not a blame document. Frame everything in terms of what the organization can control and improve.
• When data is provided in files (CSV, JSON, PDF, etc.), read and analyze it directly. Do not ask the user to summarize data that is available in machine-readable format.
• When CRM or analytics tools are connected via MCP, query them directly for account data, usage metrics, and support history.
• Treat every churn as preventable until the evidence proves otherwise. The default assumption is that the organization could have done better.
• Write the report so that someone who never worked with the account can fully understand what happened and why.
• Prioritize actionable findings over academic completeness. Every section should drive toward something the organization can do differently.