Churn Risk Detector
Scan support tickets, Slack channels, NPS scores, and usage patterns to flag accounts showing early churn indicators. Produces a weekly risk scorecard with severity tiers, root cause hypotheses, and suggested save plays per account. Designed for seed/Series A teams where the founder or a single CSM manages all accounts manually.
by Athina AInpx gooseworks install --all # then, in Claude Code, Cursor, or Codex: /gooseworks use the churn-risk-detector skill
Churn Risk Detector
Surface accounts at risk of churning before it's too late. Aggregates signals from support, communication, and usage patterns into a scored risk report with specific save actions.
Built for: Early-stage teams with no CS platform (no Gainsight, no ChurnZero). You have a spreadsheet of customers, a Slack channel, and a support inbox. This skill turns those raw signals into an actionable churn risk list.
When to Use
- "Which customers are at risk of churning?"
- "Run the weekly churn risk scan"
- "Flag accounts I should worry about"
- "Who haven't we heard from in a while?"
- "Produce a customer health report"
Phase 0: Intake
Account Data
- Customer list — CSV or sheet with: company name, primary contact email, contract value (MRR/ARR), contract start date, renewal date (if known)
- Product/service type — What are they paying for? (Helps calibrate expected engagement)
Signal Sources (provide what you have)
- Support tickets — Export from Intercom, Zendesk, or email (CSV with: customer, date, subject, status, resolution time)
- Slack channel history — Customer Slack channel or shared channel messages
- NPS/CSAT scores — Recent survey results with scores and comments
- Usage data — Any metrics you track: logins, API calls, features used, active users (CSV export)
- Email/communication log — Last touchpoints per account (dates + context)
- Billing data — Payment failures, downgrades, discount requests
Calibration
- What does "healthy" look like? — Describe a healthy customer (e.g., "logs in daily, uses 3+ features, responds to emails within 24h")
- Known churn reasons — Why have customers churned in the past? (helps weight signals)
Phase 1: Signal Extraction
1A: Support Signal Analysis
From support ticket data, calculate per account:
| Signal | Calculation | Risk Weight |
|---|---|---|
| Ticket volume spike | >2x their average in last 30 days | High |
| Unresolved tickets | Open tickets older than 7 days | High |
| Escalation language | Keywords: "cancel", "frustrated", "alternative", "not working", "disappointed" | Critical |
| Response time degradation | Your avg response time to this customer trending up | Medium |
| Repeat issues | Same problem reported 2+ times | High |
1B: Communication Signal Analysis
From Slack/email history:
| Signal | Calculation | Risk Weight |
|---|---|---|
| Gone silent | No messages in 30+ days (was previously active) | High |
| Decreasing frequency | Message frequency dropped >50% vs prior 90 days | Medium |
| Negative sentiment shift | Tone changed from positive to neutral/negative | Medium |
| Champion disengagement | Primary contact stopped responding | Critical |
| New stakeholder questions | New person asking basic "what does this do?" questions | Medium (potential reorg) |
1C: Usage Signal Analysis (if data available)
| Signal | Calculation | Risk Weight |
|---|---|---|
| Login drop | Active users down >30% vs prior month | High |
| Feature abandonment | Stopped using a key feature they previously used regularly | High |
| Shallow usage | Only using 1 feature when they're paying for many | Medium |
| No growth | Same number of seats/users for 6+ months | Low |
| Export spike | Sudden increase in data exports | Critical (may be migrating) |
1D: Commercial Signal Analysis
| Signal | Calculation | Risk Weight |
|---|---|---|
| Discount request | Asked for pricing reduction | High |
| Downgrade inquiry | Asked about lower tier | Critical |
| Payment failure | Failed payment not resolved in 7+ days | High |
| Contract approaching renewal | <60 days to renewal with no renewal discussion | Medium |
| Competitor mention | Mentioned a competitor in any channel | High |
Phase 2: Risk Scoring
Scoring Model
Each account gets a composite risk score (0-100):
Risk Score = Σ (signal_weight × signal_present)
Weights:
Critical signal = 25 points each
High signal = 15 points each
Medium signal = 8 points each
Low signal = 3 points each
Score cap: 100Risk Tiers
| Tier | Score | Label | Action Urgency |
|---|---|---|---|
| Red | 70-100 | Critical risk — likely to churn | This week |
| Orange | 40-69 | Elevated risk — needs attention | Within 2 weeks |
| Yellow | 20-39 | Early warning — monitor closely | Within 30 days |
| Green | 0-19 | Healthy — no action needed | Routine check-in |
Phase 3: Save Play Generation
For each Red and Orange account, generate a specific save play:
Save Play Template
ACCOUNT: [Company Name]
RISK TIER: [Red/Orange]
RISK SCORE: [X/100]
MRR/ARR: $[X]
SIGNALS DETECTED:
- [Signal 1] — [Evidence: specific data point]
- [Signal 2] — [Evidence]
- [Signal 3] — [Evidence]
ROOT CAUSE HYPOTHESIS:
[1-2 sentences: What do you think is actually going wrong?
E.g., "Champion left the company and new stakeholder hasn't been onboarded"
or "They hit a technical limitation with [feature] that's blocking their primary use case"]
RECOMMENDED SAVE PLAY:
1. [Immediate action — e.g., "Schedule a call with [contact] this week"]
2. [Follow-up — e.g., "Send a personalized Loom showing how to solve [specific issue]"]
3. [Structural fix — e.g., "Assign a dedicated onboarding session for new stakeholder"]
TALK TRACK:
"[2-3 sentences the CSM/founder can use to open the conversation naturally,
without saying 'we noticed you might be churning']"
ESCALATION TRIGGER:
If [specific condition] by [date], escalate to [founder/CEO call].Phase 4: Output Format
# Churn Risk Report — Week of [DATE]
Total accounts scanned: [N]
Data sources: [list what was available]
---
## Risk Summary
| Tier | Count | Total MRR at Risk |
|------|-------|-------------------|
| 🔴 Red (Critical) | [N] | $[X] |
| 🟠 Orange (Elevated) | [N] | $[X] |
| 🟡 Yellow (Early Warning) | [N] | $[X] |
| 🟢 Green (Healthy) | [N] | $[X] |
**Total MRR at risk (Red + Orange):** $[X] ([Y]% of total MRR)
---
## 🔴 Critical Risk Accounts
### [Company Name 1] — Score: [X]/100 | MRR: $[X]
**Signals:** [bullet list]
**Root cause:** [hypothesis]
**Save play:** [specific actions]
**Owner:** [who should act]
**Deadline:** [date]
### [Company Name 2] — ...
---
## 🟠 Elevated Risk Accounts
### [Company Name] — Score: [X]/100 | MRR: $[X]
**Signals:** [bullet list]
**Recommended action:** [1-2 sentences]
---
## 🟡 Early Warning Accounts
| Account | Score | Key Signal | Suggested Action |
|---------|-------|------------|-----------------|
| [Name] | [X] | [Signal] | [Action] |
| [Name] | [X] | [Signal] | [Action] |
---
## Trends vs Last Week
- Accounts moved Red → Green: [list — wins!]
- Accounts moved Green → Yellow/Orange: [list — new risks]
- Accounts churned since last report: [list]
---
## Signal Distribution
| Signal Type | Accounts Affected |
|------------|-------------------|
| Support ticket spike | [N] |
| Gone silent | [N] |
| Usage decline | [N] |
| Competitor mention | [N] |
| Payment issue | [N] |
| Champion disengagement | [N] |
---
## Recommended Focus This Week
1. **[Account]** — [Why + what to do]
2. **[Account]** — [Why + what to do]
3. **[Account]** — [Why + what to do]Save to risk-report-[YYYY-MM-DD].md in the current working directory.
Scheduling
Run weekly:
0 8 * * 1 python3 run_skill.py churn-risk-detector --client <client-name>Cost
| Component | Cost |
|---|---|
| All signal analysis | Free (LLM reasoning) |
| Slack/email parsing | Free |
| Total | Free |
Tools Required
- Input data from CSV/sheets (support tickets, usage, NPS)
- Optional: Slack channel reading for communication signals
- No external API costs — pure analysis
Trigger Phrases
- "Which customers are at risk?"
- "Run the churn risk scan"
- "Weekly customer health report"
- "Flag at-risk accounts"
What's included
Render VO Anchored Motion Listicle
Assemble an expert/educator motion-graphic LISTICLE video ad from a config — a spoken authoritative voiceover carries a numbered listicle while N web-animated hyperframe beats (HTML plus the Web Animations API, one branded design system of alternating tiles, big hero numerals, and glass-pill callouts) are rendered frame-by-frame via Playwright and anchored to the VO's word-level timestamps, periodic color-graded B-roll windows give visual breath, and captions burn ONLY inside those B-roll windows (2-word chunks, ASS Format header carrying a Name field so none drop) with the VO mixed under a low music bed. This is the FREE deterministic assembly stage (Playwright beat render plus ffmpeg concat plus window-masked caption burn plus VO-and-music mix plus final composite) — the VO, the music bed, and the stock B-roll come from create-vo-elevenlabs, create-music-elevenlabs, and media-proxy. Use for the vo-anchored-motion-listicle format.
Render Stopmotion Hand Swatch Cycle
Assemble a stop-motion hand-swatch-cycle product-demo ad from a config — a sequence of still PLATES (one hand swiping a single-barrel cosmetic across a cream skin-patch, the barrel + swatch changing per plate while the hand, background, crop, and lighting stay locked) is PNG→mp4 loop-encoded at each plate's own stop-motion hold (fast motion frames 150–250ms, per-shade ~380ms, hero beats 1100–1800ms), concat-demuxed with HARD cuts into a silent master, closed on a Playwright HTML-rendered branded end card (serif tagline + sans subtitle + real logo SVG over a hero BG, never AI-rendered text), and muxed with a pre-sourced music track playing under the end card with a fade tail (no VO). This is the FREE deterministic assembly stage (loop-encode + concat-demux + end-card render + music mux); the master-anchor plate, shade plates, and end-card BG come from create-image-gpt-image-fal and the track from create-music-elevenlabs. Use for the stopmotion-hand-swatch-cycle format.
Render Split Screen Creator
Assemble a split-screen creator ad from a config — a two-zone vertical composite where a supplied AI-creator lip-sync take fills the BOTTOM ~48% while real 16:9 product/demo clips run uncropped in the TOP ~52%, each top clip contain-fit with a darkened blurred cover-scale fill of the same clip (never black bars), a 3px brand-color divider between the zones, the creator slice cover-fit per the per-scene VO timing, scenes hard-concatenated with the body audio being the concatenated creator VO slices, an end card held on the last sharp frame ~3s, then the ASSEMBLED cut transcribed with local Whisper (not the raw VO — concat drops inter-scene silence) and word-level captions burned in the chosen style. This is the FREE deterministic assembly + caption stage (two-zone composite + blurred fill + divider + hard-concat + end card + captions); the VO comes from create-vo-elevenlabs, the anchor from create-image-gpt-image-fal, and the whole-VO lip-sync from a paid VEED Fabric 1.0 take (a no-atom upstream input). Use for the split-screen-creator format.