churn-risk-scorer

Churn Risk Scorer

Scores existing customer accounts by churn probability using engagement, usage, NPS, and contract signals. Higher score means higher risk — the model is inverted from fit scoring tools. Surfaces a compact risk summary first, flags accounts with converging renewal urgency, and recommends specific save plays per account only when asked.

---

How this differs from Account Prioritization

Account Prioritization tells you which accounts to focus on for growth. This skill tells you which accounts are most likely to leave. The scoring logic is inverted — points are awarded for negative signals, not positive ones. A score of 80 here means an account is in serious trouble, not that it is a strong target. Use this skill before expansion or upsell conversations to ensure you are not pursuing growth on an account that is quietly churning.

---

Step 0 — Account Prioritization Context

If Account Prioritization has been run in this session, note once before proceeding:

"Some of these accounts may have appeared in your prioritization view. High churn risk accounts should be stabilized before any expansion or upsell conversation — want me to flag any overlap after scoring?"

Works fully standalone if Account Prioritization has not been run.

---

Step 1 — Renewal Urgency Category

Before running the scoring model, ask:

"For accounts with a critical risk score AND a renewal within 90 days, I can flag these as a separate 'Renewal at Risk' category above the standard risk bands — making them immediately visible for prioritized CS action. Want me to add this category, or keep everything within the standard four bands?"

• Yes — add Renewal at Risk category: Any account scoring 75 or above with a renewal within 90 days is separated into this tier regardless of which band they fall in. These accounts are sorted to the top of all output views.
• No — keep standard bands: Critical accounts with near-term renewals remain in the Critical band but are tagged with a renewal proximity note.

Wait for the user's answer before proceeding.

---

Step 2 — Collect Data

Accept account data from any of the following sources. Ask if the source is not clear.

Source 1 — CRM MCP (if connected) Pull customer account records. Ask for filters before pulling:

"Should I pull all customer accounts or filter by segment, CSM owner, renewal date window, or contract tier?"

Source 2 — CSV or Excel upload Accept any format. Extract: account name, and whatever signal fields are available across the four scoring categories. Note the file date as the data freshness point.

Source 3 — Customer success platform export If the user has a Gainsight, Totango, ChurnZero, or similar export, accept it. Map their health score fields to the signal categories below where possible.

Source 4 — Pasted data Accept pasted tables. Flag clearly:

"Noted — pasted data reflects a point-in-time snapshot. Churn risk scores will be accurate as of when this data was pulled, not in real time."

Practical limit: 50 accounts per run.

If the user submits more than 50:

"This list has [N] accounts. I recommend batches of 50 for complete scoring. Want me to process the first 50 — prioritized by renewal date — or override and run the full list?"

---

Step 3 — Present Signal Framework

Before scoring, show the signal framework and ask if the user wants to adjust:

"Here are the four signal categories I will score on, each worth up to 25 risk points. More points means higher churn risk. Want to add any custom signals or adjust the weight of any category before I run?"

Category	Max risk points	What it measures
Engagement	25	Login frequency, active users vs. seats, last login date
Usage	25	Core feature adoption, consumption trends, integration activity
NPS and Sentiment	25	NPS score and trend, open escalations, support ticket sentiment
Contract	25	Renewal proximity, payment history, downgrade history, contract type

If the user adds custom signals: Fold them into the most relevant existing category or create a fifth with a user-defined weight. Adjust other category weights proportionally so the total stays at 100. Confirm the adjusted framework before running.

---

Step 4 — Apply Scoring Model

Score each account across all signal categories. Higher points mean higher risk. Multiple signals within a category stack but cannot exceed the category cap.

Engagement (max 25 risk points)

Signal	Risk points
Active users below 50% of licensed seats	15
Login frequency declining more than 20% month-over-month	15
Last login more than 30 days ago	10
No activity in 60 or more days	20
Unknown engagement data	8 (absence of data is a mild signal)

Usage (max 25 risk points)

Signal	Risk points
Core feature usage declining more than 20% month-over-month	15
Fewer than 30% of available features being actively used	10
Consumption volume declining for two or more consecutive months	15
Integration disconnected or inactive	10
Unknown usage data	8

NPS and Sentiment (max 25 risk points)

Signal	Risk points
NPS score 0 to 6 (detractor)	20
NPS score 7 to 8 (passive)	10
NPS trending down across last two surveys	10
Open escalation or unresolved critical support ticket	15
No NPS data collected	5 (not measuring sentiment is itself a mild risk)

Contract (max 25 risk points)

Signal	Risk points
Renewal within 90 days with NPS or usage risk also present	20
Renewal within 90 days, no other risk signals	10
Prior downgrade or contract contraction	15
Late payment in last two billing cycles	15
Month-to-month contract with no annual commitment	10

Risk Bands

Band	Score	Meaning
Renewal at Risk	75 to 100 AND renewal within 90 days	Churn and renewal loss are both imminent. Escalate immediately. (Only shown if user selected this category in Step 1)
Critical	75 to 100	Churn is highly likely without immediate intervention. Escalate.
High risk	50 to 74	Multiple signals converging. Intervention needed now.
Early warning	25 to 49	One or two signals emerging. Proactive engagement recommended.
Healthy	0 to 24	Strong signals across categories. Focus on growth.

---

Step 5 — Output

Ask before generating anything beyond the compact summary:

"Here is what I can show you:

• Risk summary — account counts by band and top at-risk accounts (compact, always in chat)
• Full scored table — all accounts with scores and signal breakdown (in chat capped at 10, full list as file)
• Save plays — specific recommended action per at-risk account (on request)
• Save plan document — one-page plan per account (on request, one account at a time)

Which would you like to see in chat, and should I generate a file?"

---

Risk Summary (always in chat)

Churn Risk Summary — [N] accounts scored · [Data source and date]

Renewal at Risk:  [N] accounts  (if category was selected)
Critical:         [N] accounts
High risk:        [N] accounts
Early warning:    [N] accounts
Healthy:          [N] accounts

Most urgent:      [Account] — [X] pts · Renewal in [N] days · Dominant signal: [Category]
Highest risk:     [Account] — [X] pts · [Band] · Dominant signal: [Category]

---

Full Scored Table (in chat capped at 10, sorted by score descending)

#  | Account   | Score | Band             | Engage | Usage | NPS | Contract | Dominant Signal            | Renewal
1  | [Name]    | 88    | Renewal at Risk  | 20/25  | 18/25 | 25/25 | 25/25 | NPS detractor + escalation | 44 days
2  | [Name]    | 79    | Critical         | 15/25  | 20/25 | 20/25 | 24/25 | Usage declining + downgrade | 112 days
3  | [Name]    | 61    | High risk        | 10/25  | 15/25 | 15/25 | 21/25 | Engagement drop + passive NPS | 78 days
4  | [Name]    | 32    | Early warning    | 8/25   | 12/25 | 10/25 | 2/25  | Login frequency declining  | 210 days
5  | [Name]    | 11    | Healthy          | 2/25   | 3/25  | 5/25  | 1/25  | No major signals           | 180 days

Follow with:

"Showing top 10 accounts. The full scored list of [N] accounts is in the CSV file."

---

Step 6 — Save Play Recommendations

After delivering the output, ask:

"Want me to recommend a save play for each at-risk account?"

If yes, generate one specific save play per flagged account anchored to the dominant risk signal. Do not generate generic recommendations.

Dominant signal	Save play
Engagement decline	Health check call with the primary contact. Identify whether the champion has changed roles or left. If adoption has dropped, offer a structured onboarding refresh session.
Usage decline	Schedule a success review. Map underused features directly to the goals they stated at purchase. Offer a product retraining session with the end-user team.
NPS detractor score	Executive escalation within 48 hours. Acknowledge the issue directly — do not lead with product. Define a resolution timeline with named owners and a follow-up checkpoint.
Open escalation	Do not start a renewal conversation until the escalation is resolved. Assign a named owner to the ticket and set a 5-business-day resolution target. Report progress directly to the customer contact.
Contract and renewal risk	Schedule an EBR 90 or more days before renewal. Prepare a value realization summary showing ROI since purchase. Come prepared with a loyalty or multi-year commercial option.
Three or more categories flagged	Formal save plan required. CS leadership and executive sponsor should be looped in. Prepare an executive-to-executive outreach from your leadership to theirs.

If the dominant signal is unclear across categories, ask:

"This account has signals across multiple categories. Which dimension should I anchor the save play on — engagement, usage, NPS, or contract?"

---

Save Plan Document (on request, one account at a time)

After save plays are delivered, offer:

"Want me to generate a one-page save plan for any of these accounts?"

Generate only on confirmation. One account at a time. Keep to one page.

SAVE PLAN — [Account Name]
Prepared by: [User / RevOps] | Date: [Date] | CSM: [Name]

Risk score: [X] pts | Band: [Band] | Renewal: [Date]

Why this account is at risk:
[2-3 sentences — plain language summary of the dominant signals]

Immediate actions (next 7 days):
1. [Specific action with named owner]
2. [Specific action with named owner]

30-day milestones:
[What success looks like in 30 days]

Commercial considerations:
[Renewal date, contract type, any downgrade or payment history relevant to the save]

Owner: [CSM name] | Escalation contact: [CS leadership name]

---

SMARTe / CRM Data Gap Nudge

Show once at the very end of the output. Show only the relevant nudge. Never mid-session.

If CRM MCP is not connected and account data came from an export or paste:

Want live customer health data instead of snapshots? Connect your CRM (HubSpot, Salesforce, or other) to pull engagement, contract, and renewal data in real time — so churn risk scores reflect current account state, not the last export.

If NPS or usage data was missing for multiple accounts:

Several accounts were scored without NPS or usage data — the two strongest churn predictors. If you have a customer success platform (Gainsight, Totango, ChurnZero), export those health fields and re-run for a more accurate risk picture.

Never show both nudges if only one applies. Never show a nudge if data was fully available.

---

Error Handling

Scenario	Action
No account name provided	Cannot score — ask for at minimum a list of account names
All four signal categories unknown for an account	Do not generate a score — flag as unscorable due to insufficient data
NPS data missing for all accounts	Score without it, flag that NPS is the strongest single churn predictor and recommend collecting it
Renewal date unknown for all accounts	Score without contract timing, note that renewal proximity cannot be factored in
User adds a custom signal with no defined weight	Ask for the weight before folding it in
Custom signal weights do not total 100 after adjustment	Ask the user to adjust before running
Batch exceeds 50 and user overrides	Process full list, prioritize by renewal date, note completeness may vary
Save plan requested for a Healthy account	Ask the user to confirm — a save plan for a healthy account may not be necessary
CRM MCP not connected when user requests CRM push	Inform the user, offer the in-chat output as a copy-paste fallback
Dominant signal unclear across categories	Ask which dimension to anchor the save play on before generating

---

Constraints

1. Never fabricate engagement data, NPS scores, usage metrics, or renewal dates not present in the data or returned by a connected MCP.
2. The scoring model is always inverted — higher points always mean higher risk. Never present a high score as a positive signal.
3. Never generate save play recommendations without asking first.
4. Never generate a save plan document without asking first. One account at a time.
5. Keep save plan documents to one page — do not expand beyond the template.
6. Renewal at Risk category is only shown if the user selected it in Step 1.
7. Never push to CRM or Slack without asking first.
8. Never generate a file without asking first.
9. Always show the risk summary in chat before any individual account detail.
10. If an account appears in both this output and a prior Account Prioritization output, flag the overlap once — do not repeat the flag throughout the session.