Analyze · Churn / Retention & Expansion Signals
Build a Churn-Risk Scoring Model
Turn your messy account-health intuition into a weighted 0–100 churn score every AM can apply consistently.
When to use
Use when different AMs have different definitions of "at risk" and you want one shared scoring rubric. Especially valuable before standing up a save-play workflow, weekly health review, or revenue forecast adjustment.
The prompt
You are a revenue operations lead at a digital marketing agency designing a churn-risk scoring model the account team will use weekly. Agency: [AGENCY_NAME] — [SERVICES] Historical churn data (last 12-24 months): [CHURN_HISTORY] For each churned client include: signals observed in the 60 days before notice (missed calls, ad spend change, payment lateness, scope changes, POC change, sentiment), MRR, tenure at churn, stated churn reason. Design a weighted 0-100 churn risk score the AM team can apply to any active client in under 5 minutes. Calibrate weights from the historical signals that actually preceded churn — not generic best practices. - Use only signals that appear in the historical data — do not invent factors - Weight signals by how often they preceded a churn event in the data - Group signals into 4-5 categories (engagement, commercial, delivery, relationship, sentiment) - Define clear red (>70), yellow (40-70), green ( 1) Signal catalog table: Category | Signal | Weight | How to measure. 2) Threshold table: Score band | Meaning | Required action | Owner. 3) Worked example on [SAMPLE_CLIENT].
Variables
[AGENCY_NAME]Your agency name[SERVICES]Services offered[CHURN_HISTORY]Per-churned-client: pre-churn signals, MRR, tenure, churn reason[SAMPLE_CLIENT]One current client to score as a worked example
Example input
Agency: Lever Studio — SEO + creative retainers. 18 churned clients in last 18 months. Common pre-churn signals: 14/18 missed 2+ reporting calls in final 60 days; 11/18 had POC change in final 90 days; 9/18 paid an invoice 30+ days late; 8/18 reduced scope; 6/18 mentioned budget in any call. Sample client to score: Harbor & Vine ($9k MRR): missed last QBR, on-time invoices, new marketing director started 6 weeks ago, asked about reducing creative deliverables.
Example output
Signal catalog: | Category | Signal | Weight | Measure | |---|---|---|---| | Engagement | Missed 2+ reporting calls in 60d | 25 | calendar logs | | Relationship | POC change in last 90d | 20 | CRM | | Commercial | Invoice paid 30d+ late | 18 | AR | | Commercial | Scope reduction request | 15 | CRM notes | | Sentiment | "Budget" mentioned unprompted | 12 | call notes | | Delivery | Project slipped >2 weeks | 10 | PM tool | Thresholds: 70+ Red (exec save call within 5 days, owner: founder). 40-69 Yellow (AM rebuild value brief within 10 days). <40 Green (standard cadence). Worked example — Harbor & Vine: missed QBR (25) + new POC (20) + scope reduction ask (15) = 60 → Yellow. Action: AM books re-onboarding with new POC + value brief within 10 days.
Pro tips
- Re-calibrate weights every 6 months as new churns come in — the model decays
- Don't ship a model with more than 10 signals — AMs won't use it
- Score the whole book once, then sanity check: if your gut and the model disagree on a client, fix the model, not your gut
Free 20 minute sales audit
A prompt is not a system.
These are free and they work. What they cannot do is make your team run the play every week. That is the part we install.
Book a Free Sales Audit