home / skills / ncklrs / startup-os-skills / customer-health-analyst

customer-health-analyst skill

safe

This skill guides designing health scores and churn models, turning raw data into actionable insights for proactive customer success.

npx playbooks add skill ncklrs/startup-os-skills --skill customer-health-analyst

Review the files below or copy the command above to add this skill to your agents.

Files (12)

SKILL.md

8.7 KB

---
name: customer-health-analyst
description: Expert customer health scoring and analytics guidance. Use when designing health scores, building churn prediction models, analyzing usage metrics, identifying at-risk accounts, creating executive dashboards, or performing cohort analysis. Use for leading indicator development, customer data enrichment, risk escalation frameworks, and retention analytics.
---

# Customer Health Analyst

Expert guidance for customer health scoring, predictive analytics, and data-driven customer success strategies. Transform raw customer data into actionable insights that prevent churn and drive expansion.

## Philosophy

Customer health is not a single metric — it's a predictive system:

1. **Measure what matters** — Health scores should predict outcomes, not just track activity
2. **Lead, don't lag** — Focus on indicators that predict churn before it's too late
3. **Segment for action** — Different customers need different interventions
4. **Automate detection** — Scale health monitoring across your entire customer base
5. **Close the loop** — Analytics without action is just expensive data collection

## How This Skill Works

When invoked, apply the guidelines in `rules/` organized by:

- `health-*` — Health score design, weighting, and calibration
- `indicators-*` — Leading vs lagging indicator analysis
- `churn-*` — Prediction modeling and early warning systems
- `usage-*` — Analytics and adoption metrics
- `risk-*` — Identification, escalation, and intervention
- `data-*` — Enrichment and customer 360 development
- `cohort-*` — Analysis and benchmarking
- `executive-*` — Reporting and dashboards
- `segmentation-*` — Customer tiers and scoring models

## Core Frameworks

### The Health Score Hierarchy

```
┌─────────────────────────────────────────────────────────────────┐
│                    COMPOSITE HEALTH SCORE                       │
│                         (0-100)                                 │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  ┌──────────┐  ┌──────────┐  ┌──────────┐  ┌──────────┐       │
│  │ PRODUCT  │  │ENGAGEMENT│  │ GROWTH   │  │ SUPPORT  │       │
│  │  USAGE   │  │          │  │ SIGNALS  │  │ HEALTH   │       │
│  │  (35%)   │  │  (25%)   │  │  (20%)   │  │  (20%)   │       │
│  └──────────┘  └──────────┘  └──────────┘  └──────────┘       │
│                                                                 │
├─────────────────────────────────────────────────────────────────┤
│                    COMPONENT METRICS                            │
│                                                                 │
│  Usage:        Engagement:    Growth:        Support:          │
│  - DAU/MAU     - NPS score    - Seat trend   - Ticket volume   │
│  - Features    - CSM meetings - Usage trend  - Resolution time │
│  - Depth       - Email opens  - Expansion    - Sentiment       │
│  - Breadth     - Logins       - Contract     - Escalations     │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘
```

### Leading vs Lagging Indicators

| Type | Definition | Examples | Action Window |
|------|------------|----------|---------------|
| **Leading** | Predict future outcomes | Usage decline, engagement drop | 60-90 days |
| **Coincident** | Move with outcomes | Support sentiment, NPS | 30-60 days |
| **Lagging** | Confirm after the fact | Churn, revenue loss | Too late |

### Customer Health States

```
┌─────────────────────────────────────────────────────────────────┐
│                                                                 │
│  THRIVING ──→ HEALTHY ──→ NEUTRAL ──→ AT-RISK ──→ CRITICAL    │
│    (85+)      (70-84)     (50-69)     (30-49)      (<30)       │
│                                                                 │
│  Expand       Monitor     Engage      Intervene    Escalate    │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘
```

### Health Score Components

| Component | Weight | Key Metrics | Why It Matters |
|-----------|--------|-------------|----------------|
| **Product Usage** | 30-40% | DAU/MAU, feature adoption, depth | Usage predicts value realization |
| **Engagement** | 20-25% | NPS, CSM contact, responsiveness | Relationship strength indicator |
| **Growth Signals** | 15-20% | Seat expansion, usage trend | Investment signals commitment |
| **Support Health** | 15-20% | Ticket volume, sentiment, resolution | Frustration predicts churn |
| **Financial** | 5-10% | Payment history, contract length | Financial commitment level |

### Churn Risk Factors

| Factor | Risk Weight | Detection Method |
|--------|-------------|------------------|
| Champion departure | Critical | Contact tracking, LinkedIn |
| Usage decline >30% | High | Product analytics |
| Negative NPS (0-6) | High | Survey responses |
| Support escalations | High | Ticket analysis |
| Missed renewal meeting | High | CSM activity tracking |
| Contract downgrade | Very High | Billing data |
| Competitor mentions | High | Call transcripts, tickets |
| Budget review mentions | Medium | CSM notes |

### The Analytics Stack

| Layer | Purpose | Tools/Methods |
|-------|---------|---------------|
| **Collection** | Gather raw data | Product events, CRM, support |
| **Processing** | Clean and transform | ETL, data pipelines |
| **Calculation** | Compute scores | Scoring algorithms |
| **Storage** | Historical tracking | Data warehouse |
| **Visualization** | Present insights | Dashboards, reports |
| **Action** | Trigger interventions | Alerting, automation |

### Key Metrics

| Metric | Formula | Target |
|--------|---------|--------|
| **Health Score Accuracy** | Churn predicted / Actual churn | >70% |
| **Leading Indicator Correlation** | Correlation to outcomes | >0.6 |
| **Score Distribution** | % in each health tier | Bell curve |
| **Intervention Success Rate** | Saved / Intervened | >40% |
| **Time to Detection** | Days before risk → action | <14 days |
| **False Positive Rate** | False alerts / Total alerts | <20% |

### Executive Dashboard KPIs

| KPI | Definition | Benchmark |
|-----|------------|-----------|
| **Gross Revenue Retention** | Retained ARR / Starting ARR | 85-95% |
| **Net Revenue Retention** | (Retained + Expansion) / Starting | 100-130% |
| **Logo Retention** | Retained customers / Starting | 90-95% |
| **Health Score Average** | Mean across customer base | 65-75 |
| **At-Risk Revenue** | ARR with health <50 | <15% |
| **Expansion Rate** | Customers expanded / Total | 15-30% |

### Cohort Analysis Framework

| Cohort Type | Segments By | Use Case |
|-------------|-------------|----------|
| **Time-based** | Sign-up month/quarter | Retention trends |
| **Behavioral** | Feature usage patterns | Activation success |
| **Value-based** | ARR tier | Segment economics |
| **Industry** | Vertical | Product-market fit |
| **Acquisition** | Channel/source | Marketing efficiency |

## Anti-Patterns

- **Vanity health scores** — Scores that look good but don't predict outcomes
- **Over-weighted product usage** — Ignoring relationship and sentiment signals
- **Lagging indicator focus** — Measuring what already happened
- **One-size-fits-all thresholds** — Same scores mean different things for different segments
- **Manual-only health tracking** — Can't scale without automation
- **Score without action** — Calculating risk without intervention playbooks
- **Annual calibration only** — Health models need continuous refinement
- **Ignoring data quality** — Garbage in, garbage out

Overview

This skill provides expert guidance for designing customer health scores, predictive churn models, and retention analytics. It helps transform product, support, and financial signals into actionable risk tiers and intervention workflows. Use it to build monitoring, escalation, and executive reporting that drive measurable retention outcomes.

How this skill works

When invoked, the skill applies a rules-driven framework covering health-score design, indicator classification, churn modeling, usage analytics, risk escalation, data enrichment, cohort analysis, and executive reporting. It recommends component weights, leading vs lagging indicators, detection windows, and concrete analytics stack patterns. It also suggests thresholds, calibration cadence, and automated alert/playbook definitions to turn signals into actions.

When to use it

Designing or recalibrating a composite customer health score
Building early-warning churn prediction and intervention systems
Identifying at-risk accounts and prioritizing CSM actions
Creating executive dashboards and KPIs for retention
Performing cohort or funnel analysis to find leading signals
Developing data enrichment and customer 360 workflows

Best practices

Prioritize leading indicators (usage drops, engagement gaps) over lagging outcomes
Segment scores and thresholds by cohort, ARR tier, and product usage
Weight components to reflect business priorities (usage, engagement, growth, support)
Automate detection and alerts; embed playbooks for each risk tier
Continuously validate and recalibrate models using actual churn outcomes
Ensure data quality and combine product, CRM, support, and billing sources

Example use cases

Create a 0–100 composite health score with product usage (35%), engagement (25%), growth (20%), and support (20%)
Flag accounts with >30% usage decline and negative NPS for immediate outreach within a 14-day window
Build a churn prediction model with >70% accuracy and monitor leading indicator correlation >0.6
Design an executive dashboard showing average health, at-risk ARR, gross/net retention KPIs
Run cohort analysis by sign-up month to surface activation bottlenecks and improve onboarding

FAQ

How often should I recalibrate health scores?

Recalibrate continuously when feasible; run automated model retraining or validation at least monthly and review thresholds quarterly.

Which signals are highest priority for early detection?

Usage decline, champion departure, missed renewal interactions, negative survey responses, and escalations are top leading signals.