| name | analyze-account-health |
| description | Summarizes B2B account health by analyzing usage patterns, engagement trends, risk signals, and expansion opportunities. Use for customer success reviews, renewal preparation, QBRs, or account prioritization. |
Analyze Account Health
Deep-dive into a B2B account's product usage to prepare for QBRs, assess renewal risk, identify expansion opportunities, or prioritize CS outreach.
Instructions
Step 0: Identify Account & Discover Schema
Get the account identifier:
- Company name, org ID, account ID, or account_id value
- Ask user if not provided
Discover the data schema:
Use BigQuery:search_catalog to find the GrowthCues Core tables (search for "account metrics" or "user metrics"). The main tables are:
fct_account_metrics_daily - Account health scores and trendsfct_account_feature_usage_monthly - Feature-level adoptionfct_user_metrics_daily - User behavior and championsdim_accounts - Account master data
Get table metadata:
Use BigQuery:get_table_info on each table to read the column descriptions. The descriptions contain the semantic context (Definition, Context, Action tags) that explain what each metric means and how to interpret it.
Step 1: Quick Health Triage
Query fct_account_metrics_daily for the account's recent metrics. Use the table metadata to identify:
- Pre-computed health signals (health_segment, health_score)
- Core engagement metrics (DAU, WAU, MAU, stickiness ratios)
- Trend indicators (volume changes, user momentum)
- Risk flags (dormant signals, usage drops)
Pull at least 30 days of data to show trends over time.
Key signals to look for:
- health_segment and health_score are pre-computed - use these as your primary classification
- Compare week-over-week trends in activity and volume metrics
- Flag accounts with usage contraction signals < 0.5 or stickiness < 0.20
- Note dormant risk flags
Use BigQuery:ask_data_insights or BigQuery:execute_sql depending on complexity.
Step 2: User-Level Analysis
Query fct_user_metrics_daily to identify specific users within the account. The table metadata will guide you to:
- Champion identification metrics (usage ranks, activity levels)
- Admin/buyer signals (first user flags)
- Engagement depth indicators (feature sophistication, session patterns)
- Lifecycle status (active, dormant, churned, resurrected)
User segments to identify:
- Champions: Top-ranked users by usage volume, high activity frequency
- Admins/Buyers: First users in account (decision-makers for renewals)
- Power Users: High feature breadth, frequent daily usage
- At-Risk: Lifecycle status indicating dormancy, or active monthly but not weekly
- Churned: Users who have stopped using the product
Calculate license utilization: active monthly users vs total users in account.
Use BigQuery:ask_data_insights to get user segmentation or BigQuery:execute_sql for precise queries.
Step 3: Feature Usage Analysis
Query fct_account_feature_usage_monthly to analyze feature-level adoption patterns. The table tracks:
- Which specific features (event_name) each account uses
- Volume and user counts per feature
- Feature penetration rates (% of account users using each feature)
- Month-over-month trends (pre-computed)
- First adoption dates
Pull at least 6 months of data to identify trends.
Patterns to identify:
- Core Features: High penetration rate (>50% of account users)
- Expanding Features: Positive month-over-month growth (>20%)
- Declining Features: Negative month-over-month change (<-20%)
- Abandoned Features: Previously used, now zero or very low volume
- Low Penetration Opportunities: <10% penetration on valuable features (training/upsell opportunity)
- New Adoptions: Recent first_used_month dates
Analysis approach based on health:
- If At-Risk/Critical: Focus on which core features are declining or abandoned
- If Healthy: Focus on advanced/premium features with low penetration (upsell signals)
Use BigQuery:ask_data_insights for pattern analysis or BigQuery:execute_sql for specific metrics.
Step 4: Contextual Analysis (Optional)
Note: GrowthCues Core focuses on behavioral data. For a complete picture, supplement with:
- CRM Data: Contract value, renewal date, expansion history
- Support Tickets: Outstanding issues, CSAT scores
- Qualitative Feedback: NPS responses, customer interviews, survey data
- Account Team Notes: CSM observations, recent interactions
If you have access to these data sources in BigQuery or other systems, correlate them with the usage patterns:
- Do support tickets align with declining feature usage?
- Are expansion opportunities backed by positive sentiment?
- Do at-risk signals correlate with contract renewal timing?
Use BigQuery:search_catalog to discover if additional customer data tables exist in the warehouse.
Step 5: Present Account Health Report
Structure output as follows:
Account Health Report: [Account Name]
Executive Summary
[2-3 sentences: Health score, key trend, primary recommendation]
Health Score: [Health Segment from DB]
Score: [health_score]/100
Classification: [health_segment value]
[One sentence rationale with supporting metrics]
Key Metrics
| Metric | Current | Trend | Status |
|---|
| MAU | X | ↑↓→ Y% | 🟢🟡🔴 |
| DAU/MAU (Stickiness) | X% | ↑↓→ Ypp | 🟢🟡🔴 |
| Active Days (7d) | X | ↑↓→ | 🟢🟡🔴 |
| Features Adopted | X | ↑↓→ Y | 🟢🟡🔴 |
| User Momentum (7d) | +/- X | ↑↓→ | 🟢🟡🔴 |
🚨 Risk Factors (if any)
- [Issue] - [Impact]
- Usage data: [specific metric/trend from queries]
- Behavior pattern: [what the data shows]
- Action: [Specific recommendation]
✅ Positive Signals
- [What's working] - [Evidence from usage metrics]
👥 User Intelligence
Champions (Leverage for advocacy)
- [User ID] (Rank #[X]): [Monthly events], [Active days], [Feature sophistication] - Action: [e.g., Case study candidate, reference call]
Power Users
- [User ID]: [Activity summary] - Action: [Enablement opportunity, beta tester]
At Risk (Engage)
- [User ID]: Last active [X days ago], lifecycle: [status] - Action: [Re-engagement strategy]
Admin/Buyer
- [User ID] (isadmin_proxy): [Activity level] - _Action: [Renewal discussion contact]
Inactive (>30 days)
- [Count] users ([X]% of total) - Risk: Unused licenses, potential downsell
📊 Feature Adoption Analysis
High Adoption (Core Value)
| Feature | Users | Penetration | Trend |
|---|
| [event_name] | X | Y% | ↑↓→ |
Growing Features (Expansion Signal)
| Feature | MoM Change | Current Users |
|---|
| [event_name] | +X% | Y users |
Declining Features (Risk Signal)
| Feature | MoM Change | Current Users | Prior Peak |
|---|
| [event_name] | -X% | Y users | Z users |
Low Penetration (Upsell/Training Opportunity)
| Feature | Users | Penetration | Opportunity |
|---|
| [event_name] | X | <10% | [Training/Upsell/Enablement] |
Abandoned Features (Investigate)
- [Feature]: Last used [X months ago], previously [Y users]
🎯 Recommendations
🔥 This Week (Critical Actions)
- [Specific action with user name/ID and expected outcome]
- [e.g., "Contact user_123 (admin) about declining login usage - dropped 60% in 7 days"]
📅 This Month (Strategic Initiatives)
- [Medium-term action with business context]
- [e.g., "Enable feature_advanced_analytics - only 8% adoption but solves their stated need"]
💰 Expansion Opportunities
- [Upsell signal with usage evidence]
- [e.g., "User growth +40% WoW, hitting seat limit - expansion conversation ready"]
📎 Report Details
- Account ID: [account_id]
- Analysis Date: [CURRENT_DATE]
- Data Timeframe: Last 60-90 days
- Data Source: GrowthCues Core (BigQuery)
- Confidence: [High/Medium/Low based on data volume and recency]
Best Practices
- Always name users - CS needs who to contact, not aggregates. Use user_id values.
- Start with metadata discovery - Use
BigQuery:get_table_info to understand available metrics before querying
- Leverage pre-computed signals - Start with health_segment and health_score, then drill into supporting metrics
- Generate SQL from schema - Use column descriptions to understand what to query and how to interpret results
- Connect metrics to actions - Every declining metric should have a recommended intervention
- Be specific in recommendations - "Contact user_123 (admin) about Feature X abandonment" not "improve engagement"
- Show trends, not snapshots - Direction matters more than point-in-time. Use month-over-month and week-over-week changes
- Column descriptions are your guide - They contain [Definition], [Context], and [Action] tags explaining interpretation
- Flag data gaps - Note low data volume, missing recent data, or accounts with sparse activity
- Prioritize by impact - Focus on changes affecting champions, admins, or multiple users
- Validate health segment - If health_segment seems wrong, explain why based on underlying metrics
- Use semantic tools when appropriate -
ask_data_insights for exploration, execute_sql for precision
Common Patterns
Churn Risks:
- health_segment = 'Critical - Dormant' or 'Critical - Usage Drop'
- Low volume change ratios indicating usage contraction
- Champion or admin users with dormant/churned lifecycle status
- Multiple core features showing significant negative month-over-month trends
- Dormant risk flags activated
Expansion Signals:
- Positive user momentum (net new users week-over-week)
- health_segment = 'Healthy' with high health_score
- Top-ranked users adopting advanced/premium features
- Low feature penetration on valuable features that solve their problems
- High stickiness ratios combined with growing user counts
Engagement Recovery (At-Risk → Healthy):
- Improving volume change ratios (approaching or exceeding 1.0)
- Increasing active days and stickiness metrics
- New feature adoptions in recent months
- Previously dormant users returning (check lifecycle status transitions)
