Maya Patel·
Handed it a churn dataset and got a full analysis plan with chart picks and a dashboard layout
Transform raw data objectives into a complete analysis plan with chart selection, statistical tests, and dashboard layout.
Data Analysis & Visualization Blueprint
You are a lead data analyst and visualization expert at a top analytics firm. I have a dataset and business questions — help me build the complete analysis and visualization blueprint.\n\nDATASET DESCRIPTION:\n{{describe_your_dataset — columns, row count, data types, time range, source}}\n\nBUSINESS QUESTIONS TO ANSWER:\n{{list_the_key_questions_or_decisions_this_analysis_should_support}}\n\nAUDIENCE: {{who_will_consume_this — e.g., 'C-suite executives', 'product team', 'external clients'}}\n\nDELIVERABLE FORMAT: {{output_format — e.g., 'Dashboard', 'Presentation', 'Report', 'Interactive tool'}}\n\nTOOLS AVAILABLE: {{tools — e.g., 'Tableau', 'Power BI', 'Python matplotlib/seaborn', 'Excel', 'Looker', 'R ggplot2'}}\n\nOUTPUT — Provide a comprehensive Data Analysis Blueprint:\n\n## 1. ANALYSIS PLAN\nFor each business question, specify:\n- Question restated as a testable analytical problem\n- Required analysis technique (descriptive, diagnostic, predictive, prescriptive)\n- Specific method (e.g., cohort analysis, regression, segmentation, funnel analysis)\n- Required data fields and any transformation needs\n- Expected insight type (comparison, trend, correlation, anomaly, distribution)\n\n## 2. STATISTICAL APPROACH\n- Recommended statistical tests with justification\n- Confidence level and sample size adequacy assessment\n- Handling of outliers, missing data, and biases\n- Segmentation dimensions to apply\n\n## 3. VISUALIZATION BLUEPRINT\nFor each insight, specify:\n| Insight | Chart Type | Why This Chart | Color Strategy | Interaction |\n|---------|-----------|----------------|----------------|-------------|\n(Provide 5-10 rows covering all key insights)\n\n## 4. DASHBOARD LAYOUT (if applicable)\n- Wireframe description: information hierarchy, section flow, filter placement\n- KPI card definitions (top 3-5 metrics)\n- Drill-down paths from summary to detail\n\n## 5. DATA QUALITY CHECKLIST\n- 5 validation checks to run before analysis\n- Common pitfalls for this data type\n\n## 6. EXECUTIVE SUMMARY TEMPLATE\n- Pre-structure the 1-page summary with placeholders for key findings\n\nProvide chart-specific best practices (e.g., 'use line charts for time series, not pie charts for comparison'). Flag any questions that cannot be answered with the described dataset.
Ergebnisse
# Data Analysis Blueprint: SaaS Churn Drivers
**Dataset:** 18 columns, ~45k user-months, Jan 2024–May 2026, from the product DB + Stripe. **Audience:** product team. **Deliverable:** dashboard. **Tools:** Python (pandas/seaborn) + Looker.
## 1. Analysis plan
| Business question | Technique | Method | Fields | Insight type |
|-------------------|-----------|--------|--------|--------------|
| Who churns? | Diagnostic | Cohort + logistic regression | signup_date, plan, last_active, churned | Comparison |
| When do they churn? | Descriptive | Survival/retention curve | tenure_days, churned | Trend |
| What predicts churn? | Predictive | Feature importance (logistic) | usage_freq, seats, support_tickets | Correlation |
| Which feature drives retention? | Diagnostic | Segmentation by feature adoption | feature_x_used, retained | Distribution |
## 2. Statistical approach
- Logistic regression for churn probability; report odds ratios.
- Chi-square for categorical (plan × churn).
- n=45k is ample; check class balance (churn likely <10% → use stratified sampling / class weights).
- Outliers: cap usage at the 99th percentile. Missing `last_active` → flag, don't impute silently.
- Segment by: plan tier, cohort month, team size.
## 3. Visualization blueprint
| Insight | Chart | Why | Color | Interaction |
|---------|-------|-----|-------|-------------|
| Retention over tenure | Line (cohort curves) | Time series | Sequential by cohort | Hover = exact % |
| Churn by plan | Grouped bar | Category compare | Categorical | Filter by date |
| Churn predictors | Horizontal bar (odds ratios) | Ranked importance | Diverging | Click = drill |
| Usage vs churn | Box plot | Distribution compare | 2-tone | None |
| Cohort heatmap | Heatmap | Retention matrix | Sequential | Hover cell |
## 4. Dashboard layout
- **Top:** 4 KPI cards — overall churn %, MRR churn, avg tenure, at-risk count.
- **Mid:** retention curve (left) + churn-by-plan bar (right).
- **Bottom:** predictor bar chart with drill-down to user list.
- Filters: date range, plan, cohort — top-right.
## 5. Data quality checklist
1. Row count matches Stripe active-user count?
2. No duplicate user-months?
3. `churned` flag consistent with `last_active`?
4. Date range continuous, no gaps?
5. Plan values normalized (no "Pro" vs "pro")?
## 6. Executive summary template
> **Churn rate:** ___%. **Top driver:** ___. **Most at-risk segment:** ___. **Recommended action:** ___.
*Note: "why customers leave" qualitatively cannot be answered with this dataset — needs exit-survey data.*
Modell: GPT-4o
12 Likes4 SavesScore: 12
1 Kommentar
Julia Moser·
The 'never miss twice' rule quietly changed my month.