The complete guide to prompting Hex's Notebook Agent

Context: You are analyzing [dataset description] to help [business objective].
Task: Perform [specific analysis type] to determine [specific question].
Guidelines: Use [techniques/methods/packages], focus on [key metrics/variables].
Constraints: The analysis must [limitations/requirements], avoid [pitfalls/assumptions].

Context: You are analyzing customer transaction data to help improve marketing targeting.
Task: Perform k-means clustering to segment customers based on purchasing behavior.
Guidelines: Use Python's scikit-learn, evaluate different numbers of clusters (2-10), and visualize the results with dimensionality reduction.
Constraints: Handle missing values appropriately, normalize features before clustering, and provide business interpretations for each identified segment.

I'm looking at our customer transaction data and want to improve our marketing targeting. Could you help me segment our customers based on their purchasing behavior? I'd like to use k-means clustering for this. Use scikit-learn to implement the clustering. Try different cluster counts (between 2-10) to find the optimal number. Make sure to handle any missing values in the dataset appropriately, and normalize the features before running the clustering algorithm.

I need to analyze customer churn for our subscription service. What data variables should I include, and what analytical approach would be most appropriate? Can you outline a step-by-step plan for this analysis?

1. Exploratory Data Analysis
Churn rate trends over time, by segment, and by cohort
Feature correlation analysis and usage pattern identification

2. Feature Engineering
Create recency/frequency metrics, usage velocity changes
Rolling averages for engagement, days-to-churn calculations

3. Predictive Modeling
Logistic regression for interpretability, random forest for feature importance
Time-to-churn survival analysis for more nuanced insights

4. Actionable Segmentation
Risk scoring (high/medium/low churn probability)
Intervention timing recommendations

Great, let's proceed with the first step: 
Exploratory Data Analysis
Churn rate trends over time, by segment, and by cohort
Feature correlation analysis and usage pattern identification

Meh: "Analyze the customer data."
Good: "Analyze the @customer_transactions table, focusing on the relationship between 'purchase_frequency' and 'customer_lifetime_value' columns."

Meh: "Adjust the model so it prioritizes the recent explosive growth we saw in the past 6 months"
Good: "[@Prophet Model Components Analysis] Edit this cell where we define the parameters of the Prophet model so it prioritizes the recent explosive growth we saw in the past 6 months"

Meh: "Predict customer churn."
Good: "Build a random forest classifier to predict customer churn, using feature importance to identify the top factors contributing to churn."

Meh: "Analyze marketing channel performance."
Good: "Analyze marketing channel ROI to determine which channels to increase investment in for our Q4 campaign planning. Consider both acquisition cost and customer lifetime value by channel."

# AwesomeCo Analytics Workspace Rules

## Critical Data Governance
- PII Handling: Mask email domains in all outputs (show user@****.com)
- EU Data: Filter WHERE consent_status = 'granted' for GDPR compliance
- Financial Access: Only use finance.* tables for board reports
- Test Accounts: Always exclude WHERE is_internal = false AND email NOT LIKE '%@dataco.com'

## Source of Truth Tables
- Customers: warehouse.customers_dim (updates 2am UTC daily)
- Subscriptions: billing.active_subscriptions_v2 (v1 deprecated Nov 2024)
- Usage Events: events.user_actions_cleaned (raw has duplicates)
- Revenue: finance.recognized_revenue (not billing.charges)
- Marketing: marketing.campaigns_attributed (30-day window)

## Business Definitions
- MRR: Sum of monthly_amount WHERE status IN ('active', 'paused')
- Active User: 3+ sessions in rolling 30 days, exclude internal
- Churn Date: First day of month after last payment
- LTV: Revenue until churn / (1 - gross_margin), GM = 0.75
- Qualified Lead: Score > 50 AND company_size > 10

## Calculations & Logic
- Fiscal Year: Feb 1 - Jan 31 (Q1 starts February)
- Cohorts: By signup_week, weeks start Monday
- Attribution: First-touch for acquisition, multi-touch for expansion
- Exchange Rates: Use finance.exchange_rates for month-end conversion
- Growth Rates: Month-over-month using compound formula, not simple

## Required Analysis Patterns
- Time Comparisons: Always show YoY and MoM together
- Segments: Break down by plan_type (Starter/Pro/Enterprise)
- Statistical Significance: Min 100 samples for A/B tests
- Forecasting: Use 13-week rolling average for baseline
- Anomaly Detection: Flag if metric deviates >2 std from 30-day mean

## Data Quality Warnings
- orders table: 2% duplicate rows, use orders_deduped
- user_attributes: NULL company_size for 30% of records pre-2024
- revenue_forecast: Only reliable 30 days out
- marketing_spend: Delayed by 3 days, note in analyses

## Stakeholder Preferences
- Executive Reports: Focus on trends not absolute numbers
- Sales Team: Include pipeline_stage breakdown
- Product Team: Segment by feature_flag exposure
- Finance: Show both GAAP and billing revenue

## Industry Context
- Benchmarks: Target 125% NDR (industry avg: 110%)
- Seasonality: Q4 -20% new sales, Q1 +35% (budget cycles)
- Competition: We're 2x pricing of alternatives, emphasize ROI
- Market Position: #3 by revenue, #1 by NPS in our segment

[@Prophet Model Components Analysis] Can you describe what each of these configs/parameters in this Prophet model do?

Help me discover data in the warehouse related to churn. List relevant semantic models, schemas, and tables. Summarize what metrics and columns they contain, and recommend the best starting points for analysis.

Analyze this notebook for cleanup.  
1. Scan all cells and flag any that appear unused, irrelevant, or duplicative.  
2. Map the dependency graph of cells (which depend on which).  
3. Highlight orphaned cells (not used anywhere downstream), duplicative calculations, or intermediate steps that don’t feed final outputs.  
4. Give me a concise summary of which cells are safe to delete, which are candidates for consolidation, and which are critical to keep.

This project was just shared with me by a teammate. Can you summarize the approaches they took in this analysis and any insights gleaned from the results? 

Analyze this notebook and generate a portable context prompt I can paste into a new notebook. The prompt should:  
- Restate the core business question in plain English  
- Define each key metrics, dataframes, and calculations precisely (filters, windows, denominators)  
- Summarize the data sources (tables, views) with example names  
- Note important assumptions (freshness, grain, joins)  
- Highlight the main outputs and conclusions so far  
- End with “Context Notes” about what might be adapted or extended  

Output only the portable prompt in a single code block.

Seed context:  
[Portable prompt from Prompt A goes here]  

Use this context as background for my current analysis. 
Work with me on follow-up or related questions, and suggest how I can 
extend or adapt the prior work as needed.

Cluster our customer data into meaningful segments using k-means. Use RFM (Recency, Frequency, Monetary value) as the main features and determine the optimal number of clusters using the elbow method. Visualize the clusters and describe the characteristics of each segment.

Perform hierarchical clustering on our customer dataset with the following requirements:
1. Preprocess the data by handling outliers using IQR method
2. Standardize all features
3. Use Ward's method with Euclidean distance
4. Determine optimal clusters using silhouette score and dendrogram visualization
5. Compare results with k-means and DBSCAN approaches
6. Create cluster profiles describing key characteristics
7. Recommend marketing strategies for each identified segment

Can you help me analyze customer lifetime value for our business? I'd like to understand which customer segments are most valuable over time and what factors might influence their long-term value to us. What would be a good approach to look at this?

Develop a predictive model for customer lifetime value with these specifications:
1. Create engineered features from purchase history, browsing behavior, and customer demographics
2. Handle class imbalance using SMOTE
3. Implement a gradient boosting model (XGBoost) with hyperparameter tuning
4. Use 5-fold cross-validation with stratification
5. Compare performance against a baseline random forest model
6. Evaluate using precision-recall AUC, confusion matrix, and business impact metrics
7. Implement feature importance analysis and partial dependence plots for interpretability

Create visualizations that show the relationship between marketing spend and revenue across different channels.

Design an interactive geospatial heatmap showing customer density by region using Folium. Use a red to green gradient, highlight customer counts and summary statistics in the tooltip.

Perform exploratory data analysis on our customer dataset. Include summary statistics, distribution of key variables, correlation analysis, and identify any potential outliers or data quality issues.

Conduct a comprehensive exploratory analysis of our e-commerce transaction data with:
1. Data quality assessment (missing values, outliers, duplicates)
2. Univariate analysis of all numerical and categorical features with appropriate visualizations
3. Bivariate analysis focusing on relationships between purchase amount and customer attributes
4. Temporal patterns in purchasing behavior (daily, weekly, monthly trends)
5. Cohort analysis based on customer acquisition date
6. Segmentation analysis based on purchasing behavior
7. Anomaly detection to identify unusual transactions
8. Feature engineering suggestions for future modeling

Analyze our ecommerce data to calculate key metrics including conversion rate, average order value, customer acquisition cost, and lifetime value. Create visualizations to show how these metrics have changed over the past year.

Perform a comprehensive ecommerce analysis with these components:
1. Calculate and trend key performance indicators (AOV, CVR, CAC, LTV, ROAS)
2. Analyze the customer journey with funnel visualization from product view to purchase
3. Segment customers using the RFM (Recency, Frequency, Monetary) framework
4. Perform market basket analysis to identify cross-selling opportunities
5. Analyze product affinity and build a recommendation algorithm
6. Create a cohort analysis of customer retention by acquisition channel
7. Forecast demand for top 10 products for the next quarter
8. Identify underperforming products and potential inventory issues

Create a cohort analysis to track user retention over time. Group users by signup month and show retention rates for the following 6 months. Visualize the results as a heatmap.

Develop a multi-dimensional cohort analysis framework with:
1. Time-based cohorts (acquisition date) with 12-month retention tracking
2. Behavioral cohorts based on first action (purchase vs. trial)
3. Channel-based cohorts comparing retention across marketing channels
4. Monetary cohorts comparing high vs. low initial purchase value
5. Custom cohort based on product category of first purchase
6. Calculate and visualize not just retention but also average revenue per cohort over time
7. Statistical significance testing between cohort performance
8. Predictive model to identify early indicators of long-term retention

Perform market basket analysis on our transaction data to identify products that are frequently purchased together. Use the Apriori algorithm with a minimum support of 0.01 and minimum confidence of 0.5. Visualize the top 10 rules.

Conduct a comprehensive market basket analysis with these specifications:
1. Use the FP-Growth algorithm for better performance on our large transaction dataset
2. Optimize support and confidence thresholds through iterative testing
3. Filter rules by lift > 3 to focus on meaningful associations
4. Create network visualization of product relationships
5. Segment analysis by customer type and purchase time (weekday vs. weekend)
6. Generate specific product bundle recommendations for marketing campaigns
7. Compare seasonal variation in product association patterns
8. Implement sequential pattern mining to identify common purchase sequences

Perform feature selection for our churn prediction model. Use a combination of correlation analysis, variance inflation factor, and recursive feature elimination to identify the most important predictors of customer churn.

Implement a comprehensive feature selection pipeline with:
1. Statistical filtering (remove features with >90% missing values and near-zero variance)
2. Multicollinearity analysis using VIF with threshold of 5
3. Filter methods (chi-square for categorical, ANOVA F-value for numerical features)
4. Wrapper methods (RFE with cross-validation)
5. Embedded methods (L1 regularization, tree-based importance)
6. Stability selection across multiple subsamples
7. Compare model performance across different feature subsets
8. Create importance visualizations and document feature selection justifications

Create the backend for a KPI dashboard for our sales team showing monthly revenue, sales growth, conversion rate, and top-performing products. Include year-over-year comparisons.

Design the backend to an executive KPI dashboard with these specifications:
1. Strategic layer: Company-wide KPIs with targets and variance (revenue, profit, customer growth)
2. Tactical layer: Department-specific metrics to be used with drill-down capabilities
3. Operational layer: Daily/weekly performance indicators that we can set conditional notifications on
4. Predictive elements showing forecasted performance against targets
5. Benchmark comparisons against industry standards where available

Analyze our customer reviews using NLP techniques. Extract common themes and sentiments, identify the most frequently mentioned product features, and visualize the results.

Develop a comprehensive NLP pipeline for customer feedback analysis:
1. Preprocess text data (tokenization, lemmatization, stop word removal)
2. Implement sentiment analysis using a fine-tuned BERT model
3. Extract named entities (products, features, competitors) using spaCy
4. Perform topic modeling with BERTopic to identify evolving themes
5. Create aspect-based sentiment analysis for specific product features
6. Generate word embeddings to map semantic relationships
7. Implement trend analysis to track sentiment changes over time
8. Build an interactive visualization dashboard for exploring feedback insights

Perform sentiment analysis on our customer support conversations. Categorize messages as positive, negative, or neutral, and identify common themes in negative feedback.

Implement a multi-layered sentiment analysis framework:
1. Use a fine-tuned transformer model for domain-specific sentiment classification
2. Implement aspect-based sentiment analysis to identify specific product/service elements
3. Analyze sentiment trends over time and correlate with product releases
4. Perform emotion detection beyond simple positive/negative (joy, anger, frustration)
5. Implement sarcasm and irony detection for more accurate sentiment scoring
6. Create customer sentiment profiles based on interaction history
7. Generate actionable insights and recommendations from sentiment patterns
8. Build an alert system for sudden sentiment shifts requiring immediate attention

Analyze our sales time series data to identify trends, seasonality, and anomalies. Then forecast sales for the next 6 months using an appropriate time series model, accounting for historical patterns.

Implement an advanced time series analysis framework:
1. Decompose multiple time series into trend, seasonal, and residual components
2. Test for stationarity and apply appropriate transformations
3. Identify and handle outliers and structural breaks
4. Implement multiple modeling approaches (SARIMA, Prophet, exponential smoothing)
5. Add exogenous variables (marketing spend, promotions, competitor actions)
6. Perform cross-validation using time series split
7. Generate prediction intervals with varying confidence levels
8. Implement ensemble forecasting combining multiple models
9. Design interactive visualization for exploring forecasts with different parameters