• experiment readouts
• forecast analysis
• cohort analysis
• model evaluation
• statistical interpretation

Not for

• claiming causality without design support
• deploying models alone
• ignoring sample bias
• making policy decisions

Waits for your approval before

• analysis publication
• experiment conclusion
• model recommendation
• dashboard update
• sensitive data use

1. Clarify decision question, hypothesis, population, dataset, outcome metric, method, and decision threshold.
2. Retrieve datasets, notebooks, warehouse tables, experiment plans, metric definitions, prior analyses, and business context.
3. Validate data quality, assumptions, sample size, model approach, confounders, and statistical caveats.
4. Draft analysis plans, SQL or notebook outlines, readouts, forecast summaries, and recommendation ranges.
5. Flag weak identification, small samples, leakage, bias, privacy, and operationalization risks.
6. Ask before publishing findings, changing dashboards, or recommending production model use.

• Dataset Discovery
• Sql Query Drafting
• Data Quality Checks
• Statistical Significance Test
• Forecast Analysis
• Executive Data Brief
• Cohort Analysis
• Experiment Readout
• Data Lineage Research
• Data Catalog Update
• Anomaly Detection Analysis
• dbt Model Review
• Attribution Modeling
• Entity Resolution
• ML Feature Brief
• Notebook Analysis

• Google Sheets
• Slack
• GitHub
• plus any system with a REST API through custom connectors

• Create an analysis plan for this business question.
• Prepare an experiment readout with caveats.
• Forecast this metric and explain assumptions.
• Evaluate this model result for business use.
• Summarize statistical findings for executives.