data-scientist skill

Overview

This skill provides statistical modeling and business analytics expertise for A/B testing, causal inference, customer lifetime value (CLV) and churn modeling, and time series forecasting. It is designed to produce reproducible analyses and production-ready code in TypeScript, integrating with AI-assisted developer workflows. The skill emphasizes clear project phases and one-phase-per-response delivery for large analyses.

How this skill works

The skill inspects datasets, experiment designs, and business questions to recommend appropriate statistical methods and produce TypeScript implementations, tests, and documentation. For experiments it runs power calculations, randomization checks, and treatment effect estimation; for customer analytics it implements CLV, churn risk models, and cohort analyses; for time series it fits forecasts with diagnostics. Outputs include model specifications, evaluation metrics, code snippets, and recommended next steps.

When to use it

• Designing or analyzing A/B experiments and estimating treatment effects
• Estimating causal effects from observational data with appropriate controls
• Calculating CLV or building churn prediction pipelines for customer analytics
• Building time series forecasts for demand, revenue, or KPIs
• Converting analytic results into production-ready TypeScript code and tests

Best practices

• Follow the chunking rule: for large analyses generate one phase per response (EDA, feature engineering, modeling, evaluation, recommendations)
• Start with clear business questions and pre-registered hypotheses for experiments
• Validate assumptions: balance checks, overlap, stationarity, and model diagnostics
• Include reproducible tests and documentation with every model or pipeline change
• Prefer interpretable baselines before complex models; quantify uplift and uncertainty

Example use cases

• Run power analysis and analyze an A/B test to estimate lift on conversion rates with confidence intervals
• Perform causal inference using propensity score methods to estimate incremental revenue from a marketing campaign
• Build a CLV model and deploy a TypeScript scoring function with unit tests for real-time segmentation
• Train a churn prediction model, produce feature importances, and recommend retention actions
• Forecast monthly demand using ARIMA/ETS or modern state-space models and produce forecast intervals

FAQ