Analyze Data and Deliver Business Insights

This agent analyzes datasets, performs statistical analysis, builds predictive models, and delivers actionable business insights through data-driven recommendations. It starts with data discovery and understanding - examining datasets, schemas, and sources, identifying key metrics and business context, documenting data quality issues and anomalies, and defining analytical objectives from business questions.

Exploratory data analysis follows: descriptive statistics and data profiling, visualizations to surface patterns and trends, correlation analysis and feature exploration, and identifying outliers, seasonality, and distributions. For SQL/BigQuery analysis it writes optimized queries for extraction and transformation, uses window functions, CTEs, and complex joins, builds aggregate tables and summary statistics, and performs cohort, funnel, or time-series analysis.

Statistical analysis and modeling applies appropriate tests (t-tests, chi-square, ANOVA), builds predictive models (regression, classification, clustering), validates performance via cross-validation, and interprets coefficients and feature importance. It follows a simple-first model-selection approach: start with interpretable models (linear/logistic regression), use cross-validation to prevent overfitting, weigh interpretability against accuracy per business constraints, and document feature engineering choices. Business intelligence work translates statistical findings into insights, quantifies impact and potential ROI, identifies actionable next steps, and produces an executive summary.

The output is a structured Data Analysis Report: an executive summary (3-5 key findings, primary recommendation, expected impact/ROI), a data overview (description, sample size, time period, quality assessment), key insights with confidence levels, the SQL queries used, recommendations split into immediate (0-30 days), medium-term (1-3 months), and long-term (3-12 months) actions, and a technical appendix with model performance metrics, statistical test results, and stated assumptions/limitations. Throughout, it maintains statistical rigor (confidence intervals, p-values, effect sizes), frames findings in business terms, validates data quality before analysis, and documents methodology for reproducibility, including sensitivity analysis and holdout-set testing where possible.

Use this agent when a dataset needs to become a statistically rigorous, business-actionable report - trend analysis, cohort/funnel analysis, or predictive modeling with quantified confidence and ROI framing. It is well suited to teams with SQL/BigQuery access to the underlying data. It is not meant for simple descriptive dashboards with no statistical testing needed, and it does not replace a data engineering pipeline - it analyzes data it's given access to, it doesn't build the underlying data infrastructure.

Input: access to datasets/schemas and the business question driving the analysis.

Output: a Data Analysis Report with an executive summary, key insights with confidence levels, the SQL used, phased recommendations, and a technical appendix. Example report structure the agent follows:

# Data Analysis Report

• Key findings (3-5 bullet points)
• Primary recommendation
• Expected impact/ROI

1. Immediate Actions (0-30 days)
2. Medium-term Initiatives (1-3 months)
3. Long-term Strategy (3-12 months)

Works directly with SQL/BigQuery for data extraction and transformation, and produces visualizations and reports meant to be shared with business stakeholders alongside the underlying queries for reproducibility.

Data/analytics teams delivering statistically grounded business recommendations, and product or growth stakeholders who need cohort/funnel/predictive analysis translated into confidence-scored, ROI-framed action items.