About this skill

What problem it solves

Many data tasks begin by requesting a number, but the hard parts are metric definitions, baselines, statistical validity, and business meaning. The skill covers SQL, spreadsheets, notebooks, dashboards, exports, and ad hoc tables for KPI debugging, experiment readouts, funnel or cohort analysis, anomaly review, executive reporting, and metric quality checks. It begins by asking what decision would change if the result shows X versus Y.

How it works

Method first: clarify the decision, the change-of-mind condition, the available data, and the timeframe.
Metric contract: define entity, grain, numerator, denominator, time window, timezone, filters, exclusions, and source of truth; state ambiguities explicitly.
Statistical rigor: check sample size, fair comparisons, multiple comparisons, effect size, and uncertainty ranges.
Decision-ready output: include the answer, evidence, confidence, caveats, and next action; translate method into business impact for stakeholders.
Lightweight references: load only metric-contracts.md, chart-selection.md, decision-briefs.md, pitfalls.md, or techniques.md as needed.

Boundaries

It makes no external network requests and does not depend on local persistence by default, but conclusions still depend on data quality, sample size, and confounding. If a user asks to prove a predetermined conclusion, sample size is too small, or definitions have drifted, the skill flags a downgrade or hold. It fits explainable, decision-oriented analysis, not purely syntactic code cleanup.

Screenshots

Use Cases

Review A/B test results by checking sample size, multiple comparisons, and effect size with confidence intervals.
Audit KPI definitions by clarifying numerator, denominator, time window, and filters before judging trends.
Run cohort retention analysis on SQL exports, compare retention curves across user groups, and flag anomalies.
Prepare stakeholder briefings that state the answer, evidence, limitations, and recommended next action.

Best For

Data analysts reviewing A/B tests who need to decide whether statistical evidence supports launch.
Product managers maintaining metric definitions who need to verify numerator, denominator, and time window.
BI engineers writing stakeholder reports who need to turn results into conclusions, evidence, and next steps.
Operations leads debugging funnel or retention anomalies who need to separate real shifts from aggregation noise.