Goal: data exploration
Reader: Analysts looking for structure, correlation, outliers and missing data. Meta cluster: A — data & metrics Scenarios: 8 · Examples: 10
Scenarios
| Scenario | Tier | Engines | Example files |
|---|---|---|---|
| distribution comparison | T1 | vega | metric-distribution-small-multiples.md<br>stripplot-region-spread.md |
| correlation | T1 | echarts, vega | metric-correlation-matrix.md<br>scatter-segment-correlation.md |
| composition over time | T1 | vega | streamgraph-topic-composition.md |
| deviation analysis | T1 | vega | deviation-from-average.md |
| missing data | T1 | vega | imputed-gap-trend.md |
| multi metric screen | T1 | echarts | parallel-risk-screen.md |
| regression fit | T1 | vega | regression-trend-fit.md |
| small multiples | T1 | vega | segment-pattern-small-multiples.md |
Reading order
Correlation and regression for relationships, distributions and small multiples for shape, imputation and deviation for data quality.
Engines in this goal
echarts— report-grade charts and dashboardsvega/vega-lite— statistical views where the data needs transforms
Details, limits and anti-patterns: ../engines/echarts.md · ../engines/vega.md.
Palette
Every figure takes its colours from a theme. Pick one in
../styles/palette.md — the token contract, the theme index and the
per-engine blocks all start there.
Most figures in this set want Default. Depart from it for the medium, not the topic: Print when it may be photocopied, Contrast for low vision or large-print handouts, Vivid when it will be projected, Accessible when colour must not be the only cue, Editorial for long-form reading, Slate for technical reference, Pastel when the figure sits inside gentle prose, Nord for a cool Nordic register. Every theme is designed for a light ground — the set has no dark page.
Don't
- Drawing conclusions from a fitted line without showing the spread
- Hiding dropped records → show the imputed or excluded rows