Choosing the Right Chart

Before picking a visualization type, answer three questions about goal, content, and audience. Then pick the simplest chart that answers the question. Increase complexity only when the simpler form cannot convey the insight.

Start with the question, not the chart

Answer three questions before sketching a single axis:

What is the goal? — Analysis (exploring data) or communication (presenting a finding)?
What do you want to show? — A comparison, a trend, a proportion, a distribution, a relationship?
Who is the audience? — Executives want high-level dashboards; analysts need granular detail; general audiences need familiar forms.

Write the chart's main statement first. It becomes a compass for every design decision that follows.

Decision framework by goal

Goal	Primary chart types	Avoid
Compare categories	Bar (vertical/horizontal), grouped bar, dot plot, bullet graph	Pie (if > 5 categories)
Show change over time	Line, area, column, slope chart	Pie/donut (no time axis)
Show proportions / part-to-whole	Stacked bar (100%), donut, waffle, treemap	Line charts
Show distribution	Histogram, box plot, violin plot, density plot	Bar chart (categorical only)
Show relationships / correlation	Scatter plot, bubble chart, heatmap	Bar charts
Show ranking	Horizontal bar (sorted), lollipop, slope chart	Pie charts
Show flow / movement	Sankey diagram, alluvial diagram	Static charts
Show hierarchy	Treemap, sunburst, tree diagram	Flat charts
Show geographic patterns	Choropleth, symbol map, bubble map	Non-spatial charts
Show deviation from baseline	Diverging bar, bullet graph, surplus/deficit line	Standard bar

Decision by data shape

Data shape	Chart types
1 numeric variable	Histogram, density plot, box plot
1 categorical variable	Bar chart, lollipop, pie/donut
Numeric + categorical (1 observation per group)	Bar chart, lollipop, dot plot
Numeric + categorical (many observations per group)	Box plot, violin plot, grouped bar
2 numeric variables	Scatter plot, line chart (if ordered)
3+ numeric variables	Bubble chart, parallel coordinates, heatmap
Multiple series over time	Multi-line, stacked area, small multiples
Hierarchical categories	Treemap, sunburst, circle packing
Network / flow	Sankey, chord diagram, network diagram

Complexity escalation rule

TIP

Start with the simplest chart that answers the question. A bar chart that communicates clearly is better than a beautiful but confusing stream graph.

Increase complexity only when the simpler form cannot convey the insight. Stream graphs, sunbursts, radar charts, and chord diagrams are powerful — and dangerous. Each added dimension must earn its place.

When a table is better

If you have fewer than 5 numbers to present, a table often communicates more effectively than a chart. Tables allow precise value comparison; charts excel at showing patterns, trends, and relationships.

Example

To display "revenue this quarter vs. last quarter, and year-over-year growth" — that is three numbers. A small table beats any chart.

Worked example: two stories, two chart types

The same dataset rarely fits every story. The two samples below answer different questions about different shapes of data — and each picks the simplest form that works.

Trend over time → line chart:

bpc

chart line {
  title = "Bitcoin surged past $90,000 in 2024"
  description = "USD, year-end closing price"
  source = "CoinGecko"
  colors = "#f7931a"

  data {
    "2016" = 963
    "2020" = 28949
    "2021" = 46306
    "2022" = 16547
    "2024" = 93429
  }
}

From packages/lib/src/samples/bitcoin-price.bpc

Position over time is what the reader needs — a line uses x-position for the date and y-position for the value. A bar chart of the same data would force length comparisons that obscure the trend.

Part-to-whole → donut chart:

bpc

chart donut {
  title = "Chrome dominates with two-thirds of the desktop browser market"
  description = "Worldwide, January 2025"
  source = "StatCounter"
  displayAsPercentage = true

  data {
    "Chrome" = 65.7
    "Edge" = 13.1
    "Safari" = 8.9
    "Firefox" = 6.3
    "Opera" = 3.1
    "Others" = 2.9
  }
}

From packages/lib/src/samples/browser-market.bpc

Six categories summing to 100 % — the question is "how big a slice does Chrome own?" A donut answers it directly; a line chart cannot.

Choosing the Right Chart ​

Start with the question, not the chart ​

Decision framework by goal ​

Decision by data shape ​

Complexity escalation rule ​

When a table is better ​

Worked example: two stories, two chart types ​

See also ​