Chart Types

Chart selection starts from the analytical function of the message. This document adopts the Financial Times Visual Vocabulary’s nine functions as the normative taxonomy (Financial-Times/chart-doctor, visual-vocabulary), annotated with selection rules drawn from the BBC, The Economist, the UK Government Analysis Function, and the Urban Institute.

1 Selection rules

  • CT-1 (MUST) Choose the chart by asking what relationship the message expresses — one of the nine functions below — and pick from that function’s list. A message can serve two functions (e.g. a ranked magnitude); pick the function the title asserts.
  • CT-2 (MUST) When in doubt, use the boring chart. Bar, line, and scatter cover the large majority of published messages; exotic forms need to pay for the learning cost they impose on readers.
  • CT-3 (SHOULD) When series or categories exceed the limits in P-5 (4 lines, 4 bars per cluster, ~6 categories), switch to small multiples with shared scales instead of one dense panel.
  • CT-4 (MUST) Respect each form’s encoding constraints: zero baselines for bars/columns (P-3), whole numbers only for pictograms, rates (not totals) for choropleth maps.

2 1. Deviation — variation from a fixed reference point

Use for surplus/deficit, above/below target, positive/negative sentiment.

Chart Use when
Diverging bar Values carry sign; a standard bar handles “both negative and positive magnitude values”
Diverging stacked bar Survey sentiment scales (disagree/neutral/agree)
Spine chart One value split into two contrasting components
Surplus/deficit filled line A balance over time against a baseline or between two series

CT-5 (MUST) Diverging charts MUST place the reference value (usually zero) as a visible, aligned axis line. Encode sign with the diverging palette (03-color-palettes.qmd, CO-6) only when the difference is meaningful — The Economist: differentiate positive/negative by color “only where there is a meaningful difference, i.e. imports/exports.”

3 2. Correlation — relationship between variables

Chart Use when
Scatterplot The standard for two continuous variables
Bubble A third variable sized by circle area
Connected scatterplot How a two-variable relationship changed over time
XY heatmap Patterns between two categories; imprecise for fine differences
Line + column An amount (columns) against a rate (line)

CT-6 (SHOULD) State correlation ≠ causation in the surrounding prose, not the chart. CT-7 (MUST) Bubble sizes MUST scale by area, never by radius.

4 3. Ranking — position in an ordered list is the story

Chart Use when
Ordered bar / ordered column Default; unsorted bars hide the rank
Lollipop Draws attention to the value end; keep zero baseline where value matters
Slope chart Rank changes between two points in time or two categories
Dot strip plot Space-efficient ranks across multiple categories
Ordered proportional symbol Large value ranges where fine differences don’t matter

CT-8 (MUST) Sort by value, not alphabetically, unless the reader will look up a specific row (then alphabetical is a lookup table, and consider an actual table).

5 4. Distribution — values and their frequency

Chart Use when
Histogram The standard; keep column gaps minimal so the shape reads
Boxplot Summarize many distributions: median and range
Violin plot Complex distributions a five-number summary misrepresents
Dot plot / dot strip plot Show individual values, or change/range (min–max) per category
Barcode plot All values in a small space; best when highlighting individuals
Population pyramid Age–sex structure (back-to-back histograms)
Cumulative curve Inequality of a distribution (cumulative frequency on y)

CT-9 (SHOULD) Prefer showing the distribution over reporting a lone mean whenever spread or skew affects the conclusion.

7 6. Magnitude — size comparisons

Chart Use when
Bar / column The standard. “Must always start at 0 on the axis” (FT)
Paired bar / column Up to 2 series per category; beyond that, facet
Proportional symbol Big variations; fine differences invisible by design
Isotype / pictogram Counts of whole things only — never fractional icons
Lollipop Emphasis on the value point; zero baseline still preferable
Radar / parallel coordinates Many variables per entity; ordering of axes is critical — use rarely

CT-13 (MUST) Zero baseline, no broken axis, no 3-D (P-3). CT-14 (SHOULD) Horizontal bars when category labels are long — labels stay horizontal (TY-4).

8 7. Part-to-whole — composition

Chart Use when
Stacked column / bar Default composition form; hard to read past a few components
Proportional (100%) stacked bar Size and share at once, if data are simple
Pie / donut Familiar, but segment comparison is imprecise — see CT-15
Treemap Hierarchical composition; weak with many small segments
Gridplot / waffle Percentages as counted squares; best with whole numbers
Waterfall Composition including negative components
Arc Parliaments and similar seat allocations
Venn Schematic only, never quantitative

CT-15 (SHOULD NOT / MUST NOT) Pies SHOULD NOT be used at all when segments are similar in size, and MUST NOT exceed 4–5 segments — The Economist: pie/doughnut “should probably be abandoned in favour of stacked bar if over 4 categories”; UK Analysis Function caps pies at 5; Urban Institute flags 5+ slices as an absolute. When used: start the first sector at 12 o’clock, order segments by size, label directly.

9 8. Spatial — location or geographic pattern is the point

Chart Use when
Choropleth The standard — rates or ratios only, sensible base geography
Proportional symbol map Totals/counts (never choropleth for totals)
Dot density Individual events; annotate the pattern the reader should see
Flow map Unambiguous movement across geography
Cartogram (equalised or scaled) Units re-sized to equal or value-based areas, e.g. election results
Contour / heat map Continuous surfaces not snapped to admin boundaries

CT-16 (MUST) Use a map only when where is the story (“only when precise locations or geographical patterns in data are more important than anything else” — FT). A bar chart of regions usually communicates values better than a choropleth. CT-17 (MUST) Choropleths encode rates; totals go on symbol maps.

10 9. Flow — movement between states

Chart Use when
Sankey Flows from one condition to others; tracing outcomes of a process
Waterfall Sequenced flows, typically budgets, with +/- components
Chord Two-way flows in a matrix; powerful but expensive to learn
Network Strength and connectedness of relationships

CT-18 (SHOULD) Flow diagrams need aggressive simplification: merge minor flows into “other” before publishing.

11 Named anti-patterns

Reviewers cite these directly.

ID Anti-pattern Rule Backed by
CT-A1 Truncated or broken bar axis MUST NOT Economist (“broken scale distorts data”), UK AF, Urban
CT-A2 3-D effects, shadows, textures MUST NOT UK AF, Urban Institute
CT-A3 Dual value axes to force correlation MUST NOT (see P-3 for the narrow exception) Economist, Urban Institute
CT-A4 Misaligned zero lines on a permitted dual axis MUST NOT Economist
CT-A5 Inverted scale combined with an index to imply correlation MUST NOT Economist
CT-A6 Pie with >4–5 segments or similar-sized segments MUST NOT / SHOULD NOT Economist, UK AF, Urban
CT-A7 Spaghetti line chart (>4 undifferentiated lines) SHOULD NOT — grey the context or facet UK AF, BBC practice
CT-A8 Legend when direct labeling fits SHOULD NOT UK AF, Economist
CT-A9 Fractional pictograms MUST NOT FT Visual Vocabulary
CT-A10 Choropleth of raw totals MUST NOT FT Visual Vocabulary
CT-A11 Unordered ranking bars MUST NOT FT Visual Vocabulary
CT-A12 Radar chart with arbitrary axis order SHOULD NOT FT Visual Vocabulary