Visualisation basics
By the end of this module you will be able to:
- Select an appropriate chart type for a given data type and analytical question
- Identify at least three common visualisation mistakes in a described chart
- Apply WCAG colour accessibility requirements to a visualisation design
Nightingale understood something that many analysts still struggle with: the human visual system processes pattern, scale, and colour faster than text or numbers. Choosing the right visual form is not a matter of aesthetics. It determines whether your audience understands the data or misreads it.
With the learning outcomes established, this module begins by examining why visualisation works in depth.
7.1 Why visualisation works
Pre-attentive attributes (length, angle, position, hue, and size) are processed by the visual cortex in parallel before conscious attention engages. A bar chart communicates relative magnitudes in under 100 milliseconds. A table of the same numbers requires serial reading, one cell at a time.
Edward Tufte formalised this principle in The Visual Display of Quantitative Information (1983). His data-ink ratio holds that every unit of ink (or pixel) in a chart should carry information. Gridlines, decorative borders, and 3D effects that add visual complexity without adding data are what Tufte called "chartjunk."
“Above all else, show the data.”
Edward Tufte, The Visual Display of Quantitative Information (1983) - Chapter 1, Graphical Excellence
Tufte's first principle strips away decoration. Every visual element must earn its place by encoding data. This is the design philosophy behind modern tools such as D3.js, Observable, and ggplot2, all of which default to minimal chart chrome.
With an understanding of why visualisation works in place, the discussion can now turn to matching chart type to data type, which builds directly on these foundations.
7.2 Matching chart type to data type
Chart selection is analytical, not aesthetic. The wrong chart type obscures or misrepresents the data. Five chart types cover the vast majority of analytical needs:
- Bar chart for categorical comparison (sales by region, votes by party). The length of each bar encodes magnitude.
- Line chart for time series (stock price over 12 months, daily website traffic). The slope encodes rate of change.
- Scatter plot for correlation between two continuous variables (advertising spend vs revenue). Each point is one observation.
- Histogram for distribution of one numeric variable (exam scores, response times). Bars are contiguous because the x-axis is continuous.
- Choropleth map for geographic distribution (vaccination rates by local authority). Colour intensity encodes the value.
Common misconception
“A pie chart is the best way to compare categories.”
Human visual acuity for angles is poor. Comparing segment sizes in a pie chart with more than five or six segments is unreliable. A sorted bar chart allows accurate comparison because humans judge length far more precisely than angle. Pie charts are only appropriate when showing that a single dominant category accounts for a clear majority.
With an understanding of matching chart type to data type in place, the discussion can now turn to common visualisation mistakes, which builds directly on these foundations.
7.3 Common visualisation mistakes
7.3.1 Truncated y-axes
Starting a bar chart's y-axis at a value other than zero exaggerates differences. A bar representing 98 looks dramatically taller than one representing 96 when the axis starts at 95. This technique has been used deliberately in political and commercial communications to overstate change.
7.3.2 Dual-axis charts
A chart with two y-axes implies a relationship between unrelated metrics through visual proximity. A 2012 chart plotting firearm homicide rates on one axis and gun ownership on another appeared to show dramatic divergence. The visual correlation was an artefact of scale selection, not a genuine relationship. Separate charts for unrelated metrics are always clearer.
7.3.3 3D charts
Three-dimensional bar or pie charts distort magnitude perception through perspective. Front elements appear larger than identically-valued elements at the back. There is no analytical reason to use 3D for two-dimensional data. Tufte would classify every 3D pie chart as chartjunk.
“Colour is not used as the only visual means of conveying information, indicating an action, prompting a response, or distinguishing a visual element.”
W3C WCAG 2.2 (2023) - Success Criterion 1.4.1, Use of Colour
This requirement from the Web Content Accessibility Guidelines applies directly to data visualisation. A chart that uses only colour to distinguish categories will fail for the approximately 8% of men with red-green colour vision deficiency. Always provide a second visual channel: shape, pattern, or direct labelling.
Common misconception
“Colour-blind users are a tiny minority we can address later.”
Approximately 8% of men and 0.5% of women have some form of colour vision deficiency. In a meeting room of 20 people, statistically one or two cannot distinguish your red-from-green legend. The Okabe-Ito palette (2002) and ColorBrewer palettes are freely available and work for all audiences from the start.
With an understanding of common visualisation mistakes in place, the discussion can now turn to designing for colour accessibility, which builds directly on these foundations.
7.4 Designing for colour accessibility
WCAG 2.2 Success Criterion 1.4.3 requires a contrast ratio of at least 4.5:1 between text and its background at AA level. For data visualisation, this means labels, axis text, and annotations must meet contrast thresholds against the chart background.
Beyond contrast, encode information redundantly. Use both colour and shape for scatter plot categories. Use both colour and direct labelling for line charts. Use patterns alongside colour fills for bar charts. This multi-channel approach ensures that no single visual attribute is the sole carrier of meaning.
Alt text for charts should describe the key finding, not the chart type. Write "Bar chart showing UK unemployment peaked at 8.5% in 2011 Q3, declining to 3.7% by 2024 Q1" rather than "Bar chart of unemployment data." Decorative charts without analytical purpose should use an empty alt attribute so screen readers skip them.
A data analyst creates a bar chart comparing monthly sales figures across 12 months. The y-axis starts at £480,000 and the bars range from £482,000 to £498,000. A manager presents this chart at a board meeting, pointing to the 'dramatic growth.' What is the primary problem?
A public health team publishes a choropleth map of England showing vaccination rates by local authority. They use a red-to-green colour scale (red = low, green = high). Approximately what percentage of the male audience will struggle to interpret this map?
Your team needs to explore whether advertising spend correlates with quarterly revenue. You have 24 quarters of data. Which chart type is most appropriate?
Key takeaways
- Pre-attentive visual attributes (length, position, colour) are processed faster than text. Effective visualisation places key information in pre-attentive channels, following Tufte's data-ink ratio principle.
- Chart type must match data type: bars for categorical comparison, lines for time series, scatter for correlation, histograms for distribution, choropleths for geographic data.
- Truncated y-axes, dual-axis charts, 3D effects, and overloaded pie charts are the most common misleading visualisation choices. Each distorts or obscures the underlying data.
- Approximately 8% of men have red-green colour vision deficiency. Design for colour-blind users as a baseline using palettes such as Okabe-Ito or ColorBrewer.
- WCAG 2.2 requires colour not to be the sole carrier of information (SC 1.4.1) and a minimum 4.5:1 contrast ratio for text (SC 1.4.3). Alt text should describe the finding, not just the chart type.
Standards and sources cited in this module
Tufte, E.R. (1983). The Visual Display of Quantitative Information
Chapter 1 (Graphical Excellence), Chapter 4 (Data-Ink Ratio)
Foundational visualisation design principles. Data-ink ratio and chartjunk concepts used throughout this module.
Success Criteria 1.4.1 (Use of Colour) and 1.4.3 (Contrast Minimum)
Colour accessibility requirements applied to visualisation design. Mandates that colour is not the sole information carrier.
Okabe, M. and Ito, K. (2002). Color Universal Design
Full palette specification
Eight-colour palette tested for deuteranopia, protanopia, and tritanopia. Widely used in scientific publishing.
Sequential, diverging, and qualitative palettes
Colour-blind-tested palettes for cartography and data visualisation. Available at colorbrewer2.org.
Nightingale, F. (1858). Notes on Matters Affecting the Health, Efficiency, and Hospital Administration of the British Army
Polar area diagram (rose chart)
Historical example of visualisation driving policy change. Used as the opening case study.
Module 7 of 26 · Data Foundations