34:29
DV - 1d - Data Visualization Basics - Part 2
Chad Skelton
Overview
This video explains five key rules or guidelines for creating effective data visualizations, emphasizing that while exceptions exist, these rules help beginners avoid common pitfalls. The rules cover avoiding pie charts and bubble charts due to their inherent difficulties in visual comparison, ensuring bar and column charts start at zero to prevent misleading representations, preferring horizontal bar charts over vertical column charts for better readability, and completely avoiding 3D charts which distort data perception. The presenter uses numerous examples to illustrate why these guidelines are crucial for clear and accurate data communication.
How was this?
Save this permanently with flashcards, quizzes, and AI chat
Chapters
- Data visualization rules are guidelines, not absolute laws, with common exceptions.
- For beginners, following these rules helps prevent frequent and obvious mistakes.
- Treat these rules as 'break glass in case of emergency' measures, to be deviated from only with strong justification.
- The primary goal is to make data understandable and avoid misinterpretation.
Understanding these foundational rules helps learners create visualizations that are clear, accurate, and avoid common errors that can confuse or mislead an audience.
The analogy of 'break glass in case of emergency' for rules, suggesting they are for general use but can be broken with good reason.
- Pie charts are often popular but are 'dangerous' because they make it difficult to accurately compare slice sizes, especially for small differences.
- The 'self-sufficiency test' highlights that without labels, pie charts often fail to convey information visually, relying solely on numbers.
- Comparing similar-sized slices or multiple pie charts is extremely challenging and prone to error.
- Exceptions exist only when the part-to-whole relationship is critical AND there are very few slices (ideally two or three).
Pie charts obscure precise comparisons and can be visually confusing, leading to misinterpretations of data proportions, especially when there are many slices or subtle differences.
A pie chart showing browser market share where it's hard to distinguish between Firefox (37.9%) and Internet Explorer (36.9%) without labels, versus a column chart where the difference is immediately obvious.
- Bubble charts are considered even worse than pie charts because they rely on area for comparison, which our brains struggle to interpret accurately.
- When comparing bubble sizes, people often misjudge the area, leading to significant underestimation of differences.
- The perceived size of a bubble is influenced by both diameter and area, creating confusion.
- Maps can be a defensible exception for bubble charts, as they are often used to represent magnitude at specific locations.
Bubble charts distort the perception of magnitude, making it difficult to accurately gauge the relative size of data points, unlike simpler charts that use a single dimension for comparison.
A Wall Street Journal bubble chart showing whistleblower lawsuits, where a bubble representing 573 cases looked only slightly larger than one representing 363 cases, while a column chart would have made the difference much clearer.
- Bar and column charts should always begin their value axis at zero to ensure accurate visual representation of magnitude.
- Starting a bar chart at a non-zero baseline exaggerates differences between bars, making comparisons misleading.
- While line charts have more debate, starting at zero is generally preferred unless small but critical changes need emphasis (e.g., global temperature shifts).
- Misleading charts often use non-zero baselines, even when labeled, to create a false sense of dramatic change.
A non-zero baseline in bar or column charts distorts the perceived scale of differences, leading the audience to believe variations are much larger than they actually are.
A chart showing income tax by province that started at $1,000, making Ontario's tax appear 2.5 times higher than BC's, when correcting to a zero baseline showed it was only about 30% higher.
- Horizontal bar charts are generally superior to vertical column charts for most data.
- Column charts often require rotating labels 90 degrees, making them hard to read, or lead to overlapping/truncated labels.
- Bar charts allow for longer, more legible labels placed horizontally.
- Sorting data from largest to smallest (or vice-versa) is more intuitive in bar charts (top-to-bottom) than column charts (left-to-right).
- Exceptions occur when data has a natural chronological or ordered sequence (like months or age groups), where a column chart retaining that order is better.
The orientation of bars significantly impacts readability. Horizontal bar charts accommodate labels better and follow a more intuitive sorting convention, leading to clearer communication.
A column chart showing homicide rates by Canadian province/territory with rotated labels versus a bar chart where all labels are easily readable horizontally.
- 3D charts, particularly 3D pie charts, are almost always a bad idea because they distort data perception.
- The 3D effect can make closer elements appear larger than they are, obscuring accurate comparisons.
- Elements in 3D charts can physically block the view of other data points, making them impossible to see.
- Most modern visualization tools avoid 3D functionality to prevent these issues.
3D charts introduce visual illusions and obstructions that prevent accurate interpretation of data, making them fundamentally misleading.
A 3D pie chart from an Apple keynote where the green slice (Apple) appeared larger than the purple slice (Other), but in reality, the purple slice was larger due to the 3D perspective and tilt.
Key takeaways
- Prioritize clarity and accuracy in data visualization by adhering to established guidelines.
- Pie charts are generally poor for comparing quantities; use them only for simple part-to-whole relationships with few slices.
- Bubble charts are problematic because the human eye struggles to accurately compare areas.
- Always start bar and column charts at zero to prevent exaggeration of differences.
- Horizontal bar charts offer better readability than vertical column charts due to label handling and sorting conventions.
- 3D charts are inherently misleading and should be avoided entirely.
- The 'self-sufficiency test' is a useful method to check if a visualization can be understood without relying solely on labels.
Key terms
Data Visualization RulesPie ChartSelf-Sufficiency TestBubble ChartBar ChartColumn ChartBaseline (Zero Baseline)Part-to-Whole Relationship3D Chart
Test your understanding
- Why are pie charts generally discouraged, and under what specific conditions might they be acceptable?
- How does the human brain's perception of area affect the interpretation of bubble charts, and why is this problematic?
- What is the significance of starting bar and column charts at a zero baseline, and what happens when this rule is broken?
- When comparing bar charts and column charts, what are the primary advantages of using a horizontal bar chart?
- Explain why 3D charts are considered misleading and should be avoided in data visualization.