Chart selection is an argument decision, not a design decision

THE PROBLEM

Three-quarters of student spreadsheet submissions use pie charts. One student used a pie chart to show how scores changed across four grading periods. Another used one to compare five students’ attendance. Neither is an appropriate use of a pie chart. Chart type implies a data relationship. The wrong chart type argues for a relationship the data does not support.

THE THREE MOST COMMON CHART TYPES

Bar chart: Comparing discrete, named categories. “Which one is bigger?” Test scores by school, waste by weekday, sales by product. Reordering the categories would not change the meaning. If it would, use a line chart.

Line chart: Showing change over time. “How did this change?” The horizontal axis is continuous (usually time) and the line between points implies something happened in between. Wrong use: connecting discrete, named categories with a line implies a continuous relationship that does not exist.

Pie chart: Parts of a real whole, with fewer than five slices. “What proportion of the total?” Both conditions must be true: the pieces actually sum to 100% of a meaningful whole, and there are four or fewer slices. Human perception cannot reliably compare the areas of small pie segments.

THE CLAIM TITLE TEST

Description: “Test Scores by Grading Period” — tells the reader what the chart shows.

Claim: “Scores Rose Each Quarter After Tutoring Began” — tells the reader what the chart means.

The claim title is the evidence that the student analyzed the data rather than only displaying it. Requiring claim titles for every chart requires students to do the analytical work the chart is supposed to support.

PRACTICAL STARTING POINTS
  • 1. Post a three-rule chart guide. Bar: comparing named categories. Line: showing change over time. Pie: parts of a real whole, five slices or fewer.
  • 2. Require a claim-based title for every chart. Not a description of the data. A statement of what the data establishes.
  • 3. Show one wrong-chart-type example with familiar data. Students who identify the mismatch once are significantly less likely to repeat the error.
  • 4. Add chart type selection to the rubric. Does the chart type accurately represent the data relationship? Make the selection a graded decision.

Sources: Tufte, Visual Display of Quantitative Information (2001); Few, Show Me the Numbers (2004); Evergreen, Effective Data Visualization (2017)