Data Visualization and Communication
This domain tests whether you can choose an encoding that people read accurately, design a chart that survives scrutiny, and state a finding an audience can act on.
Encoding and chart choice
Perceptual accuracy, following Cleveland and McGill, ranks position on a common scale above length, then angle and slope, then area, then colour hue and saturation. So sorted bars beat pies, and dot plots beat bubbles. Use lines for ordered change, scatter for association, bars for magnitude comparison (start at zero), histograms for distribution, and box or violin plots for spread.
Design rules
- Maximise data-ink: remove gridline clutter, direct-label series instead of using a legend, and sort categories by value.
- Colour carries meaning: sequential ramps for ordered magnitude, diverging ramps around a meaningful midpoint, categorical hues for nominal groups. Avoid red-green-only palettes and check text contrast.
- Annotate the point that matters, write a title that states the finding rather than the variable names, label units, and show uncertainty with error bars, bands, or sample sizes.
Communicating findings
Lead with the conclusion, then the evidence, then the method. Match precision to the audience and put caveats where they will be read, next to the claim they qualify. One chart, one message.
Worked example
Twelve regions with revenue shares. A pie chart forces angle comparison and fails beyond roughly five slices. Instead, sort a horizontal bar chart descending, put value labels at the bar ends, and drop the legend by labelling directly. If the question is change over time, switch to a line chart with a zero baseline and annotate the one region that moved.
Traps
Truncated y-axes exaggerating small differences; dual y-axes implying a relationship that may not exist; rainbow palettes that encode no order; 3-D effects distorting area; and captions that claim more than the plotted data support.