Data Visualization and Reporting: accurate encoding and honest scales
What this domain tests
Data Visualization and Reporting tests whether a chart communicates the comparison accurately and accessibly. Questions cover encoding choice, scales and baselines, color, annotation, dashboards, and honest treatment of uncertainty.
Core rules
- Rank encodings by perceptual accuracy: position on a common scale, then length, then angle and area, then color hue. Use the most accurate encoding the data allows.
- Bars encode length, so start the value axis at zero. Lines encode change and may use a non-zero axis if it is clearly labeled and does not exaggerate.
- Avoid dual y-axes; they invite false correlations. Use small multiples or indexed values instead.
- Use color for categories sparingly and never as the only encoding. Meet WCAG 2.2 contrast: 4.5:1 for normal text and 3:1 for large text and graphical objects.
- Label units, time zone, and source; annotate the event that explains a spike.
- Show uncertainty with intervals, bands, or sample sizes; do not hide small n behind a smooth line.
- Design dashboards around decisions: one question per view, consistent filters, and a visible last-refresh timestamp.
Worked example
You must show that a conversion rate moved from 2.0% to 2.2%. A bar chart from zero makes the change nearly invisible; a truncated bar axis would exaggerate it. Better: a dot plot with labeled values and 95% confidence intervals, plus the absolute and relative change. Reasoning: the dot plot uses position, the most accurate encoding; the interval shows whether the movement exceeds noise; labeling both 2.0% and 2.2% prevents the reader from overreading a small absolute gain. If you must use a line, keep the axis labeled and add a reference band for the prior period. For a 12-category breakdown, a sorted horizontal bar chart beats a pie.
Traps that cost marks
- Truncating a bar axis to dramatize a difference.
- Pie charts with many slices or 3D effects that distort area.
- Rainbow palettes that are not colorblind-safe.
- Dual axes that imply a relationship the data does not show.
- Missing units, denominators, or refresh times on a dashboard.