Data Visualization and Reporting
Data Visualization and Reporting tests whether a chart, dashboard or published report communicates a defined metric accurately to a defined audience. Questions cover encoding channels, chart selection, aggregation and denominator choices, accessibility, and the governance of distribution - who receives which number, refreshed when.
Encoding and chart choice
Accuracy of reading follows the perceptual hierarchy: position on a common scale, then length, then angle and slope, then area, then colour hue. Use line charts for trend over time, bars for category comparison, scatter for relationship, histogram for distribution, and a single number with a sparkline for a headline KPI. Pie and donut charts encode by angle and area, work only for parts of a whole with few slices, and fail when slices are close in value. Dual axes invite false correlation and should be avoided or explicitly labelled.
Metrics and aggregation
Define numerator, denominator, filters and time window before choosing a visual. Averages across mixed populations produce Simpson's paradox - segment-level trends can reverse when combined. Truncated y-axes exaggerate small changes and are defensible only when clearly marked and the variation is what the decision depends on. Accessibility: contrast ratio of at least 4.5:1 for body text, never colour alone to carry meaning, alt text on every chart.
Worked example
A dashboard must show the share of revenue from the top 10 customers by region, quarterly. Twelve pies, one per region-quarter, would show neither comparison nor trend. Reasoning: the question combines parts of a whole with a time series, so use a 100% stacked column per quarter with region as the stack, plus a line overlay for the top-10 share as a percentage of total revenue. State that the denominator is total revenue, the currency, and whether returns are netted, and publish the metric definition in the report footer so regional teams cannot re-derive a different number.
Traps that cost marks
- Choosing a pie for more than four categories or for change over time.
- Comparing percentages whose denominators differ without saying so.
- Averaging pre-computed averages rather than recomputing from counts.
- Publishing a dashboard with no metric definitions or refresh timestamp.
- Using red and green as the only signal for performance status.