Data Visualization and Reporting
Data Visualization and Reporting tests whether you can pick an encoding that lets a reader answer a specific question, and build reporting that stays trustworthy when definitions change.
Core rules and concepts
- Match chart to question: change over time → line; comparison across categories → sorted horizontal bar; distribution → histogram or box plot; relationship between two measures → scatter; part-to-whole → stacked bar or treemap (pie only for two or three slices). Ranking needs sorted bars, not alphabetical order.
- Encoding accuracy: position on a common scale is read most accurately, then length, then angle and area, then colour intensity. Use position when the decision depends on small differences.
- Axes: bar charts start at zero because length encodes magnitude; line charts may truncate a labelled axis because slope encodes change. Dual axes invite false correlation - replace with indexed series or two aligned panels.
- Design for scanning: the title states the takeaway, not the measurement ("Churn rose in the west", not "Churn chart"); one primary metric per view; direct labels over legends; preattentive attributes (colour, size) used sparingly and consistently; annotate where something changed.
- Reporting integrity: a certified metric needs one definition in a semantic layer with owner, freshness timestamp, filters and exclusions shown. Always state the denominator, time zone and comparison period.
- Accessibility: never rely on red/green alone, keep text contrast high, and provide data tables or text summaries.
Worked example
A request arrives for "revenue by region for three years". Twelve regions on one line chart produce spaghetti. Fix: plot the top five regions as lines, aggregate the rest into "Other", index each series to first month = 100 if the question is growth rate rather than absolute scale, label lines at the right edge, and footnote the metric definition (net of refunds, local currency converted at month-end rates).
Traps
Twelve-slice pies; truncated bar axes that exaggerate a gap; 3D and shadow effects; rainbow palettes for ordered data; a dashboard with no owner or freshness timestamp; silently changing a metric definition so the trend line breaks.