A radar chart (also spider, web, polar, or star chart) maps multiple quantitative variables onto equally-spaced radial axes, all sharing a single origin and scale. A data series is drawn as a polygon connecting its value on each axis. The perceptual mechanism is polygon shape comparison: the viewer compares overall outlines holistically, noticing which polygon is larger (higher average performance), where one extends further than another on a specific axis (strength in that dimension), and whether shapes are pointy (uneven profile) or rounded (balanced across dimensions). This holistic shape recognition is the chart's primary strength — and it requires no arithmetic from the viewer.
The data contains three series across six dimensions where the overall performance profile — not any single variable — is the message. South Asia has the most area in the upper-right quadrant (access, capacity); MENA is skewed toward the upper-left (funding, coordination). These profile shapes are immediately visible in the polygon outlines and would require six separate charts to convey with bar charts. The radar chart is appropriate here because: (a) all axes share the same 0–100 scale, making cross-axis comparison valid; (b) there are exactly 6 variables (within the 4–8 recommended range); and (c) there are only 3 series (within the 2–4 overlap-legibility limit).
The area of a radar polygon changes depending on the order of axes — placing two high-scoring variables adjacent produces a larger polygon area than placing them opposite each other, even with identical values. There is no canonical correct axis order. This implementation places axes starting at 12 o'clock and proceeds clockwise, but any rotation produces a legitimately different-looking polygon from the same data. Viewers who focus on area as the primary signal are vulnerable to this distortion. Always accompany a radar chart with the explicit values — this implementation provides them via the tooltip on hover.
Comparing values across different axes — "is MENA's funding (80) better than South Asia's access (75)?" — requires the viewer to mentally locate two points on two non-parallel scales and compare their radial distances from center. This is significantly harder than comparing two bars on a common baseline. The Catalogue states this clearly: "comparing values all on a single straight axis is much easier." Radar charts are for within-series profile reading and between-series shape comparison, not for precise cross-axis value comparisons. For the latter, use a parallel coordinates plot or small-multiple bar charts.
The FT Visual Vocabulary classifies radar charts under Patterns and Relationships: showing how multiple variables relate to each other within a single entity, or how multiple entities compare holistically. Abela's framework routes here when the question is "what is the overall profile of each series across multiple comparable dimensions?" Tufte's data-ink principle is partially in tension: the grid web, axis lines, and polygon fills all carry structural weight that isn't data — but they are necessary scaffolding for the radial reading. The fill at 14% opacity is the minimum necessary to visually anchor each polygon without obscuring the ones beneath. The stroke at full opacity carries the actual data boundary.
Series are rendered in reverse order: the first series in the
data file is drawn last, placing it on top of the SVG z-stack. This means the
series you care most about (typically the first-listed) is visually dominant —
its stroke is never covered by a subsequent polygon. The fill opacity of 14%
is set precisely so that three overlapping fills produce approximately
35–40% visible saturation in the overlap region, preserving the color
identity of each polygon without creating a false impression of a fourth
"combined" entity. Setting opacity to 0.30 or higher on three
overlapping fills would create near-opaque overlaps that the viewer would read
as distinct data — which they are not. The transparency is not an aesthetic
choice; it is a data-integrity choice.