Updated October 2026
The axes are the entire structure of a radar chart — get them right and the resulting polygon tells an honest story; get them wrong and even accurate numbers can produce a misleading shape. This page covers both the mechanics (how this tool actually places and scales axes) and the judgment calls (which criteria to use, and in what order).
For N total axes, axis i (counting from 0) sits at angle 2π × i / N, measured clockwise starting from straight up — so axis 0 is always at 12 o'clock, and the rest walk clockwise evenly around the circle. A value on that axis becomes an actual pixel position using x = centerX + r·sin(angle) and y = centerY − r·cos(angle), where r is that value's radius — its distance from center, scaled linearly against the chart's max (auto-detected from your highest entered value, or a manual max you set). A value of zero sits exactly at the center; the maximum value sits exactly on the outer ring.
Every axis should be measuring something on a genuinely comparable scale to every other axis — that's the single most important rule, and the one most radar charts get wrong. "Speed" and "Strength" both scored 1–10 are comparable; "Revenue growth %" and "Customer satisfaction score" are not, even if you force them onto the same 0–10 range, because a forced rescale hides how different the underlying meaning of each axis is. If your criteria genuinely aren't on a comparable scale, normalize them deliberately and say so (for example, "percentile rank among competitors" as a shared 0–100 scale across otherwise-unlike metrics), rather than silently plotting raw numbers of very different kinds side by side.
Three is the practical minimum (fewer doesn't form a polygon). Four to eight is the sweet spot for readability — enough to show real structure, few enough that labels don't crowd. Beyond about 10, axis labels start overlapping or need to shrink to the point of being hard to read, and the polygon itself gets visually noisy. See how many axes is too many for the full breakdown.
Visually, yes, even though the underlying data doesn't change meaning. Two practical conventions: group related axes next to each other (so a viewer's eye naturally reads related criteria as a cluster rather than scattered around the circle), or order axes by importance/priority if there's a natural ranking, so the most important criteria sit near the top where they're first noticed. There's no single correct order — but a deliberate order reads better than a random one, and reordering axes in this tool is as simple as re-entering rows in your preferred sequence.
Auto-scale (the default) sets the outer ring to your highest entered value across all datasets and axes, which keeps the chart filling the available space regardless of what numbers you're plotting. Turn it off and set a manual max when you want a fixed, meaningful ceiling — for example "10 = perfect score" on a 1–10 skill scale, or when building two separate charts that need to share an identical scale for a fair side-by-side visual comparison (auto-scale alone can't guarantee that, since two different datasets can have two different highest values).
This tool clamps plotted values to the 0–scale range, since the radius math is built around a shared zero-centered scale — a negative value has no natural position on a radar chart's radius. If your data legitimately includes negative numbers (a net score, a delta from baseline), shift every value by a constant offset before entering them so everything becomes non-negative while the relative differences between values stay identical.