Radar Chart Maker & Generator
Make radar charts that profile up to six subjects on 3 to 12 measures: sensory panels, model benchmarks, fitness tests, questionnaire domains. For researchers and students who need every point placed from their own scores.
Radar chart styles
Five styles, each drawn on a different study and scale: wheat nutrients, mouse behaviour z-scores, PhD skill self-ratings, river water quality and battery chemistries. Click a card to use its style, or take its data as a starting point.
Radar chart examples
Real renders of four radar chart prompts: three espresso blends on a 0 to 10 sensory scale, four classifiers on six evaluation metrics, one athlete against the squad average with a range on each axis, and quality-of-life domains before and after cardiac rehabilitation. Click one to load its prompt and scores into the box.
What is a radar chart?
A radar chart plots several measurements of one subject on spokes that start from a common centre, then joins the points into a closed outline, so each subject becomes a shape. It is also called a spider chart, spider graph, web chart or radar graph, and in statistics a star plot. Profiles with similar shapes have similar strengths and weaknesses, which is what the chart is for: comparing a few subjects across the same set of measures.
About this radar chart maker
- Type the axes after Axes: and one line of scores per subject, paste a table from Excel, or attach a CSV, TSV or .xlsx file.
- Scores typed in sentences are read either way round. An attached table with the measures down the side is turned when you write “rows are the axes”, or when its shape allows no other reading (more than six rows, or fewer than three columns).
- Spoke angles, ring values and the position of every point along its spoke are computed in code before the image is drawn.
- Axes measured in different units each get their own range, printed under the axis name; say “same scale” to keep one shared scale.
- Up to six series and 3 to 12 axes on one chart; the Small multiples style gives each series a panel of its own.
Radar charts researchers and students make
Six figures this spider chart maker is asked for most, what sets each apart, and what to write so it comes out right.
Comparison radar chart of models or products
Two to six series on the same metrics, one outline each, read for which one dominates where. Classifier accuracy, precision and recall; battery energy density, cycle life and cost. Write the shared scale (0 to 1, 1 to 10) so every series is measured against the same rings.
Sensory profile spider chart
A trained panel's mean intensity scores for each attribute of a food or drink, one polygon per sample. Bitterness, acidity, body, aroma: the form used in sensory science and product development. Panels usually score on a fixed scale, so state it (“on a 0–10 scale”) and the rings run 0, 2, 4, 6, 8, 10.
Single profile against a reference
One subject drawn over a norm, target or group average, the reference as a dashed outline with no fill. An athlete against the squad average, a patient against normative values. Name the reference series Reference, Average, Norm, Target or Benchmark and it is drawn that way automatically.
Radar chart with different scales per axis
Axes in different units (kg, cm, mL/kg/min) each run from their own minimum to their own maximum, with that range printed under the axis label. Without it the axis with the biggest numbers swamps the rest. It switches on by itself when the axes differ about fivefold or more, or when you write “normalized” or “each axis on its own scale”; “min–max” puts each axis's lowest value at the centre.
Questionnaire domain scores
Domain or subscale scores of an instrument, before and after an intervention or across groups. Quality-of-life, burnout or competency questionnaires report several domains on one 0 to 100 or Likert scale. Write the scale; a 1 to 5 scale is drawn from 0 with a ring at every point, so a rating of 1 stays visible instead of vanishing into the centre.
Small multiples instead of one crowded chart
One small radar per subject in a grid, every panel on the same spokes and rings. Past three overlapping outlines readers lose track of which line is which. Pick the Small multiples style; a reference series, if there is one, is repeated in every panel instead of taking its own.
Where each point goes on a radar chart
Each value becomes a distance from the centre along its own spoke: the fraction of the way from the axis minimum to the axis maximum. The spoke's angle comes from its position in the list, measured clockwise from 12 o'clock.
r = (v − min) / (max − min), θᵢ = 360° × i / n
In words
With n axes, spoke i (counting from 0) points 360/n degrees further clockwise than the one before. A value v sits at fraction r of the radius on that spoke, where min is the value at the centre and max the value at the outer ring: the shared scale's ends, or that axis's own range. A score of 7 on a 0 to 10 scale sits 70% of the way out.
The area of the outline is ½ sin(360°/n) multiplied by the sum of rᵢ × rᵢ₊₁ over neighbouring spokes. It depends on which axes happen to sit next to each other, and doubling every score quadruples it, so this maker never reports an area and does not let the AI print one.
Choosing the scale for a spider chart
Where the centre sits and how far the outer ring reaches decide what the shapes say.
| Scale | Centre and outer ring | Use it when | How to ask for it |
|---|---|---|---|
| Shared, from zero | 0 to a round number above the largest score | Every axis is the same kind of score in the same unit | Default when the axes are of similar size |
| Shared, stated | The top of the scale you name; a rating scale starting at 1 is drawn from 0 | Ratings on a fixed scale: 1 to 5, 0 to 10, 0 to 100 | “on a 1–5 scale”, “out of 100” |
| Per axis, from zero | 0 to a round number above each axis's largest value | Axes in different units or of very different size | “normalized”, “each axis on its own scale”, or automatic at about fivefold spread |
| Per axis, min–max | Each axis's smallest to largest value | Only the ranking within each axis matters | “min–max normalized” |
A centre above zero exaggerates differences, so the per-axis range is always printed under the axis name and the legend or caption should say the axes were rescaled.
Radar chart vs grouped bar chart vs heatmap
A radar chart shows the overall shape of a few profiles; a grouped bar chart is better for reading exact differences; a heatmap is better once there are many subjects.
| Question | Radar chart | Grouped bar chart | Heatmap |
|---|---|---|---|
| Best number of subjects | 1 to 3 on one chart, up to 6 as small multiples | 2 to 4 series | Dozens of rows |
| Comparing one value between subjects | Hard: points sit on different spokes | Easy: bars share one baseline | Approximate: colour steps |
| Seeing a whole profile at once | Yes, as a shape | Spread across clusters | As a row of colours |
| Affected by axis order | Yes, the shape changes | Only the reading order | Only the reading order |
| Axes in different units | Possible, with a range per axis | Needs separate panels | Needs scaling, such as z-scores |
If the point of the figure is one exact difference, draw it with the bar graph maker; for many subjects or a matrix of scores, a heatmap with one row per subject.
Radar chart questions, answered
How do I make a radar chart from my data?
Write the axes, then one line of scores per subject, and generate. For example: Axes: Bitterness, Acidity, Body. Blend A: 7.2, 4.1, 6.8. Blend B: 5.4, 6.6, 5.1. You can also paste the table from Excel or attach a CSV or .xlsx file with names in the first column. Add the scale if your scores have one, and pick a style.
Is a spider chart the same as a radar chart?
Yes. Spider chart, spider graph, web chart, radar graph and star plot all name the same figure: spokes from a common centre, one per variable, with each subject's values joined into a closed outline. Star plot is the statisticians' name: the NIST Engineering Statistics Handbook uses it, and typically draws one small star per observation.
How do I read a radar chart?
Read each spoke as its own axis: the further a point sits from the centre, the higher that score. Then compare outlines: a series bulging toward a spoke is strong on that measure. Check the scale before comparing spokes, because on a per-axis chart the same distance means different amounts on each axis.
Why does this radar chart maker not report the area of each polygon?
Because the area is not a property of the data. It is built from products of neighbouring values, so listing the same scores in a different axis order changes it, and it grows with the square of the scores: doubling every value makes the outline four times larger. Compare values spoke by spoke, or sum the scores if a total is what you need.
Does the order of the axes matter?
It changes the shape, not the values. Two strong measures placed next to each other make a broad lobe; placed apart they make two spikes. Put related measures side by side (all the physical tests, then all the cognitive ones) and keep the same order in every figure of the paper. The chart follows the order you type, starting at 12 o'clock and going clockwise.
When is a grouped bar chart or a heatmap better than a radar chart?
When readers must compare exact values, or when there are many subjects. Values on different spokes do not share a baseline, so a difference of 0.05 between two models is easier to judge in a grouped bar chart. With more than six subjects the outlines become unreadable, and a heatmap with subjects as rows shows the same table without the overlap.
Can each axis of a spider graph have its own scale?
Yes. Each axis then runs from its own minimum at the centre to its own maximum at the outer ring, and that range is printed under its label. This is needed when the axes are in different units, for example jump height in cm next to squat load in kg. It turns on automatically when the axes differ about fivefold or more; write “same scale” to keep one shared scale.
How is a radar chart in Excel different from this one?
Excel's radar chart draws every axis on one shared value scale, so axes in different units need rescaling in the sheet first. Here each axis can keep its own printed range, every point is placed from your numbers by code, a table with the measures in rows is turned for you, and the styles (outline, grayscale print, small multiples) are set for a figure panel. You can still attach the same .xlsx file.
Does the AI change or estimate my scores?
No. The scores, the ring values, every spoke angle and the distance of every point from the centre are calculated in code and given to the image model as fixed instructions. The model draws them and adds no number of its own. If your text holds no scores, nothing is calculated and the model works from your description.
How many axes and series can one radar chart have?
3 to 12 axes and up to 6 series. Below three spokes there is no outline to draw, past twelve the labels crowd each other, and past six overlapping outlines no reader can follow one series around the chart. For more subjects, use small multiples across several figures or a heatmap.
What should the figure legend of a radar chart say?
State what each axis measures and its unit, the scale or the per-axis ranges, what each series is (and n, if the values are means), and the source of the scores. If the axes were rescaled, say how: for example “each axis scaled from 0 to its maximum”. A reader cannot recover any of this from the shape alone.
Sources
- Star Plot (Engineering Statistics Handbook, section 1.3.3.29) — NIST/SEMATECH
- The Radar chart and its caveats — Data to Viz
Keep exploring
Our other diagram tools, and the guides behind them.
Draw your radar chart
Type the axes and the scores, choose a style, and get every point placed where your numbers put it.
Start drawing