Compare multi-metric profiles across materials, athletes, models, or products in one figure.
Upload wide or long tables, and SciFigure computes per-axis ranges, decides whether axes need normalization, and generates a clean radar chart with the raw values kept on record.
Upload data and describe the radar chart you need.
Costs 5 credits. Axis ranges, scaling, and plotted positions are computed from your uploaded data, then drawn as a PNG.
Preview
Aspect ratio: 16:9
These are sample previews. Upload your dataset to generate your own radar chart.
Use these examples as starting points for material selection, sports science testing, machine learning benchmarks, and sensory panel profiles.
Prompt
Based on the uploaded alloy_properties.csv file, create a radar chart comparing Al 6061-T6, Ti-6Al-4V, 316L stainless steel, and a CFRP laminate on six mechanical and thermal properties. Normalize each axis 0-1, treat density as lower is better, and show raw ranges on the axis labels.
Prompt
Based on the uploaded athlete_testing.xlsx file, create a spider chart of five athletes across VO2max, countermovement jump, 30 m sprint, back squat, Yo-Yo test distance, and HRV. Normalize axes, treat sprint time as lower is better, and use a compact legend.
Prompt
Based on the uploaded model_benchmarks.csv file, create a radar chart of four classifiers on accuracy, precision, recall, F1, AUROC, and calibration. Keep the raw 0-1 values on the same scale, use outline-only polygons, and print values for the best model.
Prompt
Based on the uploaded apple_sensory_panel.tsv file, create a spider web chart of mean panel scores for four apple cultivars on sweetness, acidity, crispness, juiciness, aroma, and bitterness. Use the raw 0-10 intensity scale with ring labels.
Prompt
Based on the uploaded battery_cell_benchmark.csv file, create a radar chart of five lithium cell chemistries across specific energy, energy density, cycle life, fast-charge rate, cost per kWh, and self-discharge. Scale each axis as percent of the best value so zero stays at the center, lower is better: cost per kWh and self-discharge, use outline-only polygons with vertex markers, and print the raw best value on each axis label.
Prompt
Based on the uploaded qol_timepoints.tsv file, which has Timepoint, Scale, and MeanScore columns, create a spider chart of mean EORTC QLQ-C30 scores for one cohort at baseline, 6 months, and 12 months. Keep the raw 0-100 scale with ring labels every 20, reverse the fatigue, pain, and nausea symptom scales so better outcomes plot outward, shade a light gray reference band at the population norm, and label each series with its n.
Add CSV, TSV, TXT, or Excel data. Use a wide table with a series name column and one numeric column per axis, or a long table with series, metric, and value columns.
Say whether axes should be normalized or kept on a raw shared scale, which metrics are lower-is-better, and how the outlines, fills, and legend should look.
SciFigure computes axis ranges and plotted positions, draws the spider chart, and lets you download a PNG for papers, theses, and slides.
It draws each series, such as a material, athlete, model, or product, as a closed polygon across three or more axes that radiate from a common center. The shape makes strengths, weaknesses, and trade-offs visible at a glance, which is why radar graphs are common in materials selection, sports science testing, machine learning benchmark summaries, and sensory analysis. This radar chart maker averages repeated rows, computes the minimum and maximum of every axis, decides whether axes need normalization, and calculates each plotted position before the figure is drawn. The image model only handles layout and styling.
Radar charts work best for a small number of series measured on the same set of 3 to 12 metrics, where the overall profile matters more than exact values. For precise comparisons of individual metrics, a grouped bar chart or dot plot is easier to read, and for many samples across many features a heatmap or PCA plot scales better. A spider web chart is a strong summary figure, but pair it with a table of raw values in the paper or supplement.
Two layouts work. Wide: one row per series, a text column with the series name, and one numeric column per axis. Long: a series column, an axis or metric name column, and a numeric value column. Rows that share a series name are averaged.
The tool calculates them from your uploaded values: series means, per-axis minimum and maximum, the scale ratio, normalization, and each plotted position. Those numbers are passed to the image model, which draws the outlines and labels without computing anything itself.
When axis magnitudes differ by more than a factor of 5, raw values would squash small-unit metrics near the center, so each axis is min-max scaled to 0-1. Write "raw" or "same scale" in your instructions to turn this off.
Yes. Write "lower is better:" followed by the axis names, for example sprint time, latency, density, or cost. Those axes are reversed after normalization so better values plot farther from the center.
Yes. Radar chart, spider chart, spider web chart, star plot, and radar graph all describe the same figure: values on spokes around a circle joined into polygons.
Yes. Verify the series names, axis order, normalization method, and raw ranges before submission, and include the raw values in a table or supplement because normalized radar positions are relative.
Generate radar charts, PCA plots, heatmaps, and other research-ready visualizations with SciFigure AI.
