How ratings & comparisons work

1 · Comparison bars (within a comparison)

On a comparison, each bar is relative to the robots you selected. For a spec, we take the smallest and largest value across your set and place each robot between them:

fill% = 10 + 90 × (value − min) / (max − min)

Direction-aware: for payload, reach and speed higher is better; for repeatability, weight, noise and price lower is better, so the bar is mirrored. A 10% floor keeps the weakest value visible. It answers one question: “of these robots, which is strongest on this spec?”

2 · Winners (the ✓)

The check mark marks the genuinely best published value in that row — the maximum for higher-is-better specs, the minimum for lower-is-better. Neutral specs (axis count, robot mass) are never crowned. The header tally (“wins 4 / 6”) simply counts decided rows.

3 · Class percentile bars (on a product page)

On a single robot’s page the bar shows where it ranks among all robots of the same type for that spec — e.g. “80th percentile in class” means it out-specs 80% of cobots on payload. Direction-aware, same as above.

4 · Data completeness

The completeness meter is not a quality score. It is the weighted share of spec fields we have filled from datasheets, with core performance specs weighted heaviest. Unknown values are shown as “—”, never guessed and never counted as zero.

Sources

Specifications are taken from public manufacturer datasheets and product pages; each robot links to its source. Spotted an error? Robots carry a verification flag and are continuously refreshed from source datasheets.