Preparing the published article.
Repeatability and Absolute Accuracy: What Each One Tells You
One describes returning to the same point, the other describes arriving at a commanded point.
Cover image is an AI-generated illustration of a robotics application, not a photograph of the equipment described.
Positioning figures are frequently mixed up. Repeatability and absolute accuracy answer two different questions, and using the wrong one produces expectations that do not match real behaviour.
Repeatability answers "can it come back"
Repeatability describes how tightly the robot returns to the same taught point over many cycles. For tasks that are taught once and then repeated, this is the figure that matters. A small value means the motion is consistent, but it says nothing about whether the robot can reach a coordinate taken from a drawing on the first attempt.
Absolute accuracy answers "does it arrive where told"
Absolute accuracy describes how closely the robot reaches a commanded position in space. It matters when work is driven by offline programs or by coordinates produced by a vision system. Absolute accuracy is normally lower than repeatability and is influenced by calibration method, payload and temperature. Because it can drift over time, it is also a maintenance question rather than a one-off purchase figure.
Measurement method decides comparability
Test conditions, measuring instruments and the number of sample points all affect the reported values. Before comparing figures from different sources, confirm that the test method is the same. Accuracy values published without conditions can only be treated as orders of magnitude.
Decide from the process
Which capability is needed depends on whether the process relies on teaching or on coordinates. Where fixed positions are repeated, consistency is the focus. Where variants change frequently and new coordinates are used, absolute accuracy and the effort of maintaining calibration also matter.
This article explains concepts and typical applications. It does not give universal thresholds; the required values follow from the process tolerance.