There is no reliable universal mapping from a remote “small/medium/large” focus command to a precise shift in focus distance.

From the answers, step size can vary because many lenses do not provide accurate position feedback, so the camera is not commanding a known physical distance. The actual result may differ by lens design, camera body, individual sample, and even operating conditions. One reported factor is battery level, which can change the apparent number or size of steps. Some lenses are more consistent than others, but that does not make them predictable enough to generalize broadly.

So yes: if you want a preset distance estimate, you would effectively need to characterize each specific body-and-lens combination, and even then repeatability may still be poor.

The more practical approach is not to infer focus-plane movement from step commands at all. Instead, use contrast-detection autofocus or another feedback-based method that evaluates image sharpness and moves focus iteratively from coarse to fine until contrast stops improving. If you need subject distance, estimate it separately from scene geometry or known object size rather than from focus motor steps.