What it means
Zero wrongly read labels let through unnoticed. That is the claim, and nothing more. Our vision system reads every label with neural networks trained on your own labels and products, checked by multiple independent engines. Only when engines agree, or a second shot confirms, does a read count as certain. Anything uncertain goes to a human: a case of doubt is not a misread, it's a correctly held read.
How it's measured
Validated on real production runs: every camera read compared afterwards with what was actually on the pallet. Result on those runs: zero wrong labels let through; a share of labels went to the human review queue. We discuss both numbers openly in every conversation.
definition: misread = read wrongly and passed as certain · held for review ≠ misread · read reliability in practice: 99.99%+