The hidden cost of a single missed defect — and how AI changes the math

On a circular knitting machine running at 28 rpm, a single missed defect rarely stays single. By the time the operator catches it on the inspection table, the same flaw is woven into the next 60 to 90 metres of fabric. That is two to three hours of production that has to be unwound, re-tensioned, and re-knitted — assuming the yarn survives the process at all.

When we sat down with operations leaders at our first ten customers and added it up, the picture was uncomfortable. The average mill we visited was losing between 4 and 7 percent of finished metres to defects that an attentive operator could, in principle, have caught in time. Almost none of those mills considered themselves to have a quality problem.

What real-time inspection changes is not the operator's attention — it is the time horizon over which a defect can grow. Profna Vision flags a stitching anomaly within 40 milliseconds, before the next revolution lays another row of yarn over it. The same defect that used to cost three hours of unwinding now costs less than one revolution of fabric.

Multiply that across 22 machines and three shifts, and the unit economics flip. The mills we work with are not chasing 100 percent quality; they are chasing the difference between catching a defect in 0.04 seconds and catching it in 30. That difference is where the entire business case lives.