A drop in declared non-conformities is rarely good news. Most often, the deviations are still there: they just no longer pass through the quality system.

Fewer NCs, really?

The number of declared non-conformities has been falling for three months. In the management review, this is celebrated. On the shop floor, no one has the impression that production has become more stable.

Both observations are true. The indicator measures what is declared, not what happens. In between, there is everything that the shop floor handles on its own, without leaving a trace in the quality system.

This grey area has a name: under-reporting. It almost never comes from a lack of rigour on the part of the teams. It comes from the way the reporting is designed.

What the indicator does not see

A part slightly out of tolerance, reworked at the workstation. A batch set aside "just to see". An adjustment corrected at the start of a shift, without anyone knowing why it had drifted.

None of these deviations are hidden. Everyone in the workshop knows about them. They are simply handled alongside the system, by competent people who have judged that reporting was not worth the effort.

It is this judgement that needs to be understood. As long as it remains rational from the operator's point of view, no memo will change it.

Five reasons why the shop floor does not report

Reporting takes more time than correcting

Reworking a part takes two minutes. Reporting it requires leaving the station, opening a form, finding the reference, choosing a category. When the pace is tight, the calculation is quickly made.

The form is designed for the quality department

Defect codes, cause families, mandatory fields: most forms serve downstream analysis. The operator, however, knows how to describe what they see. They do not always know which box to put it in.

Reporting feels like telling on oneself

If a declared NC leads to the question "who did this?", the message is received. Reporting becomes a personal risk, and silence a protection.

Nothing ever comes back

The operator reports, then hears nothing more. They do not know if their report was useful, nor what was decided. After a few times, they stop.

The indicator rewards silence

When the stated objective is "fewer NCs", every report degrades the team's result. The system demands transparency and rewards the opposite.

What we lose: the signal

An isolated NC says very little. Ten similar NCs over three months, on two different workstations, tell a story: a tool drifting, a material that has changed, an instruction misunderstood.

This story only exists if all ten deviations have been reported. If eight were reworked at the station, the analysis covers only two cases, and the root cause remains out of reach.

Under-reporting therefore does not just make a figure go down. It deprives the quality department of its raw material: repetition. The problem eventually surfaces, but later, bigger, and often at the customer's end.

Getting deviations reported

Each reason for not reporting calls for a specific lever. None of them rely on discipline.

Reduce the cost of reporting

Reporting must take just a few actions, at the workstation. The operator describes what they see, in their own words. Classification, coding, and linking to existing cases are the job of the quality department, not the shop floor.

Separate reporting from responsibility

A declared NC triggers a process analysis, not a witch hunt. This rule must be stated, then kept, even on the day the deviation costs a lot.

Close the loop

Every report deserves feedback: what was found, what has changed. A display in the workshop or two minutes at the start of a shift is enough. This is often the most effective, and most neglected, lever.

Change what is measured

Track the number of declared NCs as a sign of vigilance, not as a result to be minimised. The true performance indicator is recurrence: do the same deviations keep coming back?

What Atom changes in the equation

The more the shop floor reports, the more the quality department has to process. Without a tool, encouraging reporting is like creating a queue, and a queue discourages reporters.

Atom, Yxir's AI agent, handles part of this processing: reporting non-conformities, identifying similar cases, analysing root causes, and corrective actions. Quality teams save an average of 50% of their time on NC processing.

This returned time is used for exactly what brings reports back: responding quickly, and showing the shop floor what their report helped to find.

A report is information, not a confession

The shop floor sees deviations before anyone else. The question is not whether they spot them, but whether the system gives them a reason to report them.

A simple report, without looking for a culprit, followed by visible feedback: under these three conditions, the NC indicator starts telling the truth again. Even if, at first, it goes up.