Recurrence is one of the most costly problems in industrial quality. And one of the least well measured. This is because non-conformances are detected one by one, but we miss the pattern that connects them.

What exactly is a recurrence?

A recurrence is the repeated appearance of a non-conformance of the same nature, on the same process, the same product or the same component, after a corrective action has been implemented.

The distinction with simple repetition is important. A repetition is the same defect returning because it has not been addressed. A recurrence is the same defect returning despite being addressed. This signals that the corrective action did not address the true root cause.

There are three levels of recurrence in industry. Intra-site recurrence: same defect, same line, same shift. Inter-site recurrence: same defect reappearing on different sites, without one knowing that the other has already resolved it. And inter-programme recurrence: same root cause identified on different products or references. This last level is the most difficult to detect, and often the most costly.

Why recurrence is so difficult to detect

The first reason is structural: quality data is scattered. A closed non-conformance in one tool, a corrective action in another, feedback from a customer in a CRM. No one crosses the information. And when no one crosses it, no one sees it.

The second reason is linguistic. Two similar non-conformances can be described with completely different terms depending on the operator or the site. A keyword search is not enough to establish the link. Without semantic processing, the pattern remains invisible.

The third reason is organisational. Shift handovers and team changes erase memory. A new engineer arriving on a case has no way of knowing that an identical case was already resolved 18 months earlier on another site. And under production pressure, cases are closed without the effectiveness of the corrective action being genuinely verified over time.

What recurrence really costs

The direct cost is visible: analyses redone from scratch, expertises remobilised, repeated reworks, long investigation times. But it is the indirect cost that weighs most heavily on performance.

Repeated customer complaints about the same defect weaken a business relationship that took years to build. Contractual penalties accumulate. And engineering teams, mobilised in loops on the same problems, no longer have the time or the bandwidth to address new ones.

According to studies on the cost of non-quality, between 5% and 30% of industrial turnover is consumed by avoidable costs. A significant portion is directly linked to undetected recurrence.

In sectors such as aerospace, nuclear energy or automotive, the stakes are even higher. An unidentified recurrence can trigger a customer audit, call into question supplier status, or compromise the organisation's liability on a safety incident.

How to detect and prevent recurrence

The first lever is the quality of the input data. A non-conformance well described from its creation is a non-conformance that can be cross-referenced with history. If data entry is incomplete or inconsistent across operators, pattern detection becomes impossible.

The second lever is monitoring the effectiveness of corrective action plans. Checking at 30, 60 and 90 days whether the corrective action has genuinely eliminated the root cause. Not to fill in a form, but to ensure that the system has changed, not just that the file is closed.

The third lever is the centralisation of quality data into a single database, accessible to all teams across all sites. This is the absolute prerequisite for detecting inter-site and inter-programme patterns.

This is where artificial intelligence changes the game. Where manual search fails, AI can automatically cross-reference new events with the complete history, including text, images and technical data, and signal in real-time that a similar case has already been resolved. The engineer no longer searches. He decides.

Recurrence as a maturity indicator

The recurrence rate is one of the most revealing indicators of the maturity of a quality system. The higher it is, the more the organisation operates in a reactive rather than preventive mode.

Recurrence is also a managerial signal. If the same root causes keep returning, it is often because corrective actions are designed to close the file, not to change the system.

In the current context of ramping up production rates and increased regulatory pressure, the ability to avoid repeating errors becomes a structural competitive advantage. The industrial organisations that succeed are not those with the fewest non-conformances. They are those that best capitalise on the ones they have had.

Eliminating recurrence means transforming every incident into a learning lever for the entire organisation. It means moving from a quality department that suffers to a quality department that learns.