Corrective maintenance (CM) restores an item after it has failed. The defining line is fault recognition (EN 13306:2017, 7.9): work carried out after a fault has been recognised is corrective; work done before it — even if triggered by a rumbling bearing or a rising temperature — is preventive.
CM splits into immediate (breakdown) corrective — drop everything, the machine is down — and deferred corrective — the failure happened but you can schedule the repair. Deferred is far cheaper.
Reactive CM isn't always wrong: for cheap, non-critical, fail-safe items, run-to-failure is the rational choice. It's only a problem when it's the default for things that matter.
The distinction that matters
The words "corrective" and "preventive" get muddled constantly, usually because people anchor on how planned the work was rather than when it happened. Here is the clean rule, and it's worth memorising:
If the work is intended to assess or mitigate degradation and reduce the item’s probability of failure, it is preventive (EN 13306:2017, 7.1) — even if you only acted because a bearing started to rumble or a temperature rose. Note the criterion is the item, not the production plan: a job that protects output but does nothing about degradation is not preventive maintenance under the standard. If you refit it before it breaks down, that's preventive. Corrective maintenance is the work done after the failure has occurred, to correct it. The dividing line is the failure event, not how much warning you had.
This matters because it changes how you classify — and therefore measure — most of your work. A condition-triggered bearing change done while the item still performs all of its required functions is preventive. But once it has lost a required function — partially counts, EN 13306:2017 calls that a partial fault (6.3) — a fault has been recognised, and the same job is corrective (7.9) even though the machine is still turning. Only once the bearing has actually seized, and you're repairing the consequences, is it corrective. Get this right and your reactive-vs-proactive metrics finally mean something.
Two kinds of corrective
- Immediate corrective (EN 13306:2017, 7.11) — carried out without delay after a fault has been detected, to avoid unacceptable consequences. Unplanned, disruptive, expensive: no time to plan, parts may not be on hand, and you pay overtime and lost production. This is "reactive maintenance" in the pejorative sense.
- Deferred corrective (EN 13306:2017, 7.10) — not carried out immediately after fault detection, but delayed in accordance with given rules, so the repair lands in a convenient window. Far cheaper, because it's planned — parts staged, labour scheduled, no production hit.
The whole art of running corrective work well is converting immediate into deferred wherever possible — through redundancy, standby units, and fail-safe design — so that even after-failure work can be planned rather than panicked.
The true cost of breakdown
Unplanned corrective work is the most expensive way to maintain anything, and the repair bill is only the visible part:
| Cost | Planned (deferred / PM) | Unplanned (breakdown CM) |
|---|---|---|
| Labour | Scheduled, normal hours | Overtime, call-outs, idle waiting |
| Parts | Staged, in stock | Expedited freight, premium price |
| Production | Planned outage / standby | Lost output, missed deliveries |
| Secondary damage | Avoided | Collateral — seized shaft, wrecked rotor |
| Safety / environment | Controlled | Higher risk under pressure |
Rules of thumb in the industry put unplanned work at 3–5× the cost of the same job planned. That multiple is the entire economic case for the other two strategies.
When run-to-failure is right
Reactive maintenance gets a bad name, but deliberately choosing to run an item to failure is a perfectly valid strategy — when the maths supports it. Run-to-failure makes sense when:
- The item is cheap and quick to replace (a light bulb, a small inline filter).
- Failure has no safety, environmental or major production consequence (it's redundant or fail-safe).
- There is no effective preventive task — failure is random, so scheduled work wouldn't help (a key RCM finding).
- The cost of preventing failure exceeds the cost of the failure itself.
Deciding which items those are is exactly what criticality and RCM are for. Run-to-failure should be a chosen strategy for low-consequence items — never the accidental default for everything.
The goal isn't zero corrective — it's zero surprise. A mature program still does corrective work, but most of it is deferred and planned, and the unplanned breakdowns are confined to items where that's the deliberate, economical choice. You get there with preventive and predictive maintenance.
Key takeaways
- Corrective = after fault recognition. Work intended to reduce the item’s probability of failure is preventive — even when condition data triggered it (EN 13306:2017, 7.1 and 7.9).
- Deferred beats immediate. Convert breakdown work into planned work with redundancy and fail-safe design.
- Unplanned costs 3–5× planned — the economic case for PM and PdM.
- Run-to-failure is valid by choice for cheap, low-consequence items — decided by criticality/RCM, not by neglect.