Preventive or corrective: how to decide.

Ask it per failure mode, not per asset

« Should this pump be on preventive maintenance? » has no answer. The same pump fails by a worn seal, a seized bearing, a burnt winding and a blocked impeller, and those four deserve four different decisions. A plan built asset by asset ends up greasing things that never fail and ignoring the one part that always does.

When prevention pays

• The failure comes with age. Wear, fatigue, corrosion, drying seals. There is a before and an after, so acting before is possible.
• A task actually changes the odds. Greasing a bearing reduces its chance of seizing. Opening a healthy motor does not reduce anything, and introduces its own risk.
• The consequence is worse than the task. Four hours of planned shutdown against two days of emergency, a tanker and a boil notice.

The three have to hold together. Two out of three is how a maintenance plan grows heavy without getting safer.

When it is waste

Many failures are not age-related. Electronics, instrumentation and a good share of rotating equipment fail at a rate that barely changes with service hours. For those, a calendar task does not lower the risk — it consumes the hours of the people who would otherwise be watching the assets that do wear out. Worse, opening a working machine on schedule is itself a cause of failure: it is called infant mortality, and it is the reason a plan can increase breakdowns.

Running to failure is a decision

For a cheap, redundant, quickly replaced component, waiting for the failure is the right answer. What separates a decision from an omission is that the decision is written down, with its reason, and the spare is on the shelf. An omission is discovered at two in the morning.

Three cases never qualify: a failure that hurts somebody, a failure that reaches the environment, and a failure nobody would notice until it matters — which is the ordinary condition of a standby.

The number that tells you if the plan works

Not the share of preventive work. A plan can be ninety per cent preventive and still be wrong. Count instead the corrective jobs that were predictable — those whose failure mode was already in the plan, with a task that should have caught it. That number points at the tasks to change, and it is the only one that improves when the plan improves.