Key takeaways
  • Reactive means fixing on failure, preventive means fixing on a schedule, predictive means fixing on evidence of impending failure.
  • The right answer differs per asset. Reactive is correct for some assets and negligent for others.
  • Preventive maintenance on an asset with random failure modes adds cost and can add failures through intervention.
The uncomfortable truth about PM
More preventive maintenance is not always better

For failure modes that are random rather than age-related, scheduled intervention does not reduce failures and introduces its own risk through disassembly, contamination and error. Choosing preventive everywhere is a policy, not a decision.

The three strategies

  • Reactive. Run to failure and repair. Lowest planning cost, highest failure cost, no warning.
  • Preventive. Intervene at fixed intervals of time, cycles or output. Predictable cost, some waste, effective only where failure correlates with age or usage.
  • Predictive. Monitor condition and intervene when evidence indicates deterioration. Highest setup cost, best avoidance of both failure and unnecessary work.

Choosing per asset

  • What happens if it fails? Safety, environmental, production, quality and cost consequences.
  • Is the failure mode age-related or random? Wear, corrosion and fatigue are age-related; electronics and many bearings fail more randomly.
  • Is there a detectable warning of impending failure, and how long is the interval between that warning and failure?
  • What does intervention cost, including the downtime to perform it?
  • Is a spare available and how long to obtain one?

When reactive is the right answer

For a cheap, non-critical, redundant asset with no safety consequence and a spare on the shelf, running to failure is the economically correct choice. Writing that down as a deliberate strategy is very different from having no strategy, and it frees maintenance capacity for assets that need it.

Making preventive work

Base intervals on evidence rather than on the manual's default, which is written conservatively for every customer. Track whether each PM finds anything: a task that has found nothing in three years is either the wrong task or the wrong interval, and reviewing PM findings is how the schedule improves.

Starting predictive without a big budget

Vibration pens, infrared thermometers, ultrasonic leak detectors and oil sampling are affordable and cover a large share of common failure modes. The constraint is usually not the instrument but the discipline of taking readings consistently and trending them, because a single reading tells you almost nothing.