Predictive vs preventive maintenance

In EN 13306, the European maintenance-terminology standard, predictive maintenance is a sub-type of preventive maintenance, not its opposite. The real dividing line is what triggers the work: a fixed interval, or a measured and forecast condition. Whether prediction is possible at all depends on the asset's P-F interval.

Key takeaways at a glance

Topic	Key point
The definition most articles get wrong	Almost every article frames these as rivals.
The three strategies in practice	Stripped of terminology, the difference is simply when you act.
Why fixed intervals waste work and still miss failures	A fixed interval assumes wear accumulates predictably with time or run hours.
The P-F interval decides whether prediction is possible	Prediction is not a property of the technology.
Where predictive maintenance fails	Honest limits, because a programme that ignores them turns into shelfware: No P-F window.
How to assign a strategy per asset	It is not either/or, and the assignment should be deliberate rather than inherited.
Standards worth reading	Standard What it gives you EN 13306 Maintenance terminology — the definitions used above ISO 17359 Condition monitoring and diagnostics of machines: g

The definition most articles get wrong

Almost every article frames these as rivals. The standard does not. EN 13306, the European standard for maintenance terminology, places predictive maintenance inside the preventive branch: preventive maintenance splits into predetermined maintenance (fixed intervals) and condition-based maintenance, and predictive maintenance is condition-based maintenance carried out following a forecast derived from repeated analysis or known characteristics of asset degradation.

So the useful question is not "predictive or preventive". Both are preventive: both aim to act before failure. The question is what triggers the work — a date on a calendar, or a measurement that says degradation has started.

EN 13306 branch	Trigger	Common name
Predetermined preventive	Elapsed time or run hours	Preventive / PM
Condition-based preventive	Measured condition crosses a threshold	CBM
Predictive (condition-based)	Forecast from a degradation trend	Predictive / PdM
Corrective	After the failure	Reactive / run-to-failure

The three strategies in practice

Stripped of terminology, the difference is simply when you act. Reactive fixes things after they break. Preventive services on a fixed calendar regardless of condition. Predictive uses the asset's measured condition to act just before failure.

Strategy	When you act	Strength	Weakness
Reactive	After failure	No upfront cost	Unplanned downtime, secondary damage
Preventive	Fixed schedule	Simple, avoids many failures	Replaces healthy parts; misses random faults
Predictive	Just before failure	Avoids failures with least wasted work	Needs monitoring data and analysis

Why fixed intervals waste work and still miss failures

A fixed interval assumes wear accumulates predictably with time or run hours. For some failure modes it does — abrasive wear, filter loading, lubricant degradation. For many it does not. Reliability-centred maintenance work from the airline industry onward found that a large share of failure modes show no wear-out age at all: their hazard rate is effectively flat, so servicing on a calendar neither prevents them nor predicts them.

That produces the two failure modes of a purely interval-based programme at the same time: healthy components are replaced early (wasted parts, wasted labour, and the infant-mortality risk of disturbing a working assembly), while randomly-arriving faults still reach failure between services.

The P-F interval decides whether prediction is possible

Prediction is not a property of the technology. It is a property of the failure mode. The P-F interval is the time between the point where a developing fault first becomes detectable (P) and the point of functional failure (F). It sets the entire economics of a predictive programme.

Two rules follow, and they are the ones most programmes get wrong:

• Your inspection interval must be shorter than the P-F interval — as a working rule, no more than half of it — or the fault will arrive and fail between two checks and the monitoring buys you nothing.
• The P-F interval must be long enough to act: to raise the work order, get the part, and take the outage. A technique that gives two days' warning on an asset with a six-week spare lead time has detected the failure without preventing it.

This is why the same asset can justify continuous monitoring on one failure mode and nothing at all on another. Bearing degradation detected by vibration typically develops over a usefully long window; a brittle fracture does not develop at all.

Where predictive maintenance fails

Honest limits, because a programme that ignores them turns into shelfware:

• No P-F window. Sudden failure modes cannot be predicted by any sensor.
• No baseline. Trending needs a healthy reference. Monitoring installed on an already-degraded machine measures the wrong normal.
• Alarms with no owner. The commonest failure is organisational: the analysis produces an alert and no process converts it into a scheduled job.
• Duty-cycle confounding. Load, speed and process changes move the signal. Without operating context, variable-duty assets generate false alarms until people stop trusting the system.
• Cost floor. On a cheap, redundant, easily-replaced asset the monitoring costs more than the failure. Run-to-failure is the correct engineering answer there.

How to assign a strategy per asset

It is not either/or, and the assignment should be deliberate rather than inherited. Work down the asset register and ask three questions in order:

1. What does failure cost? Safety, environmental, production, quality. If the answer is "almost nothing", stop — run to failure.
2. Does the dominant failure mode have a detectable P-F interval long enough to act on? If no, use predetermined intervals or design the failure out.
3. Is there a technique that sees that specific mode? Match the technique to the mode, not to the asset type.

Most well-run plants end up with all three strategies running side by side: run-to-failure on cheap redundant items, fixed intervals on predictable wear and on statutory inspections, and predictive on critical rotating equipment — pumps, motors, fans, compressors, gearboxes — where failure is expensive and degradation is visible early.

Standards worth reading

Standard	What it gives you
EN 13306	Maintenance terminology — the definitions used above
ISO 17359	Condition monitoring and diagnostics of machines: general guidelines
ISO 13379	Data interpretation and diagnostics
ISO 20816	Vibration measurement and evaluation of machine vibration
ISO 14224	Collection and exchange of reliability and maintenance data
ISO 18436	Qualification and assessment of condition monitoring personnel

Frequently asked questions

Related guides

Software that helps