Reactive vs Preventive vs Predictive Maintenance: The Three Maturity Levels and the Path Between Them

Key takeaways

• Reactive maintenance = fix when it breaks. Firefighting, often expensive.
• Preventive maintenance = time-based scheduled work. Reduces breakdowns at the cost of some over-maintenance.
• Predictive maintenance = condition-based action driven by sensor data and models.
• The three are maturity levels. Most plants do all three; the question is the mix.
• World-class is roughly 10% reactive, 60% preventive, 30% predictive — not zero reactive.

Short answer: Reactive maintenance fixes assets after failure. Preventive maintenance does scheduled work on time intervals. Predictive maintenance acts on sensor data and condition models before failure. They are maturity levels — most plants do some of all three, and the right mix depends on asset criticality. World-class operations are not 100% predictive; they are around 10% reactive, 60% preventive, 30% predictive. See also Condition Monitoring vs Predictive Maintenance.

What reactive maintenance is

Reactive maintenance happens after failure. The asset breaks; the technician fixes it. Sometimes called run-to-failure (when intentional) or breakdown maintenance (when not).

Pros:

• No upfront cost on assets that rarely fail.
• No over-maintenance on cheap, replaceable assets.

Cons:

• Unplanned downtime is expensive.
• Collateral damage when failures cascade.
• Spare parts pressure when failure is unexpected.
• Safety risk on critical assets.

What preventive maintenance is

Preventive maintenance is scheduled work on time or usage intervals. Replace bearings every 2000 hours. Inspect every 30 days. Lubricate every 100 cycles.

Pros:

• Reduces unplanned downtime.
• Plannable; resources are predictable.
• Simple to schedule.

Cons:

• Over-maintenance on assets that did not need work yet.
• Misses failures that develop between intervals.
• Intervals based on average, not actual condition.

What predictive maintenance is

Predictive maintenance acts based on condition data: vibration, temperature, oil quality, performance metrics. The action happens when the data says failure is approaching, not on a fixed schedule.

Pros:

• Acts only when needed.
• Catches failures that preventive intervals miss.
• Optimizes spare parts and labor.

Cons:

• Requires sensor instrumentation.
• Requires data history for model training.
• Higher upfront cost.
• Models can be wrong (false positives, missed failures).

The right mix

Not every asset deserves predictive maintenance. The mix depends on criticality:

• Critical assets: mostly predictive plus PM as backup.
• Important assets: mostly preventive.
• Standard assets: preventive or reactive depending on cost.
• Cheap, easy-to-replace assets: reactive (run-to-failure as policy).

World-class plants typically run something like 10% reactive (mostly run-to-failure as policy), 60% preventive, 30% predictive. Not zero reactive.

The path from reactive to predictive

1. Get out of reactive dominance. Reactive over 50% means firefighting. PM compliance must reach 70%+ first.
2. Standardize PM. Right intervals, right tasks, RCM-based.
3. Add condition monitoring on critical assets. Vibration, thermal, oil analysis where ROI justifies.
4. Accumulate failure data. 1-3 years of failure history with sensor data.
5. Build predictive models. Trained on the accumulated data.
6. Iterate. Models improve; criticality classification refines.

This path is 3-5 years for most plants. There is no shortcut.

Why pure predictive does not exist

Even mature programs have:

• Reactive for cheap assets (intentional run-to-failure).
• Preventive for assets without good predictive models.
• Predictive for critical, well-instrumented assets.

Vendors that promise "100% predictive" are usually overselling.

Common mistakes

1. Skipping preventive to go straight to predictive. Predictive needs failure history that preventive operation produces.

2. Treating reactive as failure. Some reactive is correct policy. The question is whether it is intentional.

3. Universal preventive intervals. Without RCM, intervals are guesses.

4. Predictive without acting on alerts. Model predicts failure, technician ignores. The predictive value disappears.

How to measure the mix

Maintenance work order classification:

• Reactive: triggered by failure (operator request, alarm, breakdown).
• Preventive: triggered by schedule.
• Predictive: triggered by condition data.

Tally percent of total maintenance hours by category. Trend over months.

How a modern CMMS supports the mix

A modern CMMS classifies every WO by trigger (reactive, preventive, predictive), tracks the mix over time, and surfaces opportunities to migrate from one category to another.

Fabrico's CMMS classifies WOs by trigger, tracks the maintenance mix over time, and supports condition-based triggers for predictive maintenance.

See how Fabrico captures this automatically — explore OEE for manufacturing or book a demo.

Related reading

• Condition Monitoring vs Predictive Maintenance
• Predictive Quality vs SPC

Frequently asked questions

Is reactive maintenance always bad?

No. For cheap, easy-to-replace assets, reactive can be the right policy.

What is a healthy reactive percentage?

World-class is around 10-15% (mostly intentional run-to-failure). Above 30% indicates firefighting.

How long does the journey to predictive take?

3-5 years for a typical plant.

Can I skip preventive?

Usually no. Preventive operation accumulates the failure data that predictive models need.

Does AI make predictive faster?

Modestly. Models need data regardless of algorithm. Better algorithms accelerate the journey but do not replace it.