Both comparisons in this pair get argued as though one side wins. Neither does. Preventive against reactive, and condition-based against predictive, are choices made per asset rather than per site, and the same plant will reasonably run all four. What decides each case is a small number of properties of the asset itself.
Preventive against reactive, decided per asset
Reactive maintenance is cheaper per event and unpredictable in timing; preventive is more expensive in total labour and predictable. So the trade turns on what the unpredictability costs. An asset whose failure stops production, damages something else, or creates a safety issue justifies preventive work even where the parts had life left. One that is cheap, redundant and quickly swapped does not, and putting it on a schedule consumes attention that the first category needs. That is the whole of the argument, and it is per asset.
Condition based against predictive
Condition based maintenance acts when a measurement crosses a threshold: the vibration exceeds a limit, the temperature rises, the oil degrades. Predictive maintenance uses the trend of those measurements to estimate when the threshold will be crossed, and schedules ahead of it. The practical difference is lead time. Condition-based tells you to act now; predictive tells you to act in three weeks, which is what lets the work be planned rather than squeezed in. Predictive needs more history and better instrumentation to be trustworthy.
Where most sites should actually be
A site with no instrumentation and an incomplete asset list gains far more from a working interval-based programme than from any of the more sophisticated options, and the sophistication is often reached for precisely because the basics are unglamorous. Get the asset list true, the intervals defensible and the findings recorded, and the data to justify condition-based work on the assets that deserve it will accumulate as a by-product.
Questions people ask about preventive maintenance vs reactive maintenance
Is preventive maintenance always better than reactive?
No. It is better for assets whose failure stops production, damages something else or creates a safety issue. For cheap, redundant, quickly swapped items, reactive is the right economic answer.
What is the difference between condition based and predictive maintenance?
Condition based acts when a measurement crosses a threshold; predictive uses the trend to estimate when it will and schedules ahead of it. The difference is lead time to plan.
Where should a site with no instrumentation start?
With a true asset list, defensible intervals and recorded findings. That produces the data needed to justify condition-based work later, as a by-product of doing the basics.