The choice between preventive and predictive maintenance is made per asset and not per site, which is why site-wide debates about it go nowhere. Two questions, asked about one asset at a time, produce an allocation that everybody can then argue with specifically rather than generally. Most assets land in the same place, a small number land in monitoring, and a few land somewhere people find uncomfortable, which is usually where the useful conversation actually is.
The two questions
What does a failure of this asset cost, including downtime and the emergency response rather than the planned one. And does this failure mode give warning that something could measure. Both answers come from the people who would deal with the failure, and writing them down for each significant asset is a morning's work that settles arguments for a year.
The four boxes the answers produce
Expensive and gives warning: a monitoring candidate, on an interval backstop. Expensive and no warning: redundancy or a spare, plus whatever inspection genuinely helps, because no interval catches a sudden failure. Cheap and gives warning: an interval, and monitoring is not worth the step cost. Cheap and no warning: run to failure with a spare on the shelf, chosen deliberately rather than by omission.
The box people avoid
Expensive with no warning is the uncomfortable one, because neither strategy solves it and the honest answer is engineering: a standby unit, a spare held, or accepting the exposure explicitly. Sites frequently respond by shortening the interval instead, which costs money and does not address the failure mode, and it is worth naming that when it happens. It is also the box where a shortened interval is most often proposed as a substitute for the engineering answer, which costs real money every year and leaves the exposure exactly where it was, so it is worth saying out loud in the room rather than letting it pass as diligence.
Questions people ask about preventive vs predictive maintenance
How many assets end up monitored?
At most sites, a small number: the large rotating plant whose failure stops something. If the exercise produces a long monitoring list, the consequence question was probably answered generously.
Who should do the allocation?
Maintenance and operations together. Done by maintenance alone the consequences reflect maintenance's inconvenience rather than the site's, and the disagreement between the two groups is itself informative.
Should the allocation be revisited?
Once a year, and whenever the process changes. Assets move category when what depends on them changes, and nothing about the asset itself signals it has happened.