先談關鍵性，後談頻次

Quarterly or six-monthly? Annual thermographic survey or semi-annual? These debates consume review meetings, and they are unresolvable on their own terms, because frequency is an output. It falls out of the failure mode, the consequence of that failure, and the interval over which deterioration can be detected.

Argue frequency first and you are negotiating budget. Argue criticality first and you are deciding which failures you are unwilling to accept — after which most frequencies stop being controversial.

The consequence of an asset failing depends on its position in the power, cooling or control chain: what stops, whether redundancy actually covers it under maintenance, how long restoration takes, and whether the failure is detectable before it matters.

This is why a control transformer can outrank a chiller. The chiller failing is visible, covered and recoverable. A hidden failure in a trip circuit is none of those things — it is discovered at the exact moment the protection was needed.

A programme that maintains everything on the same cycle has made an allocation decision, just not deliberately. It has spent scarce technician hours evenly across assets whose consequences differ by orders of magnitude.

The visible symptom is schedule non-compliance: a backlog of low-value planned work that the team quietly triages, so the actual maintenance strategy ends up being set by whoever builds the weekly schedule.

A ranked list is not the deliverable. The deliverable is a set of rules: what a class-1 asset gets that a class-3 asset does not — in analysis depth, monitoring, spares holding, response time and capital priority.

Without those rules, criticality is a spreadsheet. With them, it becomes the mechanism by which FMEA, RCM, condition monitoring, spares and investment are all pointed at the same small number of assets that actually carry the service.