Ask a maintenance team to rank their machines by importance and you will get a confident answer. Ask them to fill in a criticality field on four hundred of them and you will get empty boxes — because the ranking is situational and the field looks permanent.
An empty field is not zero
A blank criticality reads as “unimportant” to a sorting function and as “unknown” to a human, and those two readings do not survive contact with each other. Printing Not recorded instead of leaving a gap is a small thing that prevents a machine being quietly treated as low priority because nobody had got round to it.
Do not invent severity names for somebody else's scale
The tempting move is to label a 1–5 scale with words that sound authoritative: low, moderate, high, serious, critical. It reads better. It also asserts a grading scheme — and a machine's criticality is a plant's own judgement, not a software vendor's taxonomy.
So the words are the plain ordinal reading of the number and nothing more: Lowest, Low, Medium, High, Highest. If your plant wants severity language, that is your vocabulary to define, and the number carries it faithfully.
Identity fields are the ones that pay back
Criticality is useful; maker, model and serial number are what you reach for when a part-finding exercise starts with “what is actually in this machine?”. All four are optional, all four read back as Not recorded when absent, and a manager or administrator can correct them from the machine's own page without a data-cleansing project.
Start rough, improve where it matters
A rough ranking applied to your twenty most critical machines beats a precise ranking applied to none. Criticality is a decision aid, not an audit — and a system that lets you improve it a machine at a time will get there, while one that demands a complete dataset first will not.