Industrial equipment criticality: building a matrix that earns its keep

A criticality matrix ranks equipment by what its failure would cost. It is a simple tool, one that almost every site has built at least once, and one that almost none of them still opens two years later. Understanding why that happens is more useful than building yet another one.

What a matrix has to let you do

Before picking any scales, be clear about the purpose. A criticality matrix exists to settle three concrete questions:

• Where should the inspection effort go? Which items warrant close monitoring, and which can be spaced out.
• What comes first? When the available capacity is smaller than the workload, which piece of equipment jumps the queue.
• What can you justify? In front of an auditor or a management team, being able to explain why this item is watched closely and that one is not.

If the matrix you are building does not answer at least one of those three questions in an operational way, it will end up as a study document. That is the test to apply before you tie up a team for several weeks.

The severity axis: build it once, build it properly

Severity does not depend on the condition of the equipment. It depends on where the equipment sits in the installation. That is exactly what makes it stable: it only changes when the process changes.

Four families of consequence deserve to be rated separately, because they do not have the same arbiter and because they do not offset one another:

Table scrolls horizontally on small screens.

The rule that spares you nasty surprises: the severity you retain is the highest of the four, never an average. An item with no production stake but which is indispensable at audit stays critical. Averaging the four families dilutes the very signal you are trying to isolate.

A four-level scale is enough. Go beyond that and the discussion drifts onto where the boundary between level five and level six should sit, instead of staying on the equipment itself.

The probability axis: the one that causes the trouble

This is where everything is decided. The probability of failure cannot be estimated in a meeting, and yet a meeting is almost always how it gets rated: on gut feel, with whoever happens to be in the room, under the influence of the last memorable incident.

Three sources let you establish it seriously:

The failure history already recorded on this item or on comparable ones. This source has a limit: failures are rare, so they are few in number, and therefore statistically fragile.

The observed condition at inspection: open findings, recorded reservations, defects detected. This is the richest source and the least used, because it lies dormant in PDF reports.

The measured rate of degradation: this is the best of the three, because it is continuous rather than binary. An item whose wall thickness is dropping steadily has a probability of failure that evolves in a predictable way. See tracking equipment degradation.

A method that holds up over time

Design traps

• Too many levels.A five or seven-level scale pushes the discussion onto rating boundaries instead of onto the equipment. Four are enough.
• Averaging the severity families.An item that is critical for compliance but neutral for production comes out as "medium", which is both wrong and dangerous.
• Rating probability in a meeting.With no condition data, that rating reflects the memory of whoever is present, dominated by recent incidents.
• Rating the whole fleet.Exhaustiveness kills the update; a sound matrix on fifty items beats an out-of-date one on three thousand.
• Not recording the justifications.A rating with no stated reason cannot be debated, it can only be redone in full.
• Confusing criticality with urgency.A highly critical item on which nothing is moving can wait: criticality steers the monitoring effort, it does not set the running order of the work. See prioritising maintenance by real risk.

What keeps a matrix alive

The difference between a matrix that earns its keep and one that gathers dust comes down to a single thing: is the update of the probability axis automatic or manual?

If every inspection report loaded in updates the observed condition and the rate of degradation of the items concerned, the matrix recalculates itself and flags the equipment whose position has changed. It then becomes a steering tool: you no longer consult it to verify a rating, you consult it to see what has moved since last time.

That is what an automated intake of the existing reports makes possible: pulling in findings, measurements and deadlines without rekeying, and recomputing the position of each item at every new campaign. The overall approach is set out in risk-based maintenance and on our Maintenance Intelligence page.

For the equipment families most often involved, see pressure equipment and pipework.