PM and PdM effectiveness: how much of the work you found actually needed doing

A time-based PM calls for a bearing to be replaced every six months. The mechanic strips the machine, inspects the bearing, and finds it in perfect condition. The replacement goes ahead anyway, because the work order said so. That job gets counted as corrective work, but it never needed to happen.

Effectiveness puts a filter over the corrective pile. It separates the genuine defects your programme caught from the jobs that turned out to be unnecessary, and that distinction tells you whether your preventive and predictive tasks are well calibrated or quietly wasteful.

What it actually measures

PM and PdM effectiveness is the proportion of corrective work orders raised by inspections that were genuinely necessary, where a real defect or developing failure was confirmed and put right. Work orders where the investigation found nothing wrong, or where the recommended fix turned out to be the wrong one, count as unnecessary and pull the ratio down. Where yield asks how much work the inspections found, effectiveness asks the sharper question of how much of that work deserved to be found at all.

How to work it out

PM and PdM effectiveness = Necessary corrective work orders / Corrective work orders written

Say eight corrective jobs come out of inspections in a month. A scheduled bearing swap finds the bearing perfect; a calendar-driven gasket change finds nothing amiss; and a filter change made on a hunch does not fix the problem, because a faulty gauge was the real cause. The other five are genuine: a vibration route catches a real misalignment, a thermography sweep finds a hot connection, a heat-exchanger check finds it fouled and choking production, an oil sample flags water ingress, and a performance test catches a worn impeller beyond limits.

PM and PdM effectiveness = 5 / 8 = 0.63 (63%)

Three of the eight were unnecessary, and the pattern is worth noticing: the condition-based checks earned their place far more reliably than the calendar-based ones, which is exactly the kind of insight this metric exists to surface.

Why effectiveness is subtler than it looks

Behind this simple ratio sits one of the most important and least intuitive findings in maintenance, and it explains why "just do the PM" is not the whole answer. Decades of failure studies have shown that only a small minority of failure modes, somewhere around a tenth, are genuinely age-related, the kind a scheduled overhaul can actually prevent. The great majority strike more or less at random, and a large share cluster in the period just after an intrusive job, the infant mortality that careful work avoids and clumsy work creates. The implication is sobering: for most of what you maintain, replacing a part on a calendar cannot prevent the failure it targets, and the act of opening the machine can introduce a new one. That is why effectiveness can be low not because the crew is careless but because the strategy is mismatched to the way the equipment really fails.

What the evidence shows

The numbers from the field are striking. When one industry actually inspected the bearings it had been swapping on a fixed schedule, only a twentieth were anywhere near failure; a third had nothing wrong at all, and most of the rest carried only minor damage that the maintenance itself had caused. A large naval operation found that something like a tenth to a fifth of all its maintenance work was rework, much of it correcting defects introduced by earlier intrusive tasks. Read against that backdrop, a modest effectiveness score is not a scandal, it is a measurement working as intended, telling you that some of your scheduled intervention is finding healthy machines and occasionally leaving them worse than it found them.

What good looks like

There is no fixed numerical target; the closer to one, the better, and the real value is in the trend and the split. Track preventive and predictive effectiveness separately, because they fail in different ways and call for different fixes. A low score on time-based PM usually means the frequency is simply too high, and the cure is to extend the interval and retest. A low score on condition-based work more often reflects technicians who need sharper training in reading the data, since the false alarms come from misjudged readings rather than over-frequent visits. And beware the comforting vanity metrics that can mask a failing programme: a plant can celebrate fewer unscheduled delays while its scheduled downtime and its spare-parts bill quietly climb, which is precisely the self-deception that yield and effectiveness exist to expose.

Where it can mislead you

• A low score on time-based PM tasks usually means the frequency is too high. Extend the interval and retest before doing anything drastic.
• A low score on condition-based tasks more often reflects technicians needing sharper training in reading the condition data than anything wrong with the tasks themselves.
• Do not confuse it with yield. Yield counts how much corrective work was found; effectiveness asks whether that work was warranted. A programme can score high on one and low on the other.
• Track it at the asset level on critical equipment, and reach for root cause analysis when unnecessary work orders start clustering around a particular task or technology.

PM and PdM effectiveness keeps an inspection programme honest about its own output. It separates the catches that earned their keep from the interventions that merely made work, and sometimes made things worse, and in doing so it points you precisely at the tasks worth keeping, the ones worth stretching, and the ones worth rethinking altogether. It is the difference between a programme that is merely busy and one that is genuinely working.