Preventive became predictive, and the second needs three things the first did not

Preventive action asks what might go wrong and puts something in the way of it. Predictive action asks when this will go wrong and decides what to do about the timing.

They sound adjacent. They are not, and the gap between them is where the 2024 revision of ISO 55001 has quietly raised the bar.

Clause 10.3 was called preventive action in the 2014 edition. In the second edition it is called predictive action, and ISO lists it among the main changes with an unusual amount of emphasis: renamed, and fully technically revised. Both halves of that matter. A rename alone would be presentational. A rename plus a full technical revision is a statement that the previous content was answering a different question.

The committee said what the new question is, and its phrasing is more demanding than the clause title suggests.

Optimal intervention points is the phrase that costs money. Anticipate is a posture and any organisation can claim it. An optimal intervention point is a date, and a date can only be produced by something that models how a parameter moves. That is the difference between the two clause titles expressed as a deliverable, and it is why the revision cannot be met by renaming an existing preventive maintenance schedule.

What the harder question requires

Deciding whether to intervene requires a judgement about consequence. Deciding when to intervene requires three things most organisations have not built.

Condition information, collected repeatedly. Not an inspection report from four years ago. A series, because a single observation tells you the state and only a series tells you the direction. The interval matters more than the precision: two rough measurements a year apart are more useful for timing than one exact measurement.

A view of how this class of asset deteriorates. Something that converts condition today into an expectation about condition later. This does not have to be sophisticated, and treating it as a modelling problem is how organisations avoid starting. It has to be written down and it has to be capable of being wrong, which is the property that lets it improve.

A funding mechanism that can act before failure. This is the one that actually stops people, and it is not technical at all.

Why the money is the binding constraint

An organisation can have excellent condition data, a defensible deterioration view, and still be structurally unable to act on either.

Intervening early costs money in a year when nothing has broken. The case for it rests on a cost avoided in a later year, in a budget that is not the same budget, defended by a forecast that cannot be proven because if the intervention works the failure never happens.

Every part of that is uncomfortable in an annual budget round. Meanwhile the competing bid is a failure that has already occurred, has a photograph, and has somebody demanding it be fixed.

A prediction that is acted on successfully destroys its own evidence. The failure it prevented cannot be shown to anybody, and the money spent can.

So predictive maintenance programmes are frequently approved, resourced, and then quietly starved, not because anyone opposed them but because each individual year had something more urgent. The organisation ends up with monitoring equipment producing data nobody can act on, which is worse than not having it, because it converts an acknowledged blindness into a documented one.

The reasonable defence, which is sometimes right

Running to failure is a legitimate strategy. It is the correct strategy for a large proportion of most asset bases: items that are cheap, redundant, quick to replace, and whose failure has no safety or service consequence. Predicting the failure of a light fitting is a waste of attention.

The distinction that matters is whether run to failure was chosen or defaulted to.

A chosen regime can be stated: these asset classes run to failure, for these reasons, and here is what we hold in spares to make that recovery quick. A defaulted regime is described in the same words and cannot answer the follow-up, because the same approach is applied to the critical items as to the trivial ones, and nobody has separated them.

Measurement, and measuring the wrong thing

Clause 9.1 requires monitoring, measurement, analysis and evaluation, and it sits upstream of all of this, because nothing can be predicted that is not first measured.

The common failure is not absence of measurement. It is measurement of activity rather than of condition. Planned maintenance completion against schedule, work orders closed within target, percentage of inspections done. Those are all real and they measure whether the maintenance function did what it said it would.

None of them says anything about the state of the asset. An organisation can achieve 98% planned maintenance compliance on an estate that is deteriorating steadily, and the dashboard will be green throughout, because the dashboard is reporting on the department rather than on the assets.

The test is whether any reported measure would change if the asset base got worse while the maintenance function continued performing exactly as it does now. If none would, the measurement system cannot see deterioration.

Five measures commonly reported on an asset base, all of which describe activity rather than condition: planned maintenance completed, work orders closed in target, inspections completed, compliance tasks signed off, and backlog within threshold. None of them would move if the asset base got worse.

Item	Measures activity	Moves if the asset degrades
Planned maintenance completed		No
Work orders closed in target		No
Inspections completed		No
Compliance tasks signed off		No
Backlog within threshold		No

Five accurate reported measures, every one of them true, and not one of them would change if every asset in the estate deteriorated this year. An organisation can hold all five at green and be getting worse.

Improvement, and the thing that is not a nonconformity

Clause 10.1 covers continual improvement and 10.2 nonconformity and corrective action. Those are familiar from every management system standard and organisations generally operate them.

What 10.3 adds is a category that is not a nonconformity. Nothing has failed. Nothing is non-compliant. The asset is working. The action is warranted purely because of where the trend is heading, and no existing process in most organisations is designed to initiate work on that basis.

That is the real implementation gap. Organisations can predict. What they lack is any route by which a prediction becomes a funded instruction, because every existing route begins with something being wrong.

Why the assurance function is unlikely to notice

, read: Audited on the transactions, unaudited on the basisAudited on the transactions, unaudited on the basis

Where to start, given the region's position

Much of the asset base here is young, which is usually presented as an advantage and is one, but not the one people assume.

Young assets are not yet generating the failures that force this capability into existence, so there is time. They are also generating, right now, the only baseline that will ever exist. Condition data collected on a new asset is what later measurements get compared against, and an organisation that starts collecting at year twelve has no idea what year one looked like.

The cheapest useful action available to a recently completed estate is a condition baseline, recorded against an asset hierarchy that finance and engineering both recognise, with a stated interval for repeating it.

Where the intervention decisions get made

, read: Operational control now says life cycle management, which is a larger change than it looksOperational control now says life cycle management, which is a larger change than it looks

The framework for this stage sets out the tests, including whether renewal, replacement and disposal timing is optimised together or handled by separate teams on separate cycles.

A baseline is the one piece of this that cannot be bought later at any price, because the year it describes will have passed.

---

Sources. ISO 55001:2024, second edition, July 2024, ISO standard 83054, Clause 9.1 on monitoring, measurement, analysis and evaluation, Clause 10.1 on continual improvement, Clause 10.2 on nonconformity and corrective action, and Clause 10.3 on predictive action. The renaming of preventive action to predictive action, and its full technical revision, are listed by ISO among the main changes from the 2014 edition in the published foreword. ISO/TC 251, Updated Guidance on Asset Management System Requirements, 2024, for the committee's definition of predictive action as actions that predict future behaviour of a parameter required to support decision-making, and for the stated purpose of determining optimal intervention points. ISO 55000:2024, ISO standard 83053, which introduces asset management maturity at its Clause 4.5. ISO 55013 on the management of asset data.