Nobody will give you a deterioration curve

Ask when a thing will need replacing and you are asking about a curve. Condition on one axis, time on the other, an intervention threshold drawn across it somewhere, and the year the two meet is the year the money is needed. Every renewal bid, every maintenance budget and every argument about deferral rests on that picture, whether or not anybody draws it.

So it is worth knowing where the curve comes from. The answer, from the bodies whose job it would be to say, is that it comes from you.

The equation without its terms

ISO 15686 is the international standard for service-life planning of buildings and constructed assets, and part 8 is the part that estimates how long something lasts. It sets out the factor method: take a reference service life and multiply it by seven factors, for the inherent quality of the component, the design, the workmanship, the indoor environment, the outdoor environment, how hard it is used, and how well it is maintained. Seven multipliers, A to G.

It is a reasonable structure. A pump in a coastal plant room worked around the clock and never serviced should not carry the same life as the identical pump in a dry basement on standby, and the factor method is how that difference is meant to be made explicit rather than argued about.

The standard's own abstract, on ISO's website, then says that it does not give guidance on how to estimate the values of factors A to G.

That is not a criticism of the drafting. It is the standard being honest: the factors depend on materials, climate, workmanship and operating regimes that no international committee can enumerate, and a table of invented multipliers would be worse than none. Saying so is the right call.

The consequence still lands on the reader. On ISO's own description, the standard it lists at 181 Swiss francs for 36 pages gives the shape of a calculation whose every term the buyer must supply. The standard has been confirmed once, in 2018, sits at stage 90.92, to be revised, and is expected to be absorbed into the revised parts 1 to 3.

Two paragraphs on the largest asset class there is

Roads are the biggest constructed asset most governments own, and the World Road Association is the global body for managing them, with member governments across most of the world. Its Asset Management Manual is online, free, and has a section called Deterioration models.

That section is two paragraphs. It offers two approaches. Service life, which it describes as based on industry best practice and local knowledge, adding that an organisation's own records of material performance should be a good reference. And performance modelling, where deterioration profiles come from historical performance, local knowledge and best practice, and where it notes that some organisations have developed their own profiles at significant cost in data and calibration.

The third place to look is HDM-4, which does contain real deterioration equations and came out of World Bank work. But the literature around it is almost entirely about calibration: crack initiation, crack progression and roughness progression each carry a local coefficient, and studies exist fitting those coefficients country by country because the shipped defaults are not expected to hold anywhere in particular. The model is global. The numbers are not, and the software is licensed commercially.

Three bodies, three different fields, one answer.

Why the gap is where it is

There is a good reason for it, and an owner is better off knowing the reason than resenting the gap.

Deterioration is not a property of an asset type. It is a property of an asset type in a place, under a load, in a climate, built by somebody, maintained to some standard. A chiller in Riyadh and the same model in Rotterdam are the same machine and not the same asset. An international body publishing a single curve would be publishing a number that is wrong nearly everywhere, and would be believed anyway, because a number in a standard carries authority its derivation does not.

So the delegation is correct. What is missing is the next sentence. Having said the numbers must be local, none of the three then says what an owner should do who has no local numbers, which is most owners, and particularly the ones building assets for the first time. The advice to use your own records assumes records. It is offered to organisations whose whole problem is that they have none.

There is one more place to ask, and none of these three bodies points at it. The supplier of the plant has built thousands of the same machine and knows how they fail. Whether an owner can require that knowledge at procurement turns out to have a better answer than this one: the method for asking is published, and rail, oil and gas and the dependability committee itself have each settled how. What none of them has done is carry it into buildings, whose reference specification for what a contractor hands over never uses the word reliability.

The other place the number could come from, and what to ask for

, read: Reliability can be specified like mass. Building owners do not.Reliability can be specified like mass. Building owners do not.

That is where this stops being a story about standards and becomes a story about what happens next in a real budget cycle. The curve does not go away because nobody published it. Somebody in a spreadsheet picks one, usually a straight line from new to the end of a depreciation life, because that is the only number anyone can find, and it was chosen by an accounting convention rather than by anything to do with wear.

The honest version

An owner in this position has four moves, and the first two are bad.

Use the depreciation life as the curve. It is available and it is already in the accounts, which is exactly why it gets used. It is also a statement about how cost is allocated to reporting periods, not a statement about when a thing stops working, and the two are not related.

Buy someone else's curve. National manuals do publish them, and they are usually fitted to a road network or a building stock somewhere else. Borrowing one is a defensible starting point as long as its origin travels with it, and indefensible the moment the origin is dropped and the curve is quoted as though it were a fact about your estate.

Or state the curve you assumed, its source, and what it would take to replace it with an observed one. This is the move worth arguing for, and it costs a line in a document. It also makes the assumption testable: our renewal profile exists to show what happens to a budget when one assumption about lives is swapped for another, and the same swap can be done to a curve.

That is a particular case of a proposal this publication already makes about judgements generally, which is to record the basis and not only the result. It is worth restating here because the deterioration curve is the clearest example of the problem there is: it is entirely invented, it decides the number, and it is almost never written down.

The fourth move, and what it looks like at scale

Recording the assumption is the minimum. The move that actually solves the problem is to start generating the curve, and the largest worked example of it is public.

BUILDER, built by the US Army Corps of Engineers' Construction Engineering Research Laboratory, is the Department of Defense standard for facility condition assessment, and the Marine Corps and the Navy adopted it as well. Its structure is worth copying whatever the estate. Each component carries a Condition Index on a 0 to 100 scale. The system predicts that index from where the component sits in its expected life cycle, which is to say it starts from an assumed curve. Inspections then verify the prediction, and the curve is corrected against what was found. The condition-versus-time model developed in the laboratory's own research is a Weibull distribution, the standard shape for time to failure.

The assumed curve is not the answer. It is the prior, and the design exists to overwrite it.

The scale makes the point rather than weakening it. ERDC describes an Army estate of 165,000 buildings and 1.1bn square feet, roughly 55% of whose real property maintenance funds go on maintaining them, and describes the system itself as a response to repeated criticism of DoD facility management practice by the Government Accountability Office. An organisation with that much building and that much money did not solve the missing curve by finding one. It built an inspection regime that generates its own, and it published the method while keeping the software behind a login.

The obvious objection is that inspections cost money and an owner without records usually lacks them for a reason. That is true, and it is why the honest sequence is the third move then the fourth: write down the assumption now, because it costs a line, and start correcting it where the assets are expensive enough to justify looking. Criticality decides where to spend the inspection budget, which is the same question as where to spend anything else.

What this changes about reading a maintenance plan

ISO 55001's 2024 revision renamed Clause 10.3 from preventive to predictive action, which moves the question from whether to intervene to when. Read alongside what part 8 and the road manual actually publish, that is a demanding position. The standard asks for the year. The bodies that would supply the curve behind the year say the curve is yours.

None of that makes the requirement wrong. Predicting is better than waiting, and a requirement that forces an organisation to notice it has no condition data has already earned something. But it does change what a reviewer should ask when a plan arrives with intervention dates in it.

The question is not whether the dates look reasonable. Dates always look reasonable. The question is what curve produced them, where that curve came from, and what would have to be observed for it to change. If the answer is that the dates came from a depreciation schedule, the plan is a finance document wearing an engineering title, and the first intervention it gets wrong will be discovered by the asset rather than by the plan.