Reliability can be specified like mass. Building owners do not.

This publication has already argued that nobody will hand an owner a deterioration curve: the global bodies publish the shape of the calculation without the numbers, and the organisation that solved the problem built an inspection regime that generates its own curve rather than finding one.

The standards give the equation and withhold the factors

, read: Nobody will give you a deterioration curveNobody will give you a deterioration curve

That piece looked at the standards bodies. It did not look at the other place a number could come from, which is the supplier of the plant. Every chiller, pump, switchboard and lift arrives with a manufacturer behind it who has built thousands of the same thing and, somewhere, knows how they fail. So: can an owner require that knowledge at procurement, and what arrives when they do?

The answer comes in two halves that point in opposite directions. There is a published method for asking, agreed through global bodies and used by whole sectors. There is almost nothing in the buildings sector that asks. An owner's leverage sits in the gap between those two facts, and the gap is the subject of this piece.

The bodies do say how to ask

The International Electrotechnical Commission's committee on dependability, TC 56, publishes a standard whose whole purpose is to tell a purchaser how to specify reliability and how to check that a supplier delivered it.

That sentence about dimensions and mass is the whole argument. Nobody buys a chiller without specifying its cooling capacity, its footprint and its weight, because those are checked on delivery. Reliability is treated as a different kind of property, one the owner hopes for rather than specifies, and the body that writes the standards says it is not a different kind of property at all.

Rail went further and wrote the process down as a standard of its own. IEC 62278, first published in 2002 and replaced on 24 July 2025 by IEC 62278-1 and 62278-2, sets out a process for specifying reliability, availability, maintainability and safety requirements and demonstrating that they have been achieved. The IEC's published scope for the 2025 edition applies it to signalling, rolling stock and fixed installations, from complete railway systems down to individual components, and says plainly what it does not do: it does not set targets or quantities, because those are the purchaser's to set. The supplier of a lift for a station and the supplier of a lift for an office tower may be the same company, and only the first is procured under a standard that asks it to demonstrate a failure rate.

The oil and gas industry took a third route. ISO 14224, in its 2016 edition, standardises the collection of reliability and maintenance data by operators during the equipment's operating life: a common taxonomy, a failure-mode thesaurus, a minimum data set, and the requirements a data system must meet so that owners, manufacturers and contractors can exchange records. Its introduction says two things an owner should hear. Data collection is an investment. And it can be necessary to have data covering several years of operation before sufficient data have been accumulated to give confident analysis results.

So three of the largest sectors that buy plant have each, through a global body, settled how reliability is specified, demonstrated or recorded. The buildings sector is not among them.

What the buildings sector asks for instead

To test what a building owner actually specifies, the document to read is the widest-reaching one that is free to read. The Unified Facilities Guide Specifications are the master specifications of the US Army Corps of Engineers, the Naval Facilities Engineering Systems Command and the Air Force Civil Engineer Center: free, current, and the considered statement of one of the largest building owners in the world about what a contractor owes it. Section 01 78 23, Operation and Maintenance Data, is the one that says what a contractor must hand over about the plant it installed.

Read what is there beside what is not. The owner is told how many hours a repair will take and which craft will do it. The owner is not told how often the repair will be needed. In the vocabulary of IEC 60300 that is maintainability and maintenance support without reliability, and it is the one combination from which no lifecycle figure can be computed, because work-hours per repair multiplied by an unknown number of repairs is unknown. The specification collects two of the three terms and leaves out the one that carries time.

This is not a criticism of the specification's authors, who were writing a document about how to operate and maintain what was installed, and did it thoroughly. It is an observation about where the sector's attention stopped. The document is the sector's considered statement of what a contractor owes an owner at handover, maintained across three services and revised four times since 2023, and reliability is not on it.

What a supplier's number can and cannot mean

Suppose the owner asks. What should they expect, and what should they be wary of?

The first thing to know is where a reliability number can legitimately come from, and the one place the whole industry agrees is rolling bearings. ISO 281 defines the basic rating life as the life associated with 90% reliability under stated conditions of material, manufacture and operation, and gives the calculation. That is why a bearing catalogue carries an L10 figure and a chiller catalogue does not: the bearing number rests on a published international method and the chiller has none. Where a supplier quotes a life or a failure rate for an assembly, the owner's first question is what method produced it, and the honest answers are a small set: a prediction from component data, a field return rate from a warranty population, or an operator's records shared back. Each has a different meaning and a different bias, and a number offered without its method is a number offered without its meaning.

The second thing to know is what ISO 14224's introduction says: the confident numbers come from years of the owner's own operating data, pooled with others'. A supplier's figure at procurement is a starting estimate, not the curve. Which turns the question round. What the owner most needs from the supplier is not a number but the structure in which the owner's own numbers can accumulate: the equipment taxonomy, the failure modes the supplier recognises, the boundary of what counts as the item, so that the maintenance record from day one is written in a form that will one day support an analysis. That is a data deliverable, and it is exactly the kind of thing an asset information requirement should have stated before design began.

Information requirements set after design can only compile, not capture

, read: Handover is where the cost lands, and the Gulf is now taking deliveryHandover is where the cost lands, and the Gulf is now taking delivery

Which reading this takes

A requirement that almost no building owner applies can be read four ways, and this publication's reading has to be stated.

It is not that specifying reliability is unnecessary: three sectors that face the same physics have each decided it is worth a standard. It is not that it is unaffordable: the IEC guidance is a 112-page application guide and the rail process has been a published international standard since 2002. It is not that it is unenforceable: the same guide covers verification and validation, and the rail standard's purpose is demonstration as well as specification.

The reading is that the buildings sector never carried the question across from the sectors that answered it. Its reference specification is thorough about how to maintain plant and silent about how long it lasts, and the silence is inherited each time the document is copied. That is a defect of practice, of the plainest kind: the method exists, is free to read in outline, and is not asked for.

The limit of the evidence is that one specification has been read, albeit the one with the widest reach, and that the IEC and ISO documents are cited from their published scopes and free previews rather than their full text. Nothing here shows what any particular owner received; it shows what the reference document told the contractor to provide.

What to examine

• Ask for the O&M data specification the project is using and search it for the words reliability, failure rate and expected life. If the answer is what it was above, nothing the contractor delivers will carry a lifecycle figure, however complete the manuals.
• Ask whether repair work-hours have been specified without failure frequency. If so, the sheet will say what a repair costs and not what the plant costs to keep running, which is the figure the business case needed.
• Where a supplier quotes a life, ask for the method and the population. A figure with neither is marketing.
• Ask whether the asset information requirements name the failure-mode taxonomy the maintenance system will record against, and whether the supplier's documentation uses the same one. If the two do not meet, the owner's own data will never accumulate into anything analysable.
• Ask whether anything in the estate has ever been procured against a stated reliability requirement with verification. If the answer is no, that is the finding, and it says nothing about the suppliers.

---

Sources. International Electrotechnical Commission, IEC 60300-3-4, Dependability management, Part 3-4: Application guide, Specification of dependability requirements, edition 3.0, 11 March 2022, 112 pages, sold; the published scope on the IEC webstore, and the introduction and clause 1 of the 2007 second edition read in the free sample published by iTeh, from which the dimensions and mass sentence and the three types of specification are taken. IEC, IEC 62278:2002, Railway applications, Specification and demonstration of reliability, availability, maintainability and safety (RAMS), withdrawn 24 July 2025 and replaced by IEC 62278-1:2025 and IEC 62278-2:2025; the scope of 62278-1 is cited from the IEC's published abstract, which states its application to signalling, rolling stock and fixed installations and that it does not specify targets or quantities. ISO, ISO 14224:2016, Petroleum, petrochemical and natural gas industries, Collection and exchange of reliability and maintenance data for equipment, introduction and clause 1 read in the free sample. ISO, ISO 281:2007, Rolling bearings, Dynamic load ratings and rating life, clause 1 and definition 3.4 read in the free sample. US Department of Defense, UFGS 01 78 23, Operation and Maintenance Data, May 2023, Change 4, August 2026, 28 pages, read in full from the Whole Building Design Guide; the content list is from clause 1.6 and the repair work-hours requirement from 1.6.6.4. All read on 6 September 2026.