Where the thirty per cent of lifecycle data lost at handover actually goes

An FM team taking over a finished building will usually tell you they were promised a complete handover. What they get is a share that looks full: folders, a ZIP named something like FINAL_HANDOVER_v3, a model, a pile of PDFs, perhaps a spreadsheet from the last fortnight. Ask them for the warranty start on a particular air-handling unit, or the as-left value from the sheet that accepted it, and the room goes quiet. Half of what they were told would be there is somewhere else — a mailbox whose owner has left, a drive copied once and then reorganised, a filename nobody on the facilities side can join to a tag.

‘Thirty per cent of lifecycle data is lost at handover’ is the sentence the industry reaches for at that moment. It appears on conference slides, in vendor decks and in board papers. You can say it without looking anything up. It almost never arrives with a denominator, a year, a geography or a method section. Treat it as a circulated figure, not as a census. The slogan never answers the narrower question: if information that was paid for during design and construction is not queryable by the people who will operate the asset, where did it actually go?

It did not go into a skip. It leaked at a short list of ordinary, fixable moments.

A circulated figure, not a census

Most AECO numbers arrive without a method. Documentation volume without incident rates. Automation percentages without a definition of non-physical work. ‘Thirty per cent lost at handover’ without the moment of loss. If there is no method section, you are reading a brochure.

The figure is useful as a slogan because it is large, round and portable. Sometimes it is offered as a share of project information. Sometimes it is implied as a share of lifecycle cost. Sometimes it is quietly mixed with other industry round numbers — interoperability waste, rework bands, the cost of operating versus building — until it sounds like a law of construction. Those other literatures exist. The 2004 NIST interoperability study, for example, put a dollar cost on inadequate information exchange in the United States capital-facilities industry and found that owners and operators bore most of it. Construction Industry Institute and academic rework papers give ranges, not a single constant. Soft Landings and the performance-gap literature describe a different object again: the building that does not perform as designed. None of those sources is a primary study that counted ‘thirty per cent of lifecycle data’ disappearing at practical completion. If someone in the room can produce that study — population, definition of ‘data’, definition of ‘lost’, date — the claim becomes a citation. Until then it is a gesture.

A number that cannot say what it counted cannot tell you what to fix. The useful move is to stop defending the percentage and name the moments.

Leak point	What happens	What ‘captured’ should mean instead
The export	CDE states, suitability and transmittal history collapse into a flattened ZIP or a once-generated COBie dump	Can you still say which published containerrevisioninspection and as-left value sat togetherinside project custody?
The rename	A ‘friendly’ filename strips the persistent identifier that joined the record to the asset	Keep the project container name as the identityadd any local label as metadata only
The inbox	Warranties, serials and certificates live in personal mail and never enter project custody	Nothing counts as captured until it is a project-owned container tied to the asset tag
The USB	A drive changes hands with no register of contents, no checksum, no named custodian afterwards	A controlled transfer needs a list: which containerswhich revisionswhich assetswho now holds the copy
No AIM owner	Nobody is named to accept, validate or refuse the dump before practical completion	An AIM owner named in the appointmentwith authority to refuse a dump that cannot be queried

What ‘lost’ actually means

Lost, on a real job, almost never means deleted.

The model still exists. The inspections were signed. The commissioning sheets were filled in. The warranties were issued. The as-built redlines were marked. The certificates sat in a mailbox. Something that was true of the works at the time still exists somewhere. What the operator cannot do is retrieve it from an owned asset information model, grouped by the identifier they will use, without a guided tour of the project that has already demobilised.

That is a custody and identity failure, not an incineration. The file that cannot be found from the asset is lost to operations even if it is sitting in that ZIP. The serial that lives only in a photograph folder is lost even though the photograph is sharp. The warranty that lives only in a package manager’s inbox is lost the week that person leaves, even if the vendor would send another copy if anyone still knew whom to ask.

ISO 19650’s distinction is unromantic about this. Project information (PIM) is what the delivery team needed in order to design and build. Asset information (AIM) is what the appointing party needs in order to operate, maintain and later alter. Handover is the cut between those two objects. If the cut is a dump rather than a transfer of owned, identifiable records, the PIM can be complete and the AIM can still be empty. The slogan calls that ‘thirty per cent lost’. The more accurate sentence is: a large share of what the project paid to create never became something operations could query.

Five places that share goes.

The export

The first loss is a translation.

For two or four or seven years the project information lived in a common data environment with states, suitability, transmittals, container names and — if anyone was paying attention — relationships between a drawing, an inspection, a certificate and an asset. At practical completion someone is asked to ‘export the handover’. What leaves is usually a flattened dump: a ZIP of PDFs, a folder tree rebuilt for the owner’s share, a model without the issued 2D set, a COBie spreadsheet generated once in the last fortnight.

Export is not a copy. It is a lossy conversion. Suitability codes die. Transmittal history dies. The link from the test sheet to the tag dies unless someone already wrote that link as a field. Container identifiers become filenames. Work-in-progress, shared and published collapse into one pile labelled Complete. What arrives on the owner’s side is a souvenir of the PIM, not the PIM itself.

COBie applied once is the same moment in schema form. The schema is a way to structure asset information. A late spreadsheet filled from photographs and vendor emails is a reconstruction wearing a standard’s name. It does not recover the hour the test ran, and it does not prove the PIM was transferred. It proves that someone needed a file that would open in Excel.

If you want to know whether export is where your job is leaking, ask one question: after the dump, can you still say, for a given asset, which published container, which revision, which inspection and which as-left value sat beside it inside the project’s custody? If the answer is only ‘it is in the ZIP’, the export already spent the percentage.

The rename

The second loss is a courtesy.

Owners ask for ‘friendly’ names. A document controller who has been told to make the pack presentable ‘cleans up’ for the receiving folder. An ISO 19650 container that joined the inspection to the plant item becomes AHU-L3-OM.pdf. Two files that were distinct in the CDE now look like the same thing. Two files that were the same thing now look distinct, because someone added _FINAL, _OWNER, _USE_THIS. Search by the project’s unique identifier fails. The register the operator was promised cannot be joined to the file they were given.

Rename feels like service. In practice it is how identity gets stripped. The inspection that accepted the unit was a handover record only while it still carried the identifier the operator will use. Once that identifier is a memory, the file is a brochure.

Keep the project container name as the persistent identity. If the owner needs a local label, add it as metadata or a second field — do not overwrite the name that joined the record to the asset. A rename convention that cannot survive the receiving share will not survive an FM team either.

The inbox

The third loss is correspondence that never entered project custody.

Warranties, serial numbers, commissioning certificates, spare-parts lists and the vendor letter that changed a maintenance interval routinely live in mail. They were never in the CDE, so they cannot be exported from it. They were never a container, so they cannot be renamed. They sit in a package manager’s thread, a subcontractor’s forward, a ‘see attached’ that sometimes still attaches.

An inbox is a personal communications tool, and it behaves like one. When the sender leaves, IT follows the leaver process. The archive may keep a snapshot, or it may not; the snapshot may omit the folder they kept on a laptop. Forwards into private accounts are worse. The project paid for the warranty. The project never owned the file.

This is the same mechanism as treating email as the contract record, pointed at a different reader. On a claim, the missing instruction is a commercial problem. At handover, the missing attachment is an operations problem. The operator who cannot find the warranty start, the serial or the certificate is not being difficult. They are asking for a record the project left in a mailbox.

No warranty, serial or commissioning certificate should count as captured until it is a project-owned container, tied to the asset tag, still retrievable when the sender has gone. Volume of mail is not evidence. Custody is.

The USB

The fourth loss is ceremonial.

A drive is handed across a table. Sometimes it is a hard disk. Sometimes it is a stack of disks. Sometimes it is a cardboard box of binders with a drive in the lid. Someone signs a receipt. Everyone feels that handover has occurred. The drive is copied once onto a facilities share, or it is not; the copy is complete, or it is not; the original is kept in a drawer, or it goes home in a laptop bag and does not come back.

USB is a moment of apparent completeness with no provenance, not a format in any useful sense. There is often no register of what was on the media at transfer, no checksum, no named custodian after the meeting, and no rule for what happens when the first copy is reorganised to fit the owner’s folder tree. The media’s only job was to leave the project.

Physical media is not automatically incompetent. A controlled transfer can use a drive. What makes the USB a loss moment is the absence of a list: what containers, which revisions, which assets they belong to, who now owns the copy, and where the project copy still lives. Without that list, ‘we handed over the data’ describes a meeting, not a record.

Nobody owns the AIM

The fifth loss is the one that makes the other four permanent.

ISO 19650 assumes an asset information model with an owner — the appointing party, or the person they name to keep it. CDM’s health and safety file, the golden thread on higher-risk buildings in England, and the ordinary employer’s information requirements all assume a receiving object that will still be there next year. They do not name the FTE. They do not fund the role. They do not say who accepts the dump, validates it against the assets that will actually be operated, and refuses the ZIP that cannot be queried.

If nobody is named before practical completion, operations invent a third model. That third model is usually a computerised maintenance management system rebuilt from PDFs, nameplate photographs and a consultant’s spreadsheet. The rebuild is billed as mobilisation, as facilities onboarding, as ‘data cleansing’. It is the twelve-month handover in labour form: owners paying again for information they funded in the first place.

PIM-to-AIM has no natural owner. The information manager’s appointment often ends with the project. The contractor’s document controller demobilises. Facilities were invited late, or not at all. Soft landings, if anyone used the words, sat in a separate meeting. The dump arrives at a share with no parent. That is an appointment gap, and it will swallow a perfectly competent export.

Name an AIM owner before practical completion — with authority to accept, reject and keep the model alive — or the export, the rename, the inbox and the USB have nowhere to land. Data with no parent will be reconstructed, and reconstruction is where the circulated percentage is spent, even if nobody ever counts it.

What the thirty per cent does not prove

It does not prove that thirty is the right number on your job. It does not prove that software, a mandate or a federated model will save the outcome by itself. It does not prove that a late COBie file, a golden-thread policy or a new CMMS licence will recover facts that were never written against an asset while the people were still in the room.

It proves that this claim needs a source, a limit and a next measurement.

A measurement that would mean something is boring and local. Define the population: the assets that will still need maintenance in two years. Define ‘present’: queryable in the receiving AIM or CMMS, carrying the identifier the operator uses, with contemporaneous provenance — inspection, test, serial, warranty — still in project or owner custody, not reconstructed from a photograph folder. Count present at practical completion. Count again ninety days later. The gap is your number. It will not be thirty, except by accident. It will point at one of the five moments.

What would falsify the slogan on a given job is simple. Take a sample of assets. Produce, without a guided tour and without emailing a leaver, the identity, the contemporaneous quality or commissioning record, and the warranty from an owned AIM. If you can do that, ‘lost at handover’ is not true of that sample. One good job does not retire the circulated figure. It shows that the figure was always a claim about a process, not a constant of buildings.

Cite what you can. Leave a hole where the data stops. The hole in this claim is the missing primary study. The map is not missing. The information goes into the export, the rename, the inbox, the USB, and a building that never appointed an AIM owner.

The appointing party is the only person who can close those five holes before they become a facilities rebuild, and it is not free. Name the AIM owner in the appointment — an actual person, with time after practical completion, and authority to refuse a dump that cannot be queried by asset tag. Write into the information requirements that the project container name survives the receiving share, that warranties and serials are project-owned containers before they count as captured, and that a USB or disk transfer includes a register of what left, who holds the copy, and where the project copy still lives. Then fund the local count: present at practical completion, present again at ninety days, against the assets that will still need maintenance in two years. That measurement costs a few days of a named person’s time. Reconstruction billed as mobilisation costs a year. Until the appointing party pays for the first, the slogan will keep circulating, and the FM team will keep opening a share that looks full.