Sit with an estates director who is responsible for two hundred buildings and ask to see the asset register. What usually appears is a spreadsheet she maintains herself. Not out of preference — the last three capital jobs each handed over something that would not join to the other two. Schools and depots, a hospital wing, a housing block, an office refit, a leisure centre with a facilities contract wrapped round it. Every one of them arrived with its own naming, its own common data environment, its own idea of what “as-built” meant, and an assurance that somebody would load the pack into the computerised maintenance management system after practical completion. Somebody usually did, once, by hand. The estate ends up speaking two hundred dialects and the register is whatever one person can hold together in a workbook.
The instinct is to call that a data-quality problem, which makes it sound like a cleaning job for a graduate with a spare fortnight. It rarely is. What is missing is one standard the two-hundred-and-first appointment cannot opt out of: a shared identifier scheme, a living asset information model, and a defined route out of a project CDE into the system the maintenance team actually queries.
ISO 19650, ISO 55000, CDM, the Building Safety Act, NEC, FIDIC, JCT, COBie and the Gulf mandates all assume there is somebody on the owner’s side who can say what the estate will want to query long after the job closes. None of them writes that sentence for you.
Two hundred buildings, two hundred dialects
“Asset” here means a building, a site, a demise the owner will still be responsible for in ten years. Not a fan. Two hundred is roughly the scale at which informal knowledge dies. Below it, somebody still remembers which plant room uses the contractor’s codes and which blocks were renamed. Above it, that person has retired, the long-term facilities contract runs its own maintenance system, the academy conversion relettered the blocks, and the last capital project published a COBie workbook that will not map onto the functional locations already in use.
The default state is not chaos, and it is worth being clear about that before anyone starts apportioning blame. Each project team did a reasonable job for the people who built that site. Each facilities supervisor did a reasonable job for the people who operate it. The portfolio fails because those reasonable jobs never had to share an identity.
Anyone who has walked a mixed estate will recognise most of these:
- A new-build that ran a proper CDE and then treated the export as the asset information model.
- A 1970s block whose register is a spreadsheet the last supervisor left on a shared drive, with his initials still in the change log.
- A facilities-management contract that will not release asset numbers in a form the owner can keep after the contract ends.
- A refurbishment that reused the designer’s type names as instance names, so every air-handling unit on the floor shares a tag.
- A civic campus where the wayfinding, the fire strategy and the maintenance system disagree about what “Block C” is.
Two hundred of those adds up to something other than a technology gap. It describes an owner who never issued one standard, then wondered why the estate would not answer a question.
What a portfolio standard actually contains
A portfolio standard is the small set of rules every appointment and every operational team has to use, so that a stranger to any one site can still find the thing. It is not a platform, and buying one before writing the rules is how owners end up with a second register to reconcile.
At two hundred assets it has three jobs, and only three.
- Naming — how a building, a space and a maintainable instance are identified, uniquely, in language the maintenance system and the night supervisor will both use.
- The AIM — what “accepted asset information” means on every capital job, and what “good enough to operate” means on stock that will never have a project CDE.
- The trip — how a row leaves a project CDE and arrives as a live asset in the maintenance system, carrying the same identifier, a location the estate already uses, and the document bound to that tag.
Everything else can stay local. Folder trees can differ. Authoring tools can differ. Framework contractors can differ. The standard is only the join. Where a project is allowed to invent a second join, what you have is a style guide with an ISO number on the cover.
Write it short enough to issue with the invitation to tender. If it needs a consultant in the room to explain, it has become a novel again, and novels get filed. A ten-page standard that names identity, the AIM fields and the import rule will outperform a ninety-page protocol that only the last new-build ever opened.
Naming: the identity that outlives the job
Delivery names things for the people who are building them. The estate names things for the people who will still be there long after the defects period. Those two can differ as aliases. They cannot differ as identity.
The portfolio identifier is the string operations will search for over the next twenty years. It is not AHU-L3-01 inherited from a mechanical package, or the designer’s type family, or a facilities contractor’s internal code. It is built from four things:
- an organisation or estate code that will survive a merger or a machinery-of-government change, or at least a written rule for how that code is retired;
- a building or site code matching the wayfinding and the fire strategy rather than last year’s organigram;
- a location a technician can walk to — floor, room, or the functional location the maintenance system already holds;
- an instance identifier that is unique in the receiving system, printed on the tag, and never reused.
Project names sit beside that as a cross-reference, not in place of it. The asset information requirements on the next appointment should say so plainly: author the portfolio identifier, or map to it before the trial drop, or the row is rejected. Most identity gets lost in the friendly tidying-up that happens after practical completion, when nobody is watching and the names look untidy to a fresh pair of eyes.
Classification belongs to naming, not to decoration. Pick the table the estate will query — Uniclass, for most UK public owners — and use it as a type list rather than something each job reinvents. “All maintainable assets” is not a class; air-handling units, fire dampers, fire doorsets, incoming switchboards and lifts are, because those are what the planner already raises work orders against. Let one project invent a second classification because its authoring tool had a convenient parameter and the import stalls; from then on, facilities walk the plant with a clipboard.
Two questions will tell you where you stand.
- Given only the tag on the casing, can a night supervisor find the row without ringing whoever ran the last capital job?
- Can two buildings on the estate share a type name without sharing an instance name?
If either answer is no, you still have personalities rather than a standard.
None of this means retagging two hundred buildings first. Issue the scheme, apply it on every new appointment and every replacement, and retag stock when a planned task already has the casing open and an electrician standing next to it. A catch-up spread over five years costs labour you were going to spend anyway. Two hundred one-off imports cost more, and you pay for those in arrears.
One acceptance test, every appointment
ISO 19650 already draws the line. The project information model is what delivery produces; the asset information model is what the operator queries. At portfolio scale the trouble is not that teams have never heard of the distinction — most have it in their execution plan — but that each job performs the cut as a local habit, at a milestone of its own choosing, to a field list somebody assembled in a hurry.
One AIM for the portfolio means one definition of accepted. It does not mean one federated model of the entire estate, and owners who conflate the two spend a year producing a slide.
So issue the same asset information requirements on every capital appointment: the same object types operations actually maintain, the same fields, the same milestones, the same rejection rule. The AIR the supply chain can price is that sheet, and the PIM-to-AIM cut is the ownership question underneath it. Do not let a framework contractor write a second version because their execution plan already had a chapter for it.
The fields should be the ones facilities have ever asked for, not the museum of delivery metadata that accumulates by Stage 4. A location the operator will walk to. Manufacturer, model, serial. Warranty start, end and holder. Spare or lead time. Isolation. The document bound to the tag. Which fields actually matter sets out that cut in detail; the portfolio standard is simply how the same cut gets issued on every job instead of being renegotiated at each gateway.
Name the person who accepts the AIM — a role in the owner’s organisation, not “the CDE” and not the outgoing contractor. Acceptance is a pass or a fail against the standard, and it needs somebody senior enough to say no in the week before demobilisation. A crate of PDFs in a folder called Handover is not acceptance. Neither is a workbook that will not map onto the estate’s functional locations. If the appointment gives nobody the power to reject a model for failing to feed the AIM, the standard is advisory, and advisory standards turn into personalities within two jobs.
Existing stock will not grow an AIM because a protocol was published, and it is worth saying so out loud in front of the board. For buildings with no capital job this decade, the AIM is the live maintenance row, plus the current O and M pointer, plus whatever the health and safety file still owes. Bring those up to standard when a replacement, a statutory inspection or a planned outage already has labour on site. The standard is the destination; the catch-up is a programme with a budget line.
Public-sector owners have a duty-shaped reason not to leave this local. Golden-thread obligations, landlord compliance, audit and the next framework tender all assume the owner can show what is installed, where it is, and how it is inspected — across the estate, not on the newest building. Enterprise owners face the same demand under a different letterhead: insurers, warranty holders, energy reporting, a disposal.
From the project cupboard to the work order
Connecting a CDE to a maintenance system does not have to be a two-year integration. Most of it is a rule about what is allowed to leave a project.
The trip fails in the four ways facilities already rebuild from: names that will not join, locations that will not map, tags that do not match the nameplate, documents that will not open from the asset. A portfolio standard does not wait for middleware to solve those. It forbids them on the next appointment.
Write the import as four joins rather than an interface diagram:
A spreadsheet copied once is not a trip. It is a photograph of a Friday afternoon. Somebody has to own the route after practical completion — updating the row when a spare changes, when a warranty is extended, when a unit is replaced — and if that owner is “we will integrate later”, the maintenance team will type. The CDE-to-CMMS method covers the mechanics; the portfolio rule sitting above it is short. No project CDE gets a second asset identity, and no import is accepted that breaks one of the four joins.
Resist the temptation to run two hundred CDEs as two hundred systems of record. The CDE is the project cupboard. The maintenance system is where the operational question gets asked. The AIM is the content that has to survive the move between them. A capital job can have its own CDE; it cannot have its own idea of what an asset is.
Updates need a cadence, too, or the whole thing dies in the first winter. Condition, replacement, a closed work order that changed the spare — all of that has to land back on the same identifier. The standard only needs two sentences here: the identifier does not change when the work order closes, and a replacement is a new instance rather than a quiet edit of the old serial number.
Try this on five sites
You do not need a consultant to find out whether the standard exists. Give it a morning.
- Write down the identifier scheme operations search today — not the protocol, the search box. Two buildings using two schemes means no portfolio standard.
- Pull the AIR, or the handover specification, from the last three capital appointments. If the object types, fields or rejection rules differ, those jobs were personalities.
- Pick five maintainable instances on five different sites. Starting from the tag on the casing, reach the row, the location a stranger could walk to, and the document bound to that tag. Give yourself an hour per site. A hunt is a finding.
- Ask who is allowed to reject a handover that will not import. “The project closed” means the standard was optional.
- Ask who owns the row once the defects period ends — a role, with a name in it, not a platform. A shrug is the vacuum, already scheduled.
Five sites are plenty. Nobody is completing an estate register on a Monday; the question is only whether the next job will be allowed a dialect of its own.
What to issue, and what it will cost
Four documents, in this order, and none of them long. One page of identifier scheme, with aliases permitted and identity fixed. The AIM field list and the rejection rule, attached to every invitation to tender from now on. The four joins, written as import acceptance criteria. And the named acceptor — an owner-side role who can fail a row while the delivery team is still on site and still has something to lose.
Only the appointing party can do this. A framework contractor cannot impose an identity scheme on an owner who has not chosen one, and a facilities provider has little incentive to hand back numbers it currently controls. That means it lands on the estates director, the head of property or whoever signs the next tender — and it is not free. Expect to pay for someone’s time to write the four documents, for a contract variation if a current appointment is already mid-flight, and for a retagging programme that runs alongside planned maintenance for years rather than months. Against that, weigh what is already being spent rebuilding the register every time a job completes, which is real money nobody itemises because it looks like business as usual.
Do it before the next capital appointment goes out, not after it completes. The standard is cheap to write into a tender and expensive to retrofit into a handover, and the two-hundred-and-first building is already in somebody’s pipeline.
Read next: Monthly Register — June 2026, Connecting a CDE to a CMMS without a two-year integration project, Why facilities teams rebuild the asset register from scratch.