Somewhere on your job there is a signed inspection sheet for an air-handling unit, and nobody can tell you which unit. The sheet is entirely legitimate. A QA engineer stood in front of the plant, checked what the plan told her to check, signed it and filed it exactly where the quality plan said. Eight months later a facilities supervisor asks whether AHU-07 was ever witnessed, and the honest answer is a shrug and an offer to go and read the nameplates.
Inspection and test plans already exist on almost every job. Asset tags exist too, or are promised. What usually does not exist is the join between them. The ITP lives in a quality folder named after a trade; the tag turns up later, on a nameplate, in a COBie row, or during a walk somebody does at practical completion. Then a team spends the last month attaching signed sheets to things they were never written against, and everyone agrees to call that handover.
The change that makes the last month behave is not a new form and not a new folder. Every ITP row that will produce a record the owner might later need gets bound to an asset tag before the first hold point, rather than after the last one. It is also, in my experience, the hardest single line to get agreed in a quality plan, because it costs a little effort very early against a cost that only lands once the people who wrote the plan have moved to another job.
A plan is not yet a record
An inspection and test plan is a planned sequence: hold points, witness points, review points, named authorities, acceptance criteria, and the records that prove the work met them. Write one for "air-handling units" or "fire doors" and what you have is a recipe. It tells you what must be checked. It says nothing about which unit, or which door, was actually checked.
A useful instance of that plan can answer, for a single maintainable thing:
- Which asset tag?
- Which activity or hold point?
- Who was allowed to sign, and who did?
- What evidence — certificate number, photograph, measurement, serial?
- What status — passed, failed, waived, still open?
If the only identity on the sheet is a package, a floor, or "AHU typical", the asset register has nothing to inherit. The plan can be excellent and still be no help at handover, because handover does not ask questions about trades. It asks about things.
The tag is what the later question will use
An asset tag is a stable identity for something operations will maintain: AHU-03, VAV-12-04, DB-L2-West, FD-204. It survives a drawing revision. It survives a room rename. It is the key the later query runs on.
It is not a room number, because rooms get recoded. Not a drawing mark reused across three floors. Not a WBS or activity code — those name work, not the thing the work happened to. Not a BIM identifier nobody ever printed on the casing, and certainly not a filename inside a quality ZIP.
Provisional is fine. Issue the tag from the equipment schedule before the crate arrives, then confirm it against the serial on delivery. What it cannot be is absent at the moment of inspection, because an inspection recorded against "the units on level two" is a sentence somebody will be retyping in eleven months' time.
Classification and type still matter, and nobody is suggesting you write a separate plan per door. You write one ITP for a class of fire door, then instantiate it onto tagged doors before the first hold point on each door. Type is the recipe; the tag is the plate. Without the plate, every certificate, snag and commissioning sheet that follows has nothing to hang on.
Day one means the first inspection, not the first week on site
"From day one" gets heard as a remark about mobilisation. It is really a rule about when the join is allowed to be missing, and the answer is never.
Day one is the first inspection, factory test or delivery check that an owner, an adjudicator or a facilities supervisor might later ask about. For bought-in plant that is usually factory acceptance or goods-in: bind the serial to the future tag before the crate leaves the works, or at the very latest when it is booked onto site. For things built in place — fire-stopping, partitions, buried drainage — settle the identity before the first inspection anyone will cite at handover, even if that identity starts as a location-and-type key that becomes the tag later. For items that only become maintainable once installed, issue the tag at install and hold the first site hold point until it exists.
"We will map them at completion" is not day one. Neither is "the model will reconcile the identifier later", and nor is a walk by the O and M contractor. A provisional tag you confirm afterwards is still a tag. An anonymous signed sheet you guess at afterwards is something else entirely.
The characteristic miss is the early inspection that looked cheap precisely because no tag had been issued yet. Goods-in against a purchase order. A factory witness against a batch. A first-fix check against a gridline. Those are exactly the hours handover will come asking for, and filed by trade they will not travel.
Why the tidy version fails
The usual sequence looks organised, which is most of the problem.
- Write the ITP by trade or system.
- Inspect against lots, floors, or a sample.
- File the signed sheets in a quality folder.
- At completion, rebuild an asset register from nameplates, PDFs and a walk.
- Try to attach the quality records to the new register.
You now hold two registers, the second of which is a story told about the first. Join keys that were never written cannot be recovered by diligence, however conscientious the recovery. "Level three AHUs — witnessed 12 June" does not tell you whether AHU-07 was in the lot. A ten per cent sample that passed does not populate the other ninety per cent of the tags. A certificate filed under "MEP quality" does not inherit a serial from anywhere.
Late mapping is not a clerical delay. It is a different activity: you have stopped recording what was inspected and started inferring it. Inference is how an O and M manual turns into fiction, and how a golden-thread request comes back as a ZIP.
Snags opened against a room instead of a tag carry the same defect. So do commissioning sheets opened against a system name. They all need the same labour at the end because they all declined the join at the start, which is the argument in quality records are handover records, arriving a good deal earlier in the programme.
What you get if the bind holds
There is no separate handover dataset waiting to be produced. What the owner receives is the live join of identities you already created: this tag, these inspections, these certificates, this serial, this warranty start, these open hold points, these closed snags.
If the ITP instance sat on the tag from the first hold point:
- The asset register is a roll-up rather than a retype. Each tag already carries its own quality history.
- Completeness becomes countable. You can list tags with no ITP instance, ITP rows with no tag, and tags with an open hold point. Those three lists are the readiness review.
- An O and M query — this valve, this certificate, this spare — is a filter on the same identity. You export it rather than rebuild it.
- A revision to the equipment schedule updates the tag's attributes without orphaning the inspections, because they were never filed by drawing number alone.
Nobody fills a handover pack in the last fortnight by magic. The pack is the current state of a join kept while the work was live, and the labour at the end moves from reconstruction to exception: the untagged inspection, the tag with no record, the hold point still open. A commissioning manager who inherits that position spends a fortnight chasing stragglers rather than a month doing archaeology. Capturing commissioning data during commissioning is the same rule applied to as-left values and witnesses; the ITP is the earlier chance to get the identity right.
A test you can run on Monday
Pick one class the client will operate: air-handling units, fire doors, distribution boards. Accept only a record as an answer, not a reassurance from whoever owns the folder.
- Does every installed instance in that class have a tag, on the thing and in the register?
- Does the ITP exist as instance rows against those tags, or only as a type PDF in a package folder?
- For the last inspection that happened, can you open it from the tag without unzipping a quality archive?
- For any open hold or witness point, is it blocking a named asset or a named trade?
- Can you list the tags in that class with no inspection record, and the signed sheets with no tag?
If all five point at the same identity, handover is already assembling. If they point at a trade folder and a walk somebody will do later, you will retype the building.
None of this needs a new quality system. It needs the equipment schedule, the ITP and last week's signed sheets on one table for an hour. Where the tag is missing, write it down. That is the finding, and it should not be softened into "mostly linked" by the time it reaches the report.
What the standards already assume
The principal contractor’s quality manager should write the bind into the quality plan before the first hold point: every ITP instance sits on a tag that already exists on the equipment schedule, and an untagged inspection is not accepted. Give the witnessing engineer cover to stop the hold point until the tag is there. That is a procedure change — and, if the ITP was already issued by trade, a short addendum — not a new quality platform. An hour with the schedule, the ITP and last week’s signed sheets is enough to see whether the join exists.
Wait until the last fortnight and you are paying for reconstruction: a walk, a retype, and an O and M pack that guesses which unit was in the lot. The standards already assume the records can be found against the assets. They will not invent the tag for you.
Read next: Quality records are handover records. Most teams learn this too late., Capturing commissioning data during commissioning, not after, The handover readiness review, run at sixty per cent complete.