Journal/Playbook

How to set up a construction defect register that survives handover

The register you hand over is never the register you kept, and about a third of it cannot be rescued in the fortnight before practical completion. The fields, the evidence rules and the status ownership to settle in the first hour of the job.

A two-colour risograph poster of a schedule grid whose entries are solid ink bars rather than words, its header band and one full column picked out in orange
WP
Will PrinmanCo-founder
24 August 2026·5 min read

The register you hand over is not the register you kept. Somebody spends the last fortnight before practical completion turning eight months of shorthand into a document a client can read, and about a third of it cannot be rescued. Photos that live on a phone. Rows that say "as discussed." Three different spellings of the same contractor.

None of that is a discipline problem. It is a setup problem, and the fix takes an hour at the start of the job.

Start with the fields, not the tool

Whatever you end up buying, the register has to answer the same questions. Write them down first and you will evaluate tools on whether they can hold your register, rather than reshaping your job around someone's default columns.

A permanent number. Assigned once, never reused, never encoding anything that can change. Not the trade, not the level, not the status. If your numbering scheme has to be edited when a defect gets reassigned, it is not an identifier, it is a label.

A location that exists in a tree, not a text field. "Level 3 bathroom" is not a location, it is a description of one. Building, level, unit, room, as structured rows, so a filter for level three actually returns level three. This is the field people cheapest out on and regret hardest, because you cannot retrofit structure onto eleven months of free text.

A description written for a stranger. The person reading it in November is not the person who typed it in March. "Crack" is useless. "Hairline crack, 300mm, running from the window reveal down the east wall" is a record.

What kind of record this is. A defect, incomplete works, a warranty item, a variation. Collapse those into one bucket and your defect count at handover is fiction, because a third of it is work that was never defective in the first place.

A classification. Your own taxonomy, the failure and the trade. Ours seeds around eighty entries and every organisation edits it, which is the point. It is the field the register sorts and filters on all job long, so it earns the setup time.

Who raised it. Not who typed it. A tenant complaint attributed to the site admin who entered it is a broken link the day the tenant disputes what they said. Keep a directory of the people who raise things, including the ones with no login.

Who is doing the work, and separately, who is wearing the cost. Most of the time they match. The cases where they do not are exactly the cases that turn into money arguments, and a comments field will not carry it.

Dates that mean different things. Logged, due, and the moment work actually started. That third one matters more than people expect. We stamp it on the first transition out of open and never touch it again, so time-in-progress is a real measure and not "how long since somebody typed this in."

Status. Which is the next section, because it is not really a field.

Evidence rules, decided before day one

The rule to set now: what does someone have to attach before a defect can move forward?

Our answer is that a trade cannot mark work complete without a photo of the finished work. The button stays disabled until there is one. It is a small rule and it eliminates the most common dispute on any site, which is not "was it fixed" but "was it ever actually looked at."

Set the same expectation at capture. A photo taken at the moment of capture carries its own timestamp, GPS and device metadata. A photo airdropped in three weeks later carries an argument.

And decide who is allowed to delete. Our answer is nobody: a removed note leaves a tombstone showing that something was there and when it went, without exposing the content. Choose differently if you like, but choose deliberately, because the default in most tools is a silent hole in the record.

Every status needs a named owner

A status list is only worth having if each state names who can leave it. Write yours as a table before the job starts:

StateWho moves it outOn what
OpenAnyone on your teamAssignment to a trade
Assigned / in progressThe tradeTheir own completion claim, with a photo
Ready to inspectYour side onlySomebody physically looking
ClosedYour side onlyVerification, and sign-off if the contract wants it

The line that matters is the third one. The party who did the work does not get to declare it accepted. In our role model trades have no defect transitions at all, which surprises people until the first time a subcontractor would otherwise have closed forty of their own items the week before handover.

A cadence, not a heroic sweep

Twenty minutes, twice a week, on the register rather than on site. What is overdue, what has been sitting in ready-to-inspect for more than a few days, what has no owner. That is the whole ceremony.

The failure mode it prevents is the pre-PC sweep, where six weeks of drift gets cleared in a fortnight by people who were not there when the defects were raised. That sweep is where evidence goes missing, because the fastest way through it is to tick things.

The handover test

Here is how you know the register survived. At practical completion, hand it to somebody who was not on the job and ask them to answer three questions about defect 214: what was wrong, who accepted the fix, and what did they look at when they accepted it.

If they can do that without ringing you, the register works. If they cannot, it was a to-do list with a project name on it, and the next eleven months of defects liability will be run out of your memory.

If you want the shape of the finished thing before you build your own, this is what a compliance-ready report carries out the other end of a register that was kept properly from the first slab.