The Snag List That Never Gets Finished

Handover happened in March. The original list had sixty items. It is August, roughly forty have been closed, the list now has seventy on it, and three different people hold three different versions.

Nobody is being obstructive. The list is failing as a mechanism, and it fails the same way on every project.

Why lists grow after handover

Four reasons, and only the first is what most people assume.

Genuine defects appear over time. Things that only show under use, weather or occupation. Legitimate, expected, and the smallest category.

Items get re-raised because nobody recorded the fix. The work was done in April, nobody closed it off with evidence, and it comes back on the next inspection as though nothing happened.

Scope disagreements arrive disguised as defects. Something the client expected and never specified enters the list because that is the only route available for raising it after completion.

Duplicates. The same item described three ways by three inspections, counted three times, chased three times.

The middle two are usually most of the growth, and neither is a construction problem. Both are record keeping.

The list is the wrong shape

A snag list is typically a spreadsheet with a description and a status. That structure cannot hold what actually matters.

It has no evidence attached, so closing an item is an assertion rather than a demonstration. It has no ownership per item, so everything is nominally the main contractor’s. And it does not distinguish defects from variations, so commercial disputes and genuine faults sit in one column and get managed identically.

Then it gets emailed, edited and returned, and there are four versions. Nobody can say what is outstanding because the answer depends on which file you opened.

What it costs

Retention is the obvious one. Money held back against a list that will not close is money financing somebody else’s uncertainty, and for a subcontractor it is frequently the difference between a profitable job and a marginal one.

Less obvious: senior people spend months on this. A contracts manager who should be running the next project is instead attending inspections and arguing about whether item 43 was done. That is the real cost and it never appears against the job.

And it damages the relationship at exactly the wrong moment. The client’s final experience of you is six months of a list that would not close, which is what they remember when the next project is awarded.

What AI changes about closing items

Every part of this is reading, matching and evidence, at a volume nobody has time for.

AI closes items with proof. A photograph taken on site, timestamped and located, matched to the item it relates to. Closure becomes evidenced rather than asserted, which is what stops it being re-raised.

AI catches duplicates. That cracked tile to the left of the door and damaged floor finish near entrance are the same item described by two inspectors. Rules cannot see that. It is the ordinary case for a system that reads meaning.

AI separates defects from variations. Reading each item against the specification and flagging the ones that are requests rather than faults, on the day they are raised rather than in a dispute three months later.

AI keeps one version. Items raised from any source, by anybody, on a phone, land in one place. The reason there are four spreadsheets is that capture was hard, and that constraint is gone.

Agree what closed means, in writing

The single change worth making before any technology.

Most contracts are silent on what evidence closes a snag. So each party applies its own standard, the contractor believes an item is done and the client believes it is open, and both are acting in good faith.

Agree it at the start: a dated photograph, or a sign-off by a named person, or a joint inspection. Whatever it is, write it down before handover rather than negotiating it item by item afterwards.

Cut it off

A list with no end date does not end. Agree a period after which new items are handled under warranty rather than added to the snag list.

This is normal and reasonable and most contracts already provide for it. What happens in practice is that nobody invokes it, so the list stays open and retention stays held, indefinitely, because closing it requires somebody to have the conversation.

Start the list before completion

The strongest move happens months before handover and almost nobody makes it.

Most snags were created during the works and discovered at the end, which is the worst possible moment: trades have demobilised, the work is buried behind finishes, and everybody is commercially exposed. The same defect found in week twelve is a ten minute fix by somebody already on site.

Rolling inspection through the build, with the same evidence standard, turns handover into a short list rather than an event. It costs nothing extra because the inspections were happening anyway; what changes is that findings are recorded and closed as they arise instead of accumulating.

Give the client one route

Half the version problem is that clients raise items through whatever channel is nearest: an email to the site manager, a phone call to the director, a note at a site meeting, a message to somebody they met once.

Each of those enters the process at a different point, some never enter it at all, and the client experiences the ones that vanish as being ignored.

Agree one route at handover and hold it, politely, every time. Anything raised elsewhere gets acknowledged and moved into the single list, so nothing is lost and nothing is counted twice.

The three numbers to track

• Items closed per week against items added. If additions exceed closures for three weeks, the list is not converging and something structural is wrong.

• Proportion of items re-raised after closure. This is your evidence problem, measured.

• Days from practical completion to retention release. The number that turns all of this into money.

Most contractors track none of these, which is why every project feels like the first time this has happened.

AI Optimize matches site photographs to the items they close, catches the duplicates, and keeps one version everybody is working from. That work sits under Document Intake & Validation.