Journal/Playbook

What makes a construction defect management tool site-ready?

The demo is a clean project on wifi, driven by the person who built it. Your site is a basement with no signal and a plasterer who has never logged into anything. Nine tests to run during the demo, including the ones we do not pass.

A two-colour risograph poster of a draughting rule spanning the frame, ruled with evenly spaced tick marks, a short span of them picked out in orange and drawn long
WP
Will PrinmanCo-founder
31 August 2026·5 min read

Every defect tool demos well. The demo is a clean project, on a laptop, on wifi, driven by someone who built the thing. Your site is a half-finished basement with no signal, a supervisor holding a phone in one gloved hand, and a plasterer who has never logged into anything in his life.

The gap between those two rooms is where most of these products fail, and you can find it in about forty minutes if you know what to ask for. Here is the test I would run. It is vendor-neutral, and we do not pass all of it.

Run these during the demo, not after

Time the capture, out loud. Ask them to log a defect with a photo, a location and a description, and count the seconds. Under thirty is workable. Over a minute and your team will stop doing it properly by week three, because on a walk you are logging forty of these and the arithmetic is brutal.

Ask them to turn off wifi. Then log three defects with photos, and reconnect. Watch what happens. You are looking for whether the capture works at all, and then whether anything duplicated on the way back up. Ours used to double-log a photo when you reconnected mid-upload and it took three goes to kill properly. Anyone who tells you their offline is flawless has not run it in a lift core.

Ask who can close a defect. The answer you want is that the party doing the work cannot. If a subcontractor can mark their own item closed, every close-out number that tool produces is a measure of subcontractor optimism. Ask to see it refused, not described.

Ask what a trade has to attach. "They can add a photo" is a different product from "they cannot mark it complete without one." Make them show you the disabled button.

Make them break the location field. Ask for everything outstanding on level three. If the answer involves a text search, the tool is storing locations as descriptions and your filters will lie to you all job. You want a real tree: building, level, unit, room.

Ask what the trade's screen looks like. Not the manager view with a filter on it. The actual thing a plasterer opens. Then ask what happens if he has no account and no intention of getting one, because that is half of them. If the answer is "he creates a login and joins the project", your adoption problem is now your problem.

Ask for the history of one field. Pick a due date. Ask what it said last month and who changed it. A tool that overwrites has no answer, and overwriting is the default in most of them.

Ask who pays. Specifically: the contractor doing the rework and the contractor being backcharged are different parties, where does that live? If the answer is a comments box, it does not live anywhere.

Ask for the export. Right now, in the demo, to CSV and PDF. Look at whether the photos come with it or whether you get a column of URLs that will expire when you stop paying.

The two you cannot test in a demo

Whether the trades actually use it. No demo answers this. The proxy question is how much friction stands between a notification and the trade doing something useful. We use magic links from the email, so a trade lands on the job without an account or a password. That is not a boast about our product, it is the bar: if a subcontractor has to remember credentials to update an assignment, most of them will not, and you will be back to phone calls with a software subscription attached.

Whether it survives volume. Everything works at forty defects. Ask to see a project with two thousand, and watch the list load. Then filter it. Then export it.

What is not on this list

Dashboards. Every product has them, they all look the same in a screenshot, and nobody has ever closed a defect out of a pie chart. Judge the capture and the close-out; the reporting is downstream of both and will be fine if those are.

AI features, unless the vendor can show you the failure mode. We use it to read room names off a PDF drawing and suggest a location tree, and the useful part of that feature is not the reading, it is that nothing reaches your project until a person has reviewed the suggestions and pressed the button. Ask where the human check is. If there is not one, that is the answer.

Integration counts. A logo wall of integrations tells you nothing about whether the one you need works properly.

The honest bit

Run this test on us and you will find the edges. Correspondence is staff-only, so if you wanted your client in those threads, we do not do that. Permission overrides exist in the API but there is no screen for them yet, which means an admin has to ask us. Our mention emails do not raise an in-app notification in this version.

I would rather you find those in a demo than in month four. Any vendor who cannot name three of their own rough edges either does not know their product or is managing you, and both are expensive.

If you want to run the test against ours, everything above is on the construction defect management software page, screens included.