An architect walks a site. The contractor has a question about a beam that has landed a little off the drawing, and wants to know whether it can stay where it is. The architect looks at it, thinks about the ceiling line, thinks about the light in the room below, and says yes, that's fine, carry on. The whole thing takes under a minute. It happens standing up, in dust, with a phone in one hand. By the time the car pulls out of the site, it has stopped feeling like a decision at all. It felt like a conversation.
It was a decision, though, and a fairly consequential one, made by the only person on that site qualified to make it. It just didn't arrive in a form anyone recognises. No email. No drawing revision. No moment where somebody said right, that's a change, let's write it down. It came out of the middle of a walk.
Which is odd, because a project records almost everything else. The drawings are versioned. The invoices are filed against stages. The client's approval of the layout sits in an email thread or a WhatsApp chat somewhere, retrievable if you're willing to scroll far enough. The one part of the process that leaves no trace is when the architect is physically present, looking at the real thing, and making the calls the drawings can't make. The most expert hour of the week is the least documented one.
Not because anyone is careless. Nobody skips writing down a site decision because they think it doesn't matter. They skip it because there is no natural moment for it. The visit is one continuous thing. You arrive, you walk, you answer four questions, you look at a finish, you photograph something that worries you, you have a slightly awkward conversation about a delay, and then you leave, late, already thinking about the next appointment. There is no pause in that sequence where documentation obviously belongs. And reconstructing all of it that evening, from memory, at a desk, after a full day, mostly just doesn't happen. What survives is a few photos in the camera roll and whatever the architect happens to remember.
Three months later, someone asks why the beam sits where it does.
What follows is a conversation everyone in a studio has had. The architect is fairly sure they approved it, remembers roughly the reasoning, can't place the date. The contractor remembers it slightly differently, because of course he does, he was thinking about his own sequence that morning, not about the ceiling line. Neither of them is lying. They're both working from memory about a minute that happened in the middle of a busy site a quarter of a year ago. The disagreement isn't really about the beam. It's two people rebuilding something that was never written down, and rebuilding it a little differently, which is what memory does.
That's the visible cost, and it's the one people talk about. There's a quieter one underneath it.
A site visit is work, and work has a price. Half a day of a principal's time, the travel, the fuel, the two hours of attention that went nowhere else. On a project priced upfront, with a fee agreed before anyone knew how many visits it would take, every one of those mornings comes out of a budget that was set in the dark. And because the visit leaves no record, it leaves no number either. The studio knows it went to site a lot on that job. It has a feeling about it. It can't tell you whether that was six visits or eleven, or what they cost, or whether the next project of the same type ought to be priced with that in mind. Feelings don't survive into a fee proposal. The next one gets quoted the same way as the last one.
So the visit leaks twice, once as a decision nobody wrote down and once as effort nobody counted. Both leaks run back to the same thing, which is that the visit lives in one person's head and nowhere else.
Name this in a room and the instinct is to treat it as a discipline problem. Be more rigorous. Write a site report after every visit. Send a confirming email. Keep a log. Studios try this, and it holds for about three weeks, which should tell you something.
It isn't that architects lack discipline. It's that the fix, as usually proposed, asks for a second piece of work on top of the first, at the exact point in the day with the least capacity for it. Any system that needs an architect to sit down at seven in the evening and rebuild a morning is a system that gets abandoned, and abandoning it is the rational call.
What's worth changing isn't the rigour. It's where the record lives. If capturing a site decision takes about as long as the decision itself took, in the moment, on the phone that's already in your hand, it stops competing with the rest of the day. A photo, a line of text, attached to the project and the stage it affects, while you're still standing in front of the thing. Not a report. A record. And once that exists, the second leak closes on its own, because a project that knows how many times someone went to site knows something real about what it cost to deliver, and the next fee can be a little less blind than the last one.
Site Visits inside Projectsmate logs the visit where the project already lives, so the decision sits against the stage it changed and the effort sits against the project it belonged to. Nobody has to remember anything three months later, and nobody has to rebuild the day at seven in the evening.
The judgment made standing on site was always the valuable part of the visit. It shouldn't be the only part of the project that disappears.