Field notes: the shaft that was two inches out of plumb

A worker in a hard hat walks across a freshly poured concrete slab marked with rebar stakes.

There's a version of this story on every job site, with the names changed and the tolerance adjusted up or down an inch. Someone tells it at a jobsite trailer over bad coffee, usually starting with "you believe this" and ending with a change order nobody wanted to write.

Here's the version that keeps coming up. A mid-rise core wall, elevator shaft, standard stuff. Somewhere between pour four and pour seven, the shaft walks out of plumb. Not catastrophically. Two inches over the run, the kind of drift that doesn't show up when you're standing at the base looking up, because your eye isn't a plumb bob and the shaft isn't leaning in any one dramatic direction. It's just slowly, steadily off.

Three people had a shot at catching it.

The surveyor shot control points on that level and moved on to the next one, because shooting control points was the task, not auditing the whole core against the model. The superintendent walked the floor daily, eyeballed the formwork, and the formwork looked square because formwork almost always looks square from six feet away with a hard hat on. The elevator contractor wasn't even on site yet. His first real look at that shaft would come at rail installation, months later, when two inches of drift turns into a guide rail that won't sit, a cab that scrapes, and a schedule that just lost two weeks it didn't have.

Nobody failed here, exactly, and that's the part worth sitting with. Everyone did the job they were assigned, correctly, by the letter of their scope. The surveyor shot his points. The super walked his floor. The gap lived in the seam between three scopes, and the deviation grew up quietly inside that seam for three pours before anyone had a reason to look at the shaft as a whole, continuous thing instead of three separate someone-else's-problems.

This is the structural weakness in how a lot of vertical QA still works. Everybody is watching their piece. Nobody's job is watching the whole surface against the model between the milestone checks, because that kind of continuous comparison used to mean hiring a survey crew to climb the core every few weeks, which nobody's budget survives doing on a cadence tight enough to matter. So the checking happens at milestones, and milestones are exactly where a slow drift has had the most time to compound into the expensive kind of problem.

The fix is a baseline that doesn't rely on any one trade remembering to look at someone else's work. A model to register against, and a reason to check that registration before the next pour locks the geometry in for good.

That's the actual argument for building a monthly as-built pass into the schedule instead of waiting for the as-built survey everyone assumes is still months out. Drone capture at a resolution fine enough to catch a wall that's drifting, registered against the design model, flagged the month it starts moving instead of the month the elevator contractor shows up and finds out the hard way. Early access to that kind of comparison is what As-Built Compare is building toward, specifically so the two-inch drift gets caught at pour five instead of discovered at rail install.

Two inches doesn't sound like much until it's the reason a crew is chipping out a wall that was already finished, on a floor that was already signed off, by three people who each had a perfectly good reason to assume somebody else was watching.

If that story sounds familiar, it might be worth getting on the waitlist before your shaft tells it for you.

← Back to the blog