Case Study 2 — The Sheet on the Wall

How a piece of tape cost $340,282

Setup

Project: Northgate Outpatient Pavilion. Week 47 of a 565-calendar-day contract. Interior build-out is running on levels 1 through 3. The area: Level 3, the ambulatory surgery suite. Roughly 4,600 LF of interior metal-stud partition on that level, out of 18,600 LF in the building. The people: Ray Alvarez (senior PM), Margo Deacon (general superintendent), an assistant superintendent running level 3, Dale Whitcomb (project architect, H+P), and the drywall subcontractor's foreman. The pressure: MEP rough-in complete is a hard milestone on May 30, Year 2, and it is on the critical path. The steel acceleration in weeks 34–36 already burned most of the interior float.

(Illustrative composite. Names, numbers, and events are invented for teaching.)

The change that started it: Meridian's clinical user group revised the room adjacencies in the ambulatory surgery suite after a late operational review — two ORs swapped with a sterile core, three med-gas zone valve box locations moved, and a chase relocated. H+P issued Bulletin 12 on a Thursday afternoon, revising six sheets: A-203, A-233, A-621, M-303, P-303, and E-303.

What happened

On Tuesday of week 47, the assistant superintendent printed the level-3 partition plan — sheet A-233 at revision 2 — and taped it to the field-office wall next to the coffee maker. This is a completely ordinary thing to do. Everybody does it. It is how a field office works.

On Thursday, Bulletin 12 landed. Ray's team logged it, updated the drawing log, transmitted the six revised sheets to every affected subcontractor, and got acknowledgments back from all six. The document control was, on paper, flawless. A-233 went to revision 3.

Nobody took the Tuesday print off the wall.

Over the following nine working days, the drywall foreman laid out and framed level 3 from the taped sheet, because that is the sheet that was up, and it looked current, and nothing about a printed drawing announces that it has been superseded. His own transmitted copy of revision 3 was in a truck.

By the time anyone noticed, 3,180 linear feet of partition had been framed and roughly 40% of it boarded on one side. In-wall electrical rough, med-gas rough, and in-wall plumbing had followed the framers into 118 partitions.

The catch came the way these always do — not from paperwork, but from a body in a room. Margo walked level 3 on a Wednesday morning, stopped at what should have been the sterile core, and said, out loud, to nobody: "That's not where that goes."

The bill

Who paid

This is the part that stings.

Bulletin 12 was a genuine owner-driven change. Meridian paid for the scope of Bulletin 12 through a change order — the revised layout, the relocated zone valves, the added chase framing. That was never in dispute.

But the $340,282 was not the cost of Bulletin 12. It was the cost of building the pre-Bulletin-12 layout after Bulletin 12 had been issued, transmitted, and acknowledged in writing by Kestrel. Kestrel had notice. Kestrel had the paper. Kestrel built the wrong thing anyway.

Nadia Haddad asked Ray one question in the review: "Did we acknowledge receipt of Bulletin 12?"

"Yes."

"Then it's ours."

$340,282 came out of the GMP contingency, which was $1,320,000 at the start of the job. One document-control failure consumed 26% of the contingency for the entire project, and every dollar of unused contingency at closeout would have been split 75% to Meridian and 25% to Kestrel. Kestrel's share of that consumed contingency — the money it would have kept — was about $85,000 of pure fee.

Analysis: the four controls that would have caught it

The instinct after an event like this is to find the person. Kestrel's review found the system, which is the same lesson Bea Salgado teaches about the scaffold near-miss: the failure that matters is the one that made the individual failure possible.

Control 1 — One controlled set, stamped. One printed set in the field office marked CURRENT — CONTROL SET, dated, maintained by one named person. Every other print in the world gets stamped REFERENCE ONLY — VERIFY CURRENT REVISION at the plotter, automatically, before it ever leaves the copy room.

Would it have caught this? Yes. The taped sheet would have carried "REFERENCE ONLY" in red across the title block. Cost of the control: a rubber stamp and a plotter macro.

Control 2 — A drawing log with live revision levels, accessible in the field. Sheet number, title, current revision, issue date, source of the revision, and transmittal record. Not a binder — something a foreman can check from a phone in eleven seconds.

Would it have caught this? Only if somebody checked. It is a necessary control, not a sufficient one. Controls that depend on a person remembering to look are the weakest kind.

Control 3 — Layout verification before release. No layout is released to a trade until the field engineer verifies the sheet number and revision against the log and initials the layout drawing. Ten seconds per layout.

Would it have caught this? Yes, and this is the control Kestrel actually adopted, because it puts the check at the exact moment the document turns into physical work. The framers laid out from a sheet nobody had signed off on. Under this control, that layout does not start.

Control 4 — The revision-cloud walk. Within 24 hours of any bulletin or ASI, a field engineer walks every clouded area on the revised sheets and physically flags in-progress work inside a cloud with tape and a printed notice.

Would it have caught this? Yes, and faster than anything else — the walk happens Friday, the framing started the following Monday. This is the control that catches the case where the paperwork is perfect and the wall is already going up, and it is the one most contractors skip because it feels redundant with the transmittal.

Why the paperwork being perfect was not enough

Notice the structure of this failure. Kestrel's document control did everything it was designed to do: it received, logged, transmitted, and collected acknowledgments. Every one of those is a transaction between offices.

Not one of them touched the field. The bulletin never went to the wall where the sheet was taped, and it never went to the place where the layout line got snapped. The document-control system managed documents. It did not manage the interface between a document and a stud.

That is the general lesson, and it is worth more than the $340,282: a document is not controlled until the last person who acts on it knows its revision. Transmittals prove you sent it. Only the layout-verification step and the cloud walk prove somebody used the right one.

The uncomfortable second cause

The review also found something nobody wanted in the report. The assistant superintendent was running level 3 alone, in a stretch where the interior schedule had been compressed by the steel acceleration in weeks 34–36. He was doing three people's coordination and he did not have a field engineer assigned to him. He printed a sheet because he needed one now and walking back to the trailer for a controlled set cost him twenty minutes he did not have.

Schedule pressure did not cause this failure. It made the shortcut rational. That is the same finding as the scaffold near-miss in week 34, and it appears again in Chapter 24 and Chapter 29: when you compress a schedule, you should expect the compression to show up as document errors and safety events, not just as overtime.

Discussion questions

1. Kestrel's transmittal and acknowledgment process worked perfectly and the failure still happened. What does that tell you about the difference between a records system and a control system?
2. Rank the four controls by cost-to-implement and by likelihood-of-catching-this-specific-failure. Which one would you implement first on a job with a lean staff, and why?
3. Meridian paid for the Bulletin 12 scope change but not the rework. Construct the argument Kestrel could have made for recovery, then explain why Nadia Haddad closed the conversation with one question.
4. The review named schedule compression as a contributing cause. If you were Ray, would you put that in the written lessons-learned report knowing Nadia will read it? Argue both sides, then decide.
5. The event cost $340,282 of a $1,320,000 contingency, and Kestrel's share of the savings split was 75/25 to the owner. Explain to a first-year engineer why a contingency draw is not "free money the owner already paid for."

Your turn

Write the one-page document-control procedure you would issue on your next job. It must fit on one page, be readable by a foreman, and answer exactly five questions:

1. Where is the current set, and who owns it by name?
2. How does anyone verify in under 30 seconds whether the sheet in their hand is current?
3. What happens, procedurally, in the 24 hours after a bulletin or ASI lands?
4. Who is allowed to release a layout to a trade, and what do they check first?
5. What gets stamped on every print that is not the control set?

Then do the harder half: for each of the five, write what makes it fail. A control you cannot name the failure mode of is a control you have not thought about yet.