Case Study 24-2 — 247 Days: The Job with the Perfect Numbers

Kestrel Construction Group, Rivermont Elementary School #12, and every person named here are Tier-3 illustrative composites assembled from real projects. The regulatory framework is real. The people are not.

Setup

Case study 1 was a job that found its own problem and paid to fix it. This one is the opposite, and it is the more important of the two, because nothing in it looks wrong until the end.

The project. Rivermont Elementary School #12. $22,400,000, hard-bid design-bid-build lump sum for the Rivermont Unified School District. Prevailing wage, certified payroll, 100% payment and performance bonds. A 78,000-square-foot, two-story school — masonry and steel, a gymnasium, a cafeteria with a full commercial kitchen, and a two-story classroom wing on each side. Kestrel's anticipated fee: about $784,000.

The project manager. Curtis Boone. Fifteen years with Kestrel. He wins work by buying it low, he runs a lean staff, he manages by relationship, and everybody on that site — including me — likes him. He is not a villain and this case does not work if you read him as one. If you want the version of Curtis who cuts corners on money, it is in Chapter 13 and Chapter 16. This chapter needs a different Curtis: the one who spent his own general conditions money on a safety program because he genuinely wanted nobody hurt on his job.

The field engineer. Terrence Bell. Twenty-three, second job. He writes his daily reports at 4:30 in the afternoon and he means every word of them. Remember that.

Here is what the numbers looked like in month twelve.

Read that table again. Curtis Boone's job had a better recordable rate than mine, a perfect DART rate, and 247 days on the banner at the gate. Against a construction industry recordable rate typically quoted in the low-to-mid twos — go get the current Bureau of Labor Statistics figure rather than trusting my memory of it — he was running the best-looking job in the company.

Nine days after that table was printed, a man fell thirteen and a half feet onto a concrete slab.

What Happens

Month 4 — the sting that started it

A carpenter opened his hand on a stud track and got four sutures at an urgent care. Treatment beyond first aid by a licensed health care professional: recordable. On 34,000 hours to date, that single case produced a TRIR of 5.88 — which is exactly the arithmetic problem §24.9 warns you about, and which Curtis had never had explained to him.

The district's construction manager asked about the 5.88 in the monthly owner meeting. Politely. Once. Curtis went back to his trailer genuinely stung, because he had never had a hurt man on a job and did not intend to start.

Month 5 — the program

Curtis designed it himself over a weekend and funded it out of his own general conditions, which on a hard-bid lump sum is his own money in the most direct sense there is.

• $100 per craft worker per month, every month the project records zero recordable injuries. Everybody on site — Kestrel's people and every subcontractor's people, no distinctions.
• A catered lunch at 90, 180, and 270 days.
• A banner at the gate: ___ DAYS WITHOUT A RECORDABLE INJURY, in numerals a foot high, updated every morning by whoever opened the gate.

Month 5 is also when Curtis started supplementing the mechanical subcontractor he had bought at $1,975,000 after shopping a $2,140,000 number to the fourth bidder (Chapter 16). They were underfunded and had manned the job with five people where the schedule needed eleven, so Curtis started buying bodies to keep them moving. He spent $287,000 doing it over the balance of the job.

Hold those two paragraphs next to each other. They are the same month, they are on the same page of the same cost report, and nobody at Kestrel — me included — ever noticed that one of them was a safety decision.

Months 5 through 12 — 247 days

Zero recordables. Zero DART cases. Zero near-miss reports.

Not "few." Zero, across roughly 78 craft workers on site, for eight months.

Month 12 — the operations meeting

Curtis presented. He had a slide. He was proud of it and he had earned the right to be proud of something, because he thought he had bought a safe job for $67,200.

Nadia Haddad asked one question.

Nadia: "How many near-misses?"

Curtis: "Zero. Eight months."

Nadia: "Huh."

And then we moved to the next agenda item.

I was in that room. I heard the question and I heard the answer and I understood, in some barely-conscious way, that "zero near-misses across seventy-eight workers for eight months" is not a number that can be true. And I said nothing, because it was not my job, and because Curtis was having a good day, and because the meeting was already running long.

§24.11 asks you to write down what you would have said in the conference room on the Monday after Milo's near-miss. This is my second answer to that question, and it is the worse of the two — because in the conference room I was at least under pressure and defending my own decision. In this room I was drinking coffee.

Month 13, Thursday, 9:40 a.m.

Dwayne Pickett, 34, six years in the trade, fourteen months on this job, a ceiling installer for Kettleman Interiors, walked down the second-floor corridor of the east classroom wing carrying a bundle of ceiling main tees on his right shoulder.

At column line 6 there was a mechanical chase blockout in the deck — 3 feet by 4 feet, formed when the deck was poured in month 6, covered ever since with a piece of ¾-inch plywood cut on site. The plywood was not secured against displacement. It was not marked. Nobody had calculated whether it would carry twice the load that could be imposed on it, because nobody had ever thought of it as a cover. It was a piece of plywood over a hole, and it had been walked on, by Dwayne's own estimate afterward, "a couple hundred times."

Three days earlier, a mechanical crew — supplemented bodies, on site four days, working for a subcontractor who was $165,000 underfunded — had pulled the plywood to drop a duct riser through the chase. They ran the riser. They put the plywood back. They put it back about six inches off, so that one corner bore on a duct flange instead of on the deck.

Dwayne stepped on that corner. It tipped. He went through and landed on the first-floor slab, 13 feet 6 inches below, with the bundle of main tees.

Fractured pelvis. Fractured left wrist. Surgery that night. Six days in-patient. Thirty-one weeks off work, and he did not come back to the trade.

An in-patient hospitalization is reportable to OSHA on a short clock. Kestrel reported it. A compliance officer was on site the following morning.

That was day 248.

What Bea found

Kestrel has no safety staff on hard-bid schools; Bea Salgado covers eleven jobs from corporate. She went to Rivermont for three weeks and interviewed forty-one people over nine days.

Nine near-misses had occurred in the previous eight months and none of them had been reported. Three of the nine were on that exact cover:

And a foreman told Bea, without prompting and without embarrassment, that in month 7 he had talked a worker out of going to a clinic for a hand laceration. His explanation: "We were at sixty-eight days."

The sentence Bea wrote in her report, quoting a Kettleman carpenter verbatim, is the one I want you to keep:

"I'm not going to be the guy that costs seventy-eight people a hundred bucks."

That is not management pressure. Nobody at Kestrel ever told anyone not to report anything. That is $7,800 a month, riding on nobody speaking, enforced by the crew on itself, exactly as §24.12 describes.

And what was in the file the whole time

Here is the part that still bothers me most.

Terrence wrote every one of those fourteen entries. He wrote them because he writes his daily reports at 4:30 and means every word of them. He filed them. They went into the project record, contemporaneously, dated, in the exact form Chapter 26 tells you to write them.

Nobody had ever told anybody that those entries were a trigger. They were just words in a daily report, the same way "unsuitable material" and "holding for direction" were just words in a daily report in Chapter 5's case study, on this same job, two years of narrative earlier. It is the identical failure. Twice. On Curtis Boone's project. And the second time it cost a man his trade.

💰 The Arithmetic

What the program cost, and what Curtis counted as its return

What the event cost

Illustrative composite figures. Workers' compensation medical costs, indemnity rates, disability schedules, and permanent-impairment awards vary enormously by state and by year — these are this book's numbers for this book's project, not a benchmark.

Three things are deliberately missing from that table.

The citation. OSHA cited. I am not printing the penalty, because maximums are inflation-indexed and mine would be wrong by the time you read this — and because, as the ledger shows, the penalty was by a wide margin the smallest number on the page. Nobody believes that until they see it laid out. Under the multi-employer framework (§24.2.6) the mechanical subcontractor was the creating employer, Kettleman Interiors the exposing employer, and Kestrel the controlling employer, held to a standard of reasonable care. Kestrel's own file — 31 signed inspections, 27 finding nothing, alongside 14 daily-report entries naming the hazard — did not help. An inspection record with no findings is not a defense. It is an exhibit.

Liquidated damages. Left out so you can see the operating cost by itself.

The mod. One claim of roughly $310,000 does not by itself move an experience modification rate from 0.78 to 1.15. But the rating formulas weight the primary portion of every claim heavily, the experience period runs three years, and Kestrel's annual workers' compensation manual premium is $2,850,000 (§24.9) — so ten points of mod is $285,000 a year, $855,000 over the period. I will not compute Kestrel's actual mod, because the split point, the expected losses, the ballast, and the credibility weighting belong to a rating bureau and differ by state and by year. Carry the direction and the duration: up, and for three years, at a company that bids healthcare work behind a hard 1.00 prequalification cutoff.

The comparison, three ways

Curtis spent thirty-six times the cost of the physical fix on a program whose net effect was to make the hazard invisible for eight months. And any one of those three unreported near-misses on that specific cover — reported, written down, closed — would have cost $46 of plywood, screws, and stencil paint, and about twenty minutes.

Analysis

1. The lagging numbers were correct, and they were useless. Nobody falsified anything. One recordable in 112,000 hours really is a TRIR of 1.79. Then one event took it to 3.57 and the DART rate from 0.00 to 1.79 in four seconds, on a Thursday morning, driven by causes that had been in the project record for eight months. Now look back at the setup table: Northgate, with a functioning system, carried a worse recordable rate than Rivermont did with none. Both are arithmetically correct. Both are noise. That is not a criticism of TRIR — it is what a rate built on single-digit counts is, and it is why an owner prequalifying on twelve months of project TRIR is measuring luck and calling it management.

2. The organization was blind by design, and the design was a spreadsheet. From §24.9: reports observed = near-misses occurring × willingness to report. Occurrence is a property of the work and it was high — Bea found nine in eight months just by asking, and those are only the ones people remembered. Willingness was driven to approximately zero by a $100 monthly transfer that turned every worker's report into a $7,800 cost to his coworkers. The program did not buy safe behavior, because safe behavior was not what it measured. It bought the absence of a recorded injury — something a worker can control directly, and a hazard is not.

3. The hazard was manufactured in month one, in a buyout meeting, by a man thinking about money. Follow it: shop the mechanical package $2,140,000 → $1,975,000 → the subcontractor is underfunded → they man five where eleven are needed → Curtis supplements with $287,000 of rotating bodies from month 5 → in month 13 a four-day-old crew with real trade skill and no site knowledge pulls a plywood cover nobody has ever told them anything about and sets it back six inches off. Nobody in that chain was careless. This is the chapter's threshold concept applied to a procurement decision instead of a schedule decision, and it is the same concept: the hazard was an output of a decision made eleven months earlier by somebody not thinking about safety at all.

4. Everything that looked like diligence was an artifact that produces no information. Rivermont Elementary had a current 250-page corporate safety manual, a documented weekly inspection every week without a miss, a Monday toolbox talk with a signed sign-in sheet, contract-required OSHA 10 cards, 100% hard hat and vest compliance Curtis personally enforced, a project manager who walked the job daily and knew every worker's name, no prior citations, and a funded safety incentive program. That job passes a prequalification questionnaire with distinction. And every one of those artifacts is a record of an activity, not a measure of a condition. Not one of them could have told anybody that a plywood cover in the east wing moved.

5. It felt like diligence, and that is what makes it worse than negligence. Negligence has a shape you can find. A superintendent who walks past a hazard knows, at some level, what he did — that is my mechanical curb in §24.7. Curtis never walked past anything. He read his instruments every month, they said the job was safe, and they were measuring the wrong thing. You cannot fix a problem your dashboard is designed not to show you, and you will not go looking, because the dashboard is green.

6. The rhyme with case study 1, which is the reason these two are in the same chapter. Strip both events to the physical mechanism and they are the same event: a trade moved something it did not own and put it back approximately. A plank on the north elevation of Northgate. A plywood cover in the east wing at Rivermont. Same species. The difference was never the plank or the plywood.

The difference is that one organization had built a way for bad news to travel and the other had, entirely by accident and with the best of intentions, paid people not to send it.

Discussion Questions

1. Curtis funded the program out of his own general conditions on a lump-sum job, so every dollar came off his own fee. Explain, in mechanism rather than moral terms, how a program funded out of genuine concern produced a worse information environment than no program at all. Then the harder version: what would you have had to say to him in month 5 to stop him, and would it have worked?

2. Nadia asked "How many near-misses?", heard "Zero, eight months," said "Huh," and moved on. Ray was in the room and said nothing. Design the mechanism that catches this without depending on anyone in that room being braver than they actually were. Name where the trigger lives, who owns it, and what it forces to happen.

3. Fourteen daily-report entries over eight months named the hazard, six of them naming the exact location. The identical failure — real information in a contemporaneous record that nobody had defined as a trigger — happened before on this same project in Chapter 5. Write the rule that converts a daily-report entry into an action: which words, whose inbox, what clock, and what closes it. Then estimate the false positives per month and decide whether you can live with that number.

4. Re-spend the $67,200, keeping the total identical and the money in craft workers' pockets. Which leading indicators do you pay for, at what rate per worker, and how does each one get better when somebody tells the truth? Then write the four sentences the superintendent says at the all-hands — including the one that acknowledges what the old program did to the man who wanted to report something.

5. Kestrel's 31 signed inspections with 27 "Findings: none" entries hurt rather than helped its controlling-employer position. Rewrite the inspection form so a truthful inspection with genuinely nothing to report is distinguishable from an inspection nobody performed. What fields does it need — and what does the superintendent get out of filling them in?

Your Turn

Two tasks, and the first one takes fifteen minutes.

One — the openings audit. Walk a job you can get on, or use the Willow Street Community Center package in Appendix K, and count every floor and roof penetration that will exist between structure and closeout: mechanical and plumbing chases, electrical risers, floor sleeves, stair and elevator openings, roof hatches and curbs, skylights, and every blockout on the structural drawings. For each one write four things: who creates it, who covers it, who needs it open, and who owns it while it is covered. Then price the covers — secured against displacement, marked, and rated for at least twice the imposed load — at a defensible unit rate you state.

Compare that total to $428,954.

Two — the blindness test. Take any organization you are part of and find the number that everyone watches. Then answer three questions in writing:

• What does that number get better at when someone tells an uncomfortable truth — and what does it get worse at?
• Who has more control over the number than over the thing the number is supposed to represent?
• If the situation the number measures got quietly worse for eight months, what would you see?

If the honest answer to the third question is "nothing," you have found a 247-day banner. It does not have to be about safety and it usually is not. But this is the version of it that puts somebody in an ambulance.