Chapter 24 — Key Takeaways
A one-page reference card, written to be used in a trailer rather than studied in a chair. If you read nothing else from this chapter in five years, read this.
Key Takeaways
- Safety is a property of the production system, not a rulebook. Hazards are outputs of decisions — schedule, sequence, staffing, procurement, pressure — made weeks earlier by people who were not thinking about safety at all. Fix the system that produced the hazard or you have only reset the clock on the next one.
- The person who sets the schedule owns most of the safety. Your safety director controls almost none of the variables that create hazards. You do.
- Construction is roughly 1 in 20 of U.S. employment and roughly 1 in 5 of workplace fatalities. That gap is not luck. A set of hazards producing the same Focus Four every year for forty years is a stable property of how we build.
- The regulation is the floor, not the program. Know 29 CFR 1926 by subpart topic, not by section number, so you know where to look and what to ask a competent person. Roughly half the states run their own approved plans, several are stricter, and the rest of the world allocates the duty differently — check before you mobilize, not after.
- The hierarchy of controls is a ranking, not a menu. Eliminate → substitute → engineer → administer → PPE. Reach for the top, use everything you can, and notice when your first instinct was a sign or a harness.
- PPE is last because it is the only control that requires the exposed person to act correctly every time, forever, under fatigue — and it fails silently. Every control above it keeps working while everyone sleeps.
- On a multi-employer site you are usually the controlling employer, held to a standard of reasonable care, because you are the only party who can see the interfaces between twenty companies. "The sub was supposed to handle it" is a losing sentence.
- A rising near-miss count is usually good news. Reports = near-misses occurring × willingness to report. Occurrence is roughly constant; only willingness moves. A near-zero count is far more likely to be a reporting failure than a safe site.
- Paying for zero recordables buys silence, not safety — and the enforcement is peer-to-peer, not managerial. Pay for the leading indicators instead. The test of any incentive: does the number get better or worse when somebody tells the truth?
- Lagging indicators can be gamed and, on small hour counts, mean almost nothing. One recordable on a 40,000-hour job is a TRIR of 5.00; zero is 0.00. Nothing about the program changed.
- Fall arrest needs room to work. At low heights it is frequently the wrong control. Do the clearance arithmetic in writing, on the JHA, before the crew goes up.
- A JHA written in an office and handed to a crew is evidence of a system you do not have. Forty minutes on the deck, with the people who will do the work, with a "who verifies / when" column.
- The chain of "why" always terminates at the investigator's pay grade unless somebody protects it. Count your corrective actions: how many aim at the person, the tool, or the condition, and how many aim at the decision that produced them? All of the first and none of the second means you documented rather than investigated.
- Controls that change the path of least resistance hold. Controls that require a voluntary added act decay. Design against that before you write them.
- The business case is real, large, and not the reason. A job that hurts people has failed, whatever the cost report says.
Action Items — What to Do on Your Job This Week
- Print the subpart table from §24.2.3 and tape it inside the cover of your site-specific safety plan. Then look up your state: state plan or federal? What extra written programs does it require?
- Walk your job and count the covered floor and roof openings. For each: is it secured against displacement, marked, and rated for twice the imposed load? Write down who created it, who covers it, who needs it open, and who owns it while it is covered. Covers cost about $46 each. A fall through one cost ~$429,000 in case study 2.
- Ask your safety file one question: how many near-miss reports last month? If the answer is zero or near-zero on more than thirty workers, you do not have a safe site — you have a reporting failure. Then ask the diagnostic: what happens to a worker who reports one?
- Open the last twelve weekly inspection reports. Count the ones with a blank findings section. An inspection that finds nothing is an inspection nobody performed, and it is an exhibit, not a defense.
- Take one JHA off your job and read the control column. If every control is a hard hat, glasses, gloves, harness, or vest, send it back and name the four levels above PPE it never considered.
- Pick the next crew going above six feet with personal fall arrest and do the clearance arithmetic in front of them. Anchor height, free fall, deceleration, harness stretch, worker height, margin. If the number does not fit, change the control, not the lanyard.
- Pull your six-week look-ahead and pick next week's toolbox talk topic from it. Not from a rotation calendar. Then have the foreman deliver it, and stand at the back and say nothing.
- Ask every subcontractor to name their competent person for each activity — by name, with a backup and an after-hours number. Then ask the only question that matters: can this person stop the work and spend money to fix it, without calling you?
- Find the next acceleration, resequence, or manpower change above roughly $25,000 or 3 calendar days, and run the eight questions before it issues. Price the controls and add them to the option.
- Time an unannounced evacuation drill and publish the number. Northgate's first took 19 minutes for 174 people. The fourth took 6.
Common Mistakes — and the Fix
| Mistake | What it costs | The fix |
|---|---|---|
| Reaching for PPE or a sign first | A control that fails silently, plus a record proving you required it | Ask the four questions above PPE, in order, in writing, on the JHA |
| Specifying fall arrest without computing clearance | A fully harnessed worker striking the ground. 6-ft lanyard needs 17.5 ft below the anchor; a 12-ft anchor is a 5.5-ft shortfall | Do the arithmetic on the JHA. If it does not fit, use guardrails, nets, or a platform |
| Anchoring a lanyard at foot level | Free fall roughly doubles; required clearance goes to ~22.5 ft and arrest forces exceed the design limit | Anchor overhead at or above the D-ring, or use a device rated and labeled for foot-level tie-off |
| Loose plywood over an opening | It is a trap with a lid. ~$429,000 and a fractured pelvis in case study 2 | Secured against displacement, marked HOLE or COVER, rated for twice the imposed load. ~$46 each | |
| A JHA produced in the office | Theater, and evidence against you | Forty minutes on the deck with the crew, hierarchy level named for every control, "who verifies / when" filled in |
| A corporate manual with the project name on the cover | You have no site-specific plan and will not know it until the investigation | 14 sections; the test is whether a new foreman could run tomorrow's work from it |
| "The sub will provide a competent person" | A paper designation and a controlling-employer exposure | A name, an employer, a qualification basis, a backup, an after-hours number, and stated authority to stop work and spend money |
| Weekly inspections with an empty findings box | Reads as an inspection nobody performed — to a compliance officer and to a plaintiff's expert | More inspections, more findings each, with a corrective-action tracker and a 14-day closure target |
| Paying a bonus for zero recordables | Peer-enforced silence for as long as the program runs. $67,200 bought eight months of blindness in case study 2 | Same money, paid on near-miss reports, pre-task plans, hazards closed, and stop-work used correctly |
| Reading a falling near-miss count as good news | You lose your only early-warning system exactly when pressure is highest — Northgate week 34 | Treat the count as a floor, never a ceiling; a fall during a pressure period is the alarm |
| Judging a program by project TRIR over a short period | One case on 40,000 hours reads 5.00; zero reads 0.00. Northgate's functioning system carried a worse rate than a blind job | Report the rate with the hours, and manage to the leading indicators |
| Accelerating without a safety impact review | Trade stacking, out-of-sequence work, out-of-date temporary protection, and the plank on the north elevation | One page, eight questions, priced controls, signed before the directive issues |
| Pricing an acceleration without the controls | The cheapest line on the cost sheet is cheap because the controls are missing from it | Add control cost to every option, then re-rank. Line-item it in any owner-directed change |
| Unwritten recovery pressure on a subcontractor | Nobody can push back on something nobody said. It is the cheapest control on the list and the first to decay | A letter: revised sequence, means of recovery, what it costs — so they can price it or refuse it |
| Root cause written as "worker stepped on an unsecured plank" | Describes; does not explain; corrective action prevents nothing | Apply the substitution test. If a different competent person would have done the same, stop looking at people |
| Discipline following a near-miss report | One instance of compliance, and every future report from that man and everyone who hears about it | A near-miss report is never a disciplinary event. Separate the report from the conduct, in writing, in advance |
| Breaking stop-work authority the first time it is used | The promise is dead site-wide, permanently | Pay it cheerfully and publicly. Northgate's first call cost ~$9,700 and took near-miss reports from ~3/week to ~11/week |
Decision Framework
For any hazard — the order you think in
ELIMINATE? Change the sequence, prefab at grade, relocate the exposure,
│ no buy it out early. Cheapest fall protection is procured, not worn.
▼
SUBSTITUTE? Different method, different equipment, different access.
│ no
▼
ENGINEER? Guardrail, cover, barricade, netting, overhead protection,
│ no machine guard. It works while you are in a meeting.
▼
ADMINISTER? Permit, sequence, restricted hours, spotter, procedure.
│ no Costs schedule; depends entirely on enforcement at 6:40 a.m.
▼
PPE — then answer three questions:
1. Does it require the exposed person to act right every time? (yes)
2. Does it fail silently? (yes)
3. Did I honestly try the four above it? ( ? )
For any acceleration, resequence, or manpower change above ~$25,000 or 3 CD
1. Price doing nothing. (days × daily exposure — Northgate: $10,650/CD)
2. Price each option. (labor premium, equipment, supervision)
3. SAFETY IMPACT REVIEW each option — eight questions:
who works where, and when?
how many trades in one zone, before and after?
what work now happens above other work?
what hours, for how many consecutive weeks?
what temporary protections are needed longer, or removed sooner?
what goes out of sequence, and what does it assume is complete?
what controls are required, what do they cost, who buys them?
what do we watch weekly to know it is going wrong?
4. ADD the control cost to each option. ← the step everybody skips
5. Re-rank, then sign BEFORE the directive issues.
6. Put any expected subcontractor recovery in writing.
The numbers to keep in your head
| What | Carry this |
|---|---|
| Fall protection trigger — construction | 6 ft · scaffolds 10 ft · steel erection 15 ft · general industry 4 ft |
| Required clearance, 6-ft shock-absorbing lanyard | 17.5 ft below the anchor (6.0 + 3.5 + 1.0 + 5.0 + 2.0) |
| PFAS anchorage | 5,000 lb per worker, or engineered by a qualified person with a factor of ≥2 |
| Maximum arresting force, body harness | 1,800 lb |
| Excavation protective system | required at 5 ft; engineered design over 20 ft; egress within 25 ft of lateral travel; competent-person inspection before each shift and after rain |
| Overhead lines | 10 ft minimum approach for people and hand-carried conductive objects up to 50 kV; larger clearances for equipment, and the default is larger still when voltage is unknown — get the table and the voltage in writing |
| TRIR / DART | (cases × 200,000) ÷ hours worked. Northgate: 5 × 200,000 ÷ 412,000 = 2.43 |
| Report the rate with the hours | 1 case on 40,000 MH = 5.00; 0 cases = 0.00. Same job |
| EMR swing 0.78 → 1.15 on a $2.85M manual premium | $1,054,500/year, persisting three years — and a hard 1.00 cutoff is a closed door, not a price |
The first hour after something happens
1. PEOPLE FIRST. Aid, 911, the emergency action plan, the medical gate.
Somebody's whole job for the next hour is that person.
2. STOP THE EXPOSURE. Whatever hurt one can hurt the next. Be generous
with the boundary. Barricade it. Post someone.
3. PRESERVE EVIDENCE. Photograph wide to close with scale before anything
moves. Pull the JHA, PTP, tag, permit, inspection log,
daily report, schedule update, and two weeks of timecards.
4. REPORT ON THE CLOCK. Fatality, in-patient hospitalization, amputation,
loss of an eye — verify current thresholds and windows
and post them on the trailer wall today. Owner and
insurer notice are in your contract and your policy.
5. INTERVIEW EARLY, SEPARATELY. Open with: "I'm not trying to find out who
to blame." Then: "Walk me through your morning."
6. TIMELINE BEFORE THEORY. Every fact on one line, with a time and a source.
7. THEN ANALYZE. Five whys as a prompt, not a proof. Then the
substitution test. Then classify each finding
proximate / contributory / systemic — and make sure
the systemic one is on the page.
The one test for a safety program
Count the leading indicators you publish weekly, and ask of each one: does it get better or worse when somebody tells you the truth?
If every indicator improves with honesty — near-miss reports submitted, findings per inspection, corrective actions closed inside 14 days, pre-task plans completed, stop-work used and celebrated — you have a system that will surface its own problems. If your headline number improves with silence, you have bought a banner, and you will find out what it cost on a Thursday morning.