Most safety reviews end in the wrong place. Somebody reads the findings out loud, the room nods, the document goes into a folder, and the same items show up next year with a new date on them. If a review names problems and nobody gets an owner, a deadline, or a follow-up, all you’ve produced is a meeting record.
TL;DR: Review the safety program at least once a year, then keep going. A finding with no owner and no date is only a meeting record. Assign each correction to a person, set a deadline, verify the fix in the field, and tell workers what changed and why.
A finding is not a fix
A long findings list feels good. It looks like rigor. It reads like proof somebody paid attention. And it costs almost nothing, because writing a problem down is the cheapest part of solving it.
The expensive part comes after. Somebody has to buy the equipment, rewrite the procedure, retrain the crew, change the sequence on the roof, or tell a project manager that a schedule assumption no longer holds. That work has a name on it, a cost on it, and a date on it. A findings list has none of that unless you put it there.
I judge an annual safety program review by what it produces. Sounding thorough in the room is easy. If the output is a document, the review was an audit of paperwork. If the output is a list of assigned tasks with owners and due dates, the review was leadership. Same meeting, same people, different result.
The failure mode is easy to spot once you look for it. Read last year’s review next to this year’s. Heavy overlap doesn’t mean your people can’t find hazards. Your people are good at finding hazards. It means nothing downstream of the finding ever got built.
What OSHA recommends
OSHA’s guidance on program evaluation and improvement recommends an initial evaluation of the safety and health program and, after that, evaluation at least annually. The same guidance recommends involving workers in the evaluation, correcting the shortcomings the evaluation identifies, and monitoring whether the changes made actually work.
Read that last part again, because it’s the piece most companies drop. Correcting a shortcoming and checking whether the correction worked are two separate steps, and the second one is where a program either becomes real or stays theoretical. A control that got installed and never checked is an assumption.
Annual is the floor. In construction the work changes. Crews change, subcontractors change, the scope on a reroof changes the second the tear-off exposes something nobody expected. A calendar-driven review makes a good backstop. It’s a poor substitute for looking at the program whenever the work in front of you stops matching the program you wrote.
Who belongs in the review
OSHA’s guidance recommends involving workers in the program evaluation. That recommendation exists for a practical reason, and anybody who has spent time on a jobsite already knows it: the people doing the work know which procedures get followed and which ones get worked around.
My own view on the room is simple. Get the field in it. Foremen and crew members who touch the hazard daily bring information no document produces. Bring the safety lead, because they hold the incident history and the regulatory context. Bring an operations leader who controls budget and schedule, because most corrective actions die on one of those two constraints, and a decision-maker in the room can kill that excuse on the spot.
If the only people in the review are the ones who wrote the program, you get a review of the writing. The work is a different subject.
One more thing about participation. People tell you the truth when telling you the truth is safe and useful. If a crew member raises a near miss and watches it disappear into a spreadsheet, you’ve taught the whole crew what raising something is worth. Follow-through is what keeps the input coming next year.
Turn every finding into assigned work
Here’s the standard I hold. Before the review closes, every finding carries four things: a named owner, a due date, a description of what “done” looks like, and a method of verification. Miss any of the four and what you have is a wish.
- A named owner. One person. A department can’t own anything, which is why “Operations will handle it” is how a finding stays open for three years. A name gives the question somewhere to land at the next check-in.
- A real due date, set with the owner in the room instead of emailed to them afterward. If the fix takes a quarter to procure, the date says so, and you write down what protects people in the meantime.
- A definition of done. “Improve tie-off practice” is a wish. “Anchor points installed and inspected on the two low-slope units, crews retrained on the new plan, training records filed” is a task.
- A verification method, decided in advance. Field observation, a document check, an equipment inspection, a conversation with the crew. Pick one before the work starts, so nobody negotiates the standard after the fact.
Then sequence the list by exposure. There’s a strong pull toward knocking out the cheap items first because it makes the tracker look productive. Resist it. The finding that can hurt somebody goes first even if it’s the hardest one to close, and especially if it’s the one that requires a budget conversation you’ve been avoiding.
When a permanent correction takes time, write down the interim control and assign that too. Ordering a piece of equipment protects nobody today. What the crew does between now and the delivery date is the safety measure, and it deserves the same owner, date, and verification as everything else.
Corrective action tracking doesn’t need software. It needs one visible list that gets read out loud at a recurring meeting, with the owner present and the status stated as open or closed. Nothing else has to be complicated. A tracker nobody reviews on a schedule is just a document, and we’re back where we started.
Verify the fix, then tell people what changed
OSHA’s guidance recommends monitoring whether the changes made actually work. In practice that means somebody other than the person who closed the task confirms it in the field.
An email that says “done” is a claim. Verification is looking at the anchor, watching the sequence run, reading the training records, or asking the crew to walk you through the new procedure in their own words. If they can’t describe it, the procedure exists on paper and nowhere else. Worth knowing before an incident tells you.
Then close the loop out loud. Tell the crews what the review found, what got fixed, what’s still in progress, and what you decided not to do and why. That last category matters more than people expect. Some findings get deferred for real reasons. Say so. Silence reads as indifference, and indifference is what stops workers from telling you the next thing.
This is the part that turns a compliance exercise into credibility. People don’t evaluate your safety program by the binder. They evaluate it by whether the thing they flagged in March is different in June.
Practical takeaway
If you’re running an annual safety program review this year, structure it like this.
- Before the meeting, pull last year’s review and mark which findings are still open. Pull incident and near-miss records, inspection results, and training records. Ask a few foremen and crew members what they’d change, before the group setting shapes their answer.
- Open with last year’s open items instead of this year’s findings. Anything still open gets a decision on the spot: new owner, new date, or a documented reason it’s being dropped. Don’t let it roll forward untouched a third time.
- Work through the program by area. Hazard identification, controls, training, equipment, subcontractor management, emergency procedures, reporting. Ask one question in each area: does what we wrote match what happens on the job?
- Assign as you go. Don’t batch the assignments for the end. When a finding surfaces, name the owner, set the date, define done, and pick the verification method while everyone who has an opinion is still in the room.
- Rank by exposure. Reorder the assigned list so the items with the most serious potential harm sit at the top, and add interim controls where the permanent fix has a long lead time.
- Put the tracker on a recurring agenda. Monthly works for most of these lists. Read the open items out loud with owners present. Closed means verified in the field.
- Report back to the crews. Tell them what changed, what’s coming, and what got deferred and why. Then do it again the next time something closes.
Two tests will tell you whether the review worked. Can any leader in your company name the top three open safety corrections and who owns each one, without looking anything up? And when you set next year’s review next to this year’s, how much of the list repeats?
Both tests measure the same thing: whether you assigned the problems you found.
The review is not the deliverable. The assigned work is. If your annual safety program review ends with a document instead of a task list with names and dates on it, you haven’t evaluated your program. You’ve described it.
About Khary Penebaker
Khary Penebaker is Division President at MetalMaster-RoofMaster, the Upper Midwest division of Wolkow Braker Roofing Corp. He previously built Roofed Right America from startup to $35M+ in revenue with 180 employees (2014-2025) and founded Penebaker Enterprises, growing it from $1.5M to $15M. A gun violence prevention advocate and former Everytown for Gun Safety Fellow, Khary brings two decades of leadership in commercial roofing, architectural sheet metal, and civic engagement.
Common questions
How often should a safety program be reviewed?
OSHA's guidance on program evaluation and improvement recommends an initial evaluation of the program and, after that, evaluation at least annually. In construction, treat annual as the floor. When crews, scopes, or subcontractors change enough that the written program stops matching the work, that's a reason to look again sooner.
Who should participate in the review?
OSHA's guidance recommends involving workers in the evaluation. Beyond that, I want foremen and crew members who touch the hazard daily, the safety lead who holds the incident history, and an operations leader who controls budget and schedule, since most corrective actions stall on one of those two constraints.
What should happen after an audit finding?
Every finding should leave the review with four things attached: one named owner rather than a department, a due date set with that owner present, a clear description of what done looks like, and the method you'll use to verify it. Rank the list by exposure, so the items with the most serious potential harm get worked first.
How do managers verify corrective actions?
Verification means somebody other than the person who closed the task confirms it in the field. Look at the installed control, watch the new sequence run, check the training records, or ask the crew to walk you through the procedure in their own words. An email that says done is a claim, not proof.