Peer Observation

Peer observation programmes fail in a predictable way. They start well, produce a surge of cards, settle into a monthly number that somebody reports upward, and quietly become an administrative task that finds nothing. The mechanism of that decline is worth understanding, because the programme that avoids it is one of the few things on a site that reaches work as it is actually done. This Peer Observation Toolbox Talk (Safety Talk / Tailgate Talk) is about the difference.

Here is the distinction that carries this whole talk: the conversation is the product, and the data is a by-product. Almost every failing programme has this backwards. Once the card is the point, observers optimise for filling it in — which means recording the easy, visible, uncontroversial things and never having the awkward exchange that would have produced something useful. A programme that generated no cards and a hundred honest conversations would be a success. The reverse is a filing system.

Why peer is the load-bearing word#

A supervisor observing a worker produces compliance. A worker observing a peer produces information. The difference is power.

When the person watching can affect your hours, your placement or your next job, the honest answer to "why did you do it that way?" becomes unavailable — because any answer that involves a shortcut, a missing tool or a schedule pressure is an admission. So you get the acceptable answer, the observation records nothing real, and the condition behind the behaviour survives untouched.

Between peers, that answer is available. The right tool has been broken for a fortnight. There's nowhere to anchor. We'd never finish if we did it the written way. Each of those is the actual finding, and none of them appears on a supervisor's card.

Two consequences worth stating plainly:

Supervisors should still observe — it is part of their job. But that activity is supervision, and it should not be counted as peer observation or expected to produce the same information.

The programme must be genuinely blameless in practice, not in policy. Crews judge it by what happened to the last person whose name appeared on something.

The one question that does the work#

Not what were you doing wrong. The question is: why did that make sense to you at the time?

Nobody takes a risk for no reason, and the answer to that question is almost always a condition rather than a preference. Ask it properly and you will hear the same handful of things:

• the correct tool or equipment is missing, broken or three floors away
• there is nowhere adequate to anchor, or the setup takes longer than the task
• the written method does not describe how the job is actually done
• the safe route is blocked, longer, or does not exist
• the permit or isolation process takes longer than the work
• nobody has time today, and everybody knows it

Every one of those is fixable, and none of them would appear on a card that only records the behaviour. The behaviour is the symptom you noticed; the answer to that question is the finding.

How the conversation should go#

Ask permission and pick the moment. Not mid-lift, not mid-cut, and not in front of an audience.

Start with what is going well, and mean it. Most of what anyone does is correct, and a conversation that only ever arrives with a criticism teaches people to walk away from it.

Describe rather than judge. "You were on the second rail" is a fact both people can discuss. "You were being reckless" ends the exchange.

Then ask why it made sense — and let the silence sit until it is answered.

Agree what changes, and who fixes what. Some of it is theirs; more of it is usually the site's.

Close the loop. If the answer was a broken tool, the observation has failed unless a tool appears. This is the single strongest determinant of whether the programme survives.

Keep it short. Two minutes done properly beats twenty minutes of form-filling, and it is repeatable.

The counting trap#

The most reliable way to kill a peer observation programme is to measure the number of observations.

It works exactly as you would expect. Set a target of ten cards a month and you will get ten cards a month — recording housekeeping, PPE and other easy visible items, produced quickly at the end of the month, with no conversation attached. The number will look healthy while the information goes to zero, and the crew will correctly conclude the whole thing is theatre.

Measure what changed instead. Tools replaced, anchors installed, routes unblocked, methods rewritten, processes shortened. Those are countable, they are what the programme is for, and they cannot be manufactured at a desk.

The same logic that applies to near-miss reporting applies here: quantity is easy to produce and quality is not, so quantity is the wrong target.

Where the duty sits#

There is no OSHA standard requiring peer observation — it is a voluntary programme, not a regulated one. What the standards supply is the surrounding framework:

1926.20(b)(2) — frequent and regular inspections by competent persons designated by the employer. Inspection looks at conditions; peer observation reaches the work as performed. They find different things and neither substitutes for the other.

1926.21(b)(2) — instruction in the recognition and avoidance of unsafe conditions, which is what an observer is exercising, and what the conversation reinforces on both sides.

1904.35 — the requirement for a reporting procedure that does not deter or discourage. It is written about injury reporting, but a site where reporting carries a cost will not sustain honest observation either. The two rise and fall together.

Section 5(a)(1) — recognised hazards. An observation is one of the ways a hazard becomes recognised, which is why what happens next matters legally as well as practically.

What can go wrong?#

Cards counted as the output, so the number stays healthy and the information dies.

Supervisors doing the observing, which produces the acceptable answer rather than the real one.

Names on cards, so the exercise becomes a record of individuals rather than of conditions.

Only the easy items recorded — housekeeping and PPE — while judgement, seniority and pressure never appear.

Feedback that never closes. The broken tool is reported and stays broken.

Observation as a disciplinary route, which ends the honest answer permanently.

Month-end card production with no conversation attached at all.

No time allowed for it, so it competes with production and loses.

How do we manage this properly?#

Say clearly that the conversation is the point and that the card is optional paperwork, not the deliverable.

Keep it peer-to-peer, and let supervisor observation be a separate activity with a different name.

Never attach names, and never use an observation in a disciplinary process. One breach ends the programme.

Train the question, not the form. Why did that make sense at the time? is the whole skill.

Allow time for it, because anything competing with production without protected time will lose.

Measure changes, not counts. Report tools replaced and methods fixed, never observation totals.

Feed back visibly at the toolbox talk — what came up, what changed.

Observe upward too. A programme where a labourer can observe a foreman is real; one where the flow is only downward is an inspection with a friendlier name.

Before you start#

• Confirm you know what peer observation is for here — conversation, not paperwork.
• Confirm no names are recorded and no observation feeds a disciplinary process.
• Confirm you have time allowed for it today.
• Confirm you know the question: why did that make sense at the time?
• Confirm what happened to the last thing raised through an observation.
• Confirm whether anything you would raise is actually a site condition rather than a person.
• Confirm you would be comfortable observing someone senior to you.
• Confirm who fixes the conditions that observations surface.

Talk it over#

• When someone last observed you, what did the conversation change?
• Would you tell a peer the real reason you took a shortcut? Would you tell a supervisor?
• What is the most common thing on our observation cards, and is it the most important thing here?
• What has been fixed on this site because of an observation?

The bottom line#

The conversation is the product; the data is a by-product. A programme with no cards and a hundred honest conversations is a success — the reverse is a filing system. Peer is the load-bearing word: when the observer can affect your hours or placement, the honest answer becomes an admission, so you get the acceptable answer and the condition survives. The whole skill is one question — why did that make sense to you at the time? — because nobody takes a risk for no reason, and the answer is almost always a condition: a missing or broken tool, nowhere to anchor, a written method that does not describe the real job, a blocked route, a permit process longer than the work, or no time today. The behaviour is the symptom you noticed; the answer is the finding. Avoid the counting trap: set a target for observations and you will get observations and no information, so measure what changed — tools replaced, anchors installed, routes unblocked, methods rewritten. There is no OSHA standard requiring peer observation, but 1926.20(b)(2) inspections, 1926.21(b)(2) instruction in recognising unsafe conditions and 1904.35's prohibition on procedures that deter or discourage are the framework it lives inside.

Frequently asked questions about peer observation#

What is the point of peer observation?

The conversation, not the record. The value is in reaching work as it is actually performed and finding out why a method made sense to the person using it — which is information no inspection of conditions can produce. The card is administrative; if the programme is generating paperwork without conversations, it has already failed.

Why does it have to be peer-to-peer?

Because of power. When the observer can affect someone's hours, placement or next job, any honest answer involving a shortcut, a missing tool or schedule pressure becomes an admission — so the acceptable answer gets given instead and nothing real is recorded. Between peers that answer is available, and it is where the fixable condition is.

Should supervisors observe as well?

Yes, but call it what it is. Supervisory observation is part of supervision and it has real value; it simply does not produce the same information as peer observation, and it should not be counted as though it does or expected to surface the same answers.

What is the single question to ask?

Why did that make sense to you at the time? Not what were you doing wrong. Nobody takes a risk without a reason, and the reason is usually a condition rather than a preference — which makes the answer, not the behaviour, the actual finding.

Should observations record names?

No. Once names are attached, the exercise becomes a record of individuals rather than of conditions, and the honest answer disappears. An observation should never be used in a disciplinary process either — a single instance of that ends the programme's usefulness across the site.

Why is counting observations a mistake?

Because quantity is easy to manufacture and quality is not. A target of ten cards a month reliably produces ten cards a month, recording the easiest visible items, often at month end with no conversation attached. The number looks healthy while the information goes to zero. Measure changes made instead — tools replaced, methods rewritten, hazards closed.

Does OSHA require peer observation?

No — it is a voluntary programme rather than a regulated one. The related duties are 1926.20(b)(2) on frequent and regular inspections by competent persons, 1926.21(b)(2) on instruction in recognising and avoiding unsafe conditions, and 1904.35, which requires that a reporting procedure not deter or discourage — because a site where reporting carries a cost will not sustain honest observation either.

Download the peer observation toolbox talk PDF#

Get this peer observation toolbox talk as a print-ready PDF — available in English, Spanish, Portuguese, and Turkish. Print it, hand it to the crew, and collect signatures on the included attendance sheet.

Download the PDF — free account required. New members get 5 free downloads.

Related toolbox talks#

• Common Unsafe Behaviors
• Near-Miss Reporting
• Stop Work Authority
• Human Error

Sources#

• OSHA, 29 CFR 1926.20 — General safety and health provisions: https://www.osha.gov/laws-regs/regulations/standardnumber/1926/1926.20
• OSHA, 29 CFR 1926.21 — Safety training and education: https://www.osha.gov/laws-regs/regulations/standardnumber/1926/1926.21
• OSHA, 29 CFR 1904.35 — Employee involvement: https://www.osha.gov/laws-regs/regulations/standardnumber/1904/1904.35

Written by FieldSafetyTalk's safety professional — a CSP, ASP, CHST and OSHA Authorized Outreach Trainer with 14+ years of international construction safety experience across federal, heavy civil, and industrial projects.

Hazards covered