How to Run a Behavioral Observation Conversation That Changes the Next Task

A supervisor watches a worker reach across a moving conveyor to clear a small obstruction. The worker is experienced, the line is behind schedule, and the guard has been removed because it makes the jam harder to see. A quick observation form would record an unsafe act. A useful conversation asks why that action was the most workable option in that moment.

Behavioral observation changes safety only when it changes the next task. The goal is not to collect more cards or create a score for each employee. The goal is to understand the decision, test the control, and leave the worksite with a correction that someone will verify.

What you need before starting

Choose one task that carries meaningful exposure and that the supervisor can observe without interrupting a critical movement. Bring the current procedure, the relevant permit or job analysis, and a way to record one action owner. Do not begin with a checklist that forces every conversation into the same categories.

Andreza Araujo’s work on behavioral observation treats dialogue as a way to make the work visible. That principle is consistent with James Reason’s distinction between active actions and latent conditions. The first action you see may be real, but it is not always the most useful explanation.

Step 1: Watch the task before speaking

Spend the first few minutes watching the work as it normally occurs. Note the position of the worker, material flow, tools, pace, interruptions, and controls that are actually available. Record observable facts rather than labels. “The worker crossed the exclusion line while the load was suspended” can be checked by another person, while “the worker was careless” cannot.

Step 2: Start with a neutral description

Open with what you saw and ask the worker to describe the task. “I saw you move inside the marked area while the load was still elevated; walk me through what you were trying to do” keeps the conversation connected to the work.

A label produces a defense before the facts are understood. A neutral description does not weaken accountability. It creates enough accuracy for accountability to be fair and useful.

Step 3: Ask what made the action reasonable

Ask what the worker expected to happen, what made the action necessary, and what alternatives were available. If the normal method takes too long, ask what pressure created that tradeoff. If a barrier was in the way, ask who can change it and how the task should be performed until then.

Use questions that reveal the work design rather than questions that invite a yes or no answer. “What did you see?” and “What would make the safer method easier here?” usually produce better evidence than “Did you know the rule?”

Step 4: Test the control in the field

Compare the expected control with the control the worker could actually use. Check whether the guard, isolation point, lifting accessory, permit, communication device, or line of sight was present and functional. A control that exists only in a procedure is not controlling the exposure.

If the control was unavailable, badly positioned, or incompatible with the task, the correction must address that condition. If the control was workable and knowingly bypassed, the conversation must address the decision and the expected standard.

Step 5: Agree on the smallest useful change

Do not close with a general promise to “be more careful.” Choose one change that can affect the next task. It may be relocating a barrier, changing the sequence, adding a tool, clarifying a hand signal, stopping the job until a guard is restored, or assigning a named person to check the control.

The change should be specific enough that a colleague could verify it without asking what you meant. If it depends on a maintenance request or management decision, record that dependency instead of pretending the worker can solve it alone.

Step 6: Ask the worker to explain the new method

Invite the worker to describe how the next task will be performed. This is not a quiz. It tests whether the agreed change makes sense in the actual sequence of work. Listen for missing tools, conflicting instructions, awkward positions, and points at which production pressure may recreate the original choice.

If the worker cannot explain the method plainly, revise the control, demonstrate the sequence, or involve the person who owns the equipment. A signed form cannot compensate for a method that the team cannot use.

Step 7: Record the decision, not just the behavior

Document the observed exposure, the worker’s explanation, the control tested, the agreed change, the owner, and the verification time. Keep the record short enough that supervisors will complete it accurately. The purpose is to preserve the decision trail and make follow-up possible.

Do not turn the form into a ranking system that rewards the largest number of observations. In projects supported by Andreza Araujo, conversation quality matters because a count can rise while control effectiveness stays unchanged. The record should reveal recurring barriers, not create a target that workers learn to perform for.

Step 8: Verify the next comparable task

Return at the agreed time, preferably during the next comparable task. Check whether the change is present, usable, and understood. Ask what still makes the task difficult, because the first correction may reveal a second barrier hidden by the original workaround.

Close the loop with the worker and the control owner. If the change worked, keep it in the standard method. If it did not, revise the control and record why. A behavioral observation becomes a safety intervention only when the work is different after the conversation.

Final checklist for supervisors

• The task was observed before the supervisor formed a conclusion.
• The worker explained the decision and the conditions around it.
• The expected control was compared with the control available in the field.
• One practical change was assigned to a named owner.
• The worker explained how the next task would be performed.
• A verification time was agreed and completed.

FAQ

What is a behavioral observation conversation?

It is a short field dialogue in which a supervisor observes a task, asks the worker how the job was being done, tests the available controls, and agrees on a practical change that can be verified.

Should behavioral observations be announced in advance?

Routine observations can be announced as part of normal field coaching, while unannounced checks may test whether controls work under ordinary pressure. Neither approach should be used as a surprise test designed to catch or punish a worker.

What should a supervisor do when the worker skipped a required step?

First check whether the step was understood, available, practical, and supported by the work design. If the control was clear and workable, restate the expectation. If the control was weak, fix the system rather than recording the worker as the whole problem.

How do you know whether an observation changed behavior?

Return to the same task or a comparable one and verify the agreed change in the field. Look for evidence that the control is present, usable, and understood, rather than relying only on a completed observation form.

How many behavioral observations should a supervisor complete?

The useful number depends on exposure, task variety, and supervisor capacity. A smaller number of well-finished conversations is more valuable than a high target that rewards speed, shallow questions, or paperwork completion.

For a related supervisor routine, read how to build a safety coaching routine for new supervisors. Safety is about coming home.