How to Have a Trainee Teach It Back to You

Why this matters

You spend forty minutes walking an apprentice through a procedure, they nod in all the right places, and three weeks later they do it wrong in a way that tells you they never had the shape of it. Nodding is not understanding, and neither is watching, and neither is a correct answer to "any questions?" The only cheap test that reliably separates a person who understood from a person who followed along is making them teach it back to you, unaided, out loud, some days later. It takes about ten minutes, catches the gap while it is still free to fix, and does a second thing almost nobody uses it for: it produces the written version of the procedure your senior tech was never going to write.

Teach-back with a customer is a related tool aimed at confirming an instruction landed before you leave the driveway. This is the trainee version, and it goes deeper, because you are verifying a method rather than a message.

Step 1: Decide which of the three teach-backs you are running

They look similar and do three different jobs. Say which one you are running before you start.

The capture teach-back is the one shops never think of and the one with the most leverage. Ask a twenty-year tech to write down how they do something and you get four compressed lines that assume everything they know. Have them teach it to an apprentice, then have the apprentice teach it back while you write, and you get the version that works, because every place the apprentice hesitated is a place the card needs a step the expert never mentioned.

Step 2: Frame it so it is not a test

The framing decides whether you get a real attempt or a defensive one. Two rules:

Put the burden on your explanation, not their memory. Say: "I want to check I taught this properly. Walk me through it like I have never done it." Never "let's see if you remember." Same request, different pressure, and the second version produces a person guessing at what you want to hear.

Say what happens with the result, up front. "If there are gaps, we work them and run this again in two weeks. Nothing here goes on a review." Then honor it. The first time a teach-back result shows up in a performance conversation, every future teach-back becomes a performance, which destroys the only thing the tool was good for.

Step 3: Pick the moment, and it is not right after you taught it

A teach-back immediately after instruction measures short-term recall, which fades in days. Wait three to five weeks for a verification teach-back. That gap is not a delay, it is the measurement: supervision will be gone at week three too, and you want to know what survives to then. Consolidation teach-backs are the deliberate exception and run the same day. And if a job came up in the meantime where they used the skill, run it right after that job, because they have a fresh reference point.

Do it standing at the equipment or the bench, not at a desk. A teach-back in an office gets a description. At the equipment you get a demonstration with narration, which is better evidence and easier for the trainee.

Step 4: Ask for the sequence first, then the why

Two questions, in this order, and the order matters:

1. "Walk me through it, start to finish." Do not interrupt, do not correct, let them run out of road if they run out of road. Write down what they say and, more importantly, where they hesitate and what they skip.
2. "Now tell me why step four comes before step five." This separates understanding from memorization. Anyone can memorize a sequence. Only someone who understands it can say what breaks if you reverse two steps.

If the first is perfect and the second produces "because that's how you showed me," you have someone who can execute the standard case and will be lost the first time the job is one step off it. Not a failure, just a different remediation: fewer repetitions, more explanation of mechanism.

Where the procedure touches a real hazard - gas, energized electrical, stored pressure, stored mechanical energy, height, confined space - the isolation or verification step has to appear in their account unprompted. If you have to prompt "and what about locking it out?", the teach-back is a stop no matter how strong the rest was. Remediate the safety step on its own and re-run the whole teach-back later. Do not let a strong technical account buy a pass on a missing safety step, because the field will not.

Step 5: Go under the sequence, then out to the edge

Two probes, in that order. First, play dumb on purpose: take the role of someone who does not know the trade and ask for the layer underneath, aimed at wherever they were vague. "What does that reading actually tell you?" "What are you comparing it against?" "Say that part again like you are explaining it to a homeowner."

That last one is the sharpest question in the method. Someone who genuinely understands a mechanism can put it in plain language; someone reciting cannot, and retreats further into jargon under pressure, which is the tell. Calibrate depth to their stage - pushing a four-month apprentice to first principles produces panic and no information.

Then push to the edge with three "what if" turns, because the standard case is the one they have already seen:

• A condition that changes the answer. "Same job, but the equipment is twice the age you have worked on. What changes?"
• A reading that cannot be right. "It comes back impossible. Now what?" The right answer starts with doubting the instrument or the setup, not with acting on the number.
• A scope surprise. "You find a second problem that is not what you were sent for. Walk me through the next ten minutes."

You are not hunting the right answer to a trick question. You are checking whether their reasoning has anywhere to go once the script runs out, which is the difference between someone who can be dispatched alone and someone who cannot.

Step 6: Grade the gap, not the person, and write it down

Sort what you heard into three buckets and treat each differently. Missing steps are cheapest: more reps and a written card they keep. Wrong sequence or wrong reason means the mechanism did not land, so re-teach it with a different explanation - change the analogy, change the order, or show it failing - because repeating the original one louder has never worked. A missing safety step is an automatic stop with its own remediation.

Write one line per gap in the sign-off log. Not a narrative, a list. Reading three of those lines in sequence across three months is the clearest picture of a person's development you will get, and it takes ninety seconds to produce.

Step 7: Have them write the card

Close every teach-back the same way: the trainee writes the one-page card from their own corrected account. Their version, not yours. It forces one more pass at a precision speaking does not require, and the card ends up written at the altitude of someone who recently did not know the thing, which is exactly the altitude the next new person needs.

Worked example: a verification teach-back before a sign-off

An apprentice was taught a nine-step procedure five weeks ago and has performed it four times supervised. The lead tech runs a verification teach-back at the bench before granting a solo release for that job type.

The sequence run. The apprentice gives seven of the nine steps in the right order, with the equipment in front of them. Two are missing: the isolation-and-verify step at the start, and a comparison against a rated value at step six, which they perform but describe only as "check it looks right."

The why probe. Asked why step four precedes step five, they answer correctly and in their own words, which is a genuinely good sign and worth saying out loud. Asked what the step-six reading is compared against, they say "the number should be in range" and cannot say which range or where the figure comes from.

The call. A stop, on the first gap rather than the second. Missing the isolation step unprompted is disqualifying on its own. The apprentice performed it correctly all four supervised times, which is exactly the trap: with a supervisor present, the isolation step happens because the supervisor's presence carries it. Alone, an unnarrated step is a step that does not exist.

Remediation, split by type. The safety gap is drilled as a named, spoken step, stated out loud at the start of the next three supervised jobs of that type. The comparison-value gap is a knowledge gap, closed by writing the rated value on the card and two bench repetitions where they state the expected value before taking the reading.

Re-run at two weeks. Nine of nine steps, isolation stated first and unprompted, comparison value named with its source. The edge turns go well on two of three; the third, an impossible reading, produces "I'd replace the sensor," corrected on the spot to "verify the instrument and the connection first, then suspect the sensor." That correction goes on the card too. Sign-off granted, dated, assessor named.

Cost. About 12 minutes for the first teach-back, 40 minutes of remediation across two weeks, 10 minutes for the re-run. A bit over an hour, against a solo release that would otherwise have gone to someone who would have skipped an isolation step on their first unsupervised job.

Worked example: capture, before a senior tech leaves

A tech with 22 years is retiring in 8 months, and the matrix shows 14 skills where they are the only verified name. The shop runs one 30-minute capture session a week: the senior tech teaches one procedure to an apprentice, and the apprentice teaches it back the following week while the lead writes the card. Across 32 weeks that is 16 hours of the senior tech's time, producing 14 written cards plus a second verified name on most of them. It works because it uses the senior tech only for what they are excellent at, explaining while doing.

What a strong and a weak teach-back sound like

Weight that last row heavily. A trainee who says "I don't know, I'd call" has given you the best predictor of safe solo work there is. Reward it visibly, every time, or you will train it out of them.

What doing this wrong looks like

Running it as an interrogation. Arms folded, correcting mid-sentence, a tone that makes the trainee track your face instead of the procedure. You get a performance and learn nothing.

Prompting the safety step and then counting it. "And you'd lock it out, right?" "Yeah, of course." Now it is in the record as covered, and it was your knowledge, not theirs. Never writing the card is the quieter version of the same waste: the teach-back goes well, nothing is recorded, and six months later the knowledge is still in one head.

References

• Teach-back method, established practice in high-consequence instructional settings for verifying comprehension
• OSHA 29 CFR 1910.147(c)(7)(ii), demonstrated proficiency as the standard for verifying training
• U.S. Department of Labor, registered apprenticeship competency-based evaluation practice
• See related: The Teach-Back Method: Confirming They Actually Got It, How to Decide When Someone Is Ready to Work Alone, The Cross-Training Rotation SOP, The Toolbox Talk SOP