Inductions and toolbox talks that actually stick: a 15-minute check after every refresher

The induction nobody remembers

Rob runs a crew of nine. Roofing, mostly, some scaffold. No training department, no learning system, no spare hour in the week. New starters get an induction on day one: the site, the hazards, where the isolation points are, what to do if someone falls. They sign the form. The form goes in a folder.

Two weeks later, Rob asked a new starter where the nearest isolation point was. He didn't know. He'd been told. He'd signed. He'd had a lot of other things to remember that day, and that one hadn't stayed.

Rob's reaction was the right one, and it's why I'm writing this: "So what's the form for?"

Check: What did the new starter's signed induction form fail to show?

Why "told" isn't "kept"

An induction is input. A toolbox talk is input. Input is necessary and it's not enough, because most of what a person hears in a fifteen-minute talk at 6:45 on a Monday is gone by Wednesday. That isn't a character flaw. It's how memory works, and the research on it is about a hundred years old.

What makes something stay is being asked to pull it back out. Not re-reading, not re-hearing: retrieving. A person who's asked "where's the isolation point?" on Tuesday, and has to find the answer in their own head, is far more likely to know on Friday than the person who was simply told twice.

So the fix isn't a longer talk or a second signature. It's a short check, soon after, that makes everyone retrieve the handful of things the talk was actually for.

Check: Why does hearing something in a talk not mean it will be remembered, and what does make it stay?

Rob's fifteen minutes

Here's the routine Rob runs now, every Monday and after every refresher.

The talk is eight minutes, same as before. His notes for it, usually a one-page procedure with a few lines he's added, go in before the talk, with one line: "six questions for a roofing crew, scenario-based, one written". He reads what comes back on Sunday night, changes what needs changing, and it's ready.

After the talk, the crew answer the six questions on their phones, around the ute, before anyone climbs anything. About five minutes. It's marked as each person submits, with a line under each answer saying why. Rob sees everyone's result on his own phone before the first ladder goes up. Two minutes looking at it, a word with anyone who missed something, and the day starts.

Fifteen minutes. Eight of talk, five of check, two of reading the result. That's the whole system, and it fits inside the time the talk used to take plus the time people spent getting their boots on.

Check: How are Rob's fifteen minutes divided, and what happens in each part?

What a good question about a procedure looks like

The first draft Rob read had a question asking what the procedure was called. He deleted it, and the deletion is the lesson.

A procedure check isn't about the document. It's about what a person does. The questions that earn their place are scenarios:

• "You're on the north side and the wind picks up past the limit. What's the first thing you do?" Four options, three of them things people actually do and shouldn't.
• "Which of these is an isolation point on this site?" With the real ones, and one that looks like one and isn't.
• "A new starter is about to step onto the sheet without clipping on. In one sentence, what do you say and do?" Written, marked against the answer Rob wrote: Stop them, clip on first, then show them the anchor point. Not a paragraph. The thing he'd say.

Rob's rule, in his words: "If the question could be answered by someone who's never been on a roof, it's a bad question."

Check: What kind of question earns its place in a procedure check, and what is Rob's rule for a bad one?

Reading the draft with a trade hat on

Rob isn't a teacher and doesn't want to be. His reading pass is two minutes on a Sunday night with the footy on, and he changes three kinds of thing.

Site detail. The draft knows the procedure; it doesn't know this site. A question about "the designated exclusion zone" becomes a question about the one on the east side, by name.

Wrong options that are right somewhere else. A method that's fine on a commercial job and wrong on this one. The draft can't know that. Rob swaps it for the mistake he's actually seen a bloke make.

The written answer, in crew language. Whatever the draft writes as the expected answer, Rob rewrites it the way he'd say it on site, short, because that's the standard the marking follows and he wants it to match the standard he'd apply in person.

What he doesn't change: most of it. Four questions in six go through as they came.

Check: What three kinds of thing does Rob change in a drafted check?

Who to chase, by name

Before, Rob had a folder of signatures and no idea who'd actually taken anything in. Now he has a result per person, per talk, with a date.

Someone who missed the wind question gets a word before they go up. Someone who hasn't done the check at all by the next morning gets a nudge on their phone that it's waiting, and Rob doesn't have to remember to send it. Someone who got two wrong and had another go at smoko, and got six, has told Rob something useful: the talk didn't land the first time and it did the second.

Every attempt stays, with its score and date, and the crew can see their own history too. Rob's words again: "Nobody's in trouble for getting one wrong. They're in trouble for not doing it." That line, and the fact that the check is six questions and not sixteen, is why the crew do it without being chased.

Check: What does Rob now have per person that the signature folder never gave him, and how does he handle someone who hasn't done the check?

New starters: an induction that's a check, not a signature

The induction pack went in once. It's a dozen pages, and the check from it is twelve questions, the only one Rob runs that long.

Every new starter does it on day one, after the walk-around, and again at the end of week one, and again at week four. Same questions. Rob doesn't care about the first score much; a new starter has a lot coming at them. He cares about the week-one score, and he cares a lot about week four. If the isolation points aren't known by week four, the problem isn't the starter, it's the induction, and he changes the induction.

Three attempts, three dates, side by side, per starter. That's what the signature was supposed to be and never was.

Check: When do new starters sit the induction check, and which of the three results does Rob care about most?

The fifteen-minute rule

Why fifteen, and why it holds.

Eight minutes of talk is about the most anyone takes in standing up at 6:45. Five minutes of check is six questions, which is enough to cover a one-page procedure and not so many that people rush the last ones. Two minutes reading the result is enough to see who to speak to.

What doesn't fit: a second topic. Rob tried covering two procedures in one talk, and the check for the second one came back with half the crew missing it. One talk, one procedure, one check. If there are two things to cover, there are two Mondays.

Check: Why does Rob run one procedure per talk, and what happened when he tried two?

What this is, and what it isn't

Being plain about this, because it matters.

A six-question check after a toolbox talk is a reading of what the crew kept from the talk. It's useful, it's quick, and it's far more than a signature.

It isn't a judgement of whether someone can do the job. Watching a person work tells you that, and nothing on a phone replaces it. It doesn't make a site safe on its own; the procedure, the supervision and the gear do that. And a good score doesn't mean a person will do the right thing under pressure. It means they knew the right thing on Monday morning, which is where doing the right thing starts.

Rob uses it for what it is: the fastest honest answer to "did that go in?".

Check: What does the post say the check is, and name two things it says the check is not?

If you've got nothing but paper and a phone

The idea works without any software, if you're disciplined.

1. Write six scenario questions before the talk, not after. One written.
2. Print them, or read them out and have people write answers on a slip.
3. Mark them at the tailgate, out loud, together. Two minutes.
4. Keep a sheet per person: date, topic, what they missed.
5. Re-run the induction questions at week one and week four for every new starter.

What you won't get is the marking doing itself, the nudge for the person who didn't do it, the retake at smoko, or the record keeping itself. You'll get most of the retention, and you'll know within a month whether you want the rest.

Check: What are the five steps for running the check on paper?

If you're in Australia

Toolbox talks and a "take five" before the job are already normal on most sites here, which makes the fifteen-minute routine an easy fit: the talk is happening anyway, and the check slots into the minutes after it. Your obligations around inductions, supervision and records are set by your state's safety regulator and your own safety system, and this post doesn't speak to them; it's about whether the talk went in. Swap this section for your own country's site habits; nothing above depends on it.

Check: What does the post say decides a crew's obligations around inductions and records?

Try it with your own notes →