Warehouse simulation: Directed picking under rate pressure
Original portfolio demonstration, built so the learner practises the decision rather than reads about it.
The Challenge
Most training for a job like this takes one of two shapes, and neither changes what happens on the floor. The first is an information dump: the rules reformatted onto slides, a quiz at the end that checks whether you read them. Distribution centre inductions teach directed picking exactly this way. Scan the location, confirm the check digit, pick what the screen directs, raise a reason code on anything that doesn't match. None of it is hard. Everybody passes. The second looks more like practice, but isn't. A short story, a decision point, three options on a screen, feedback on which one was right. That's a quiz wearing a narrative. The learner is choosing between sentences, not doing the job, and those are different skills. Real picking has no menu of options. It's a gun in your hand, a rate meter running, and a decision made in about four seconds. Everybody passes the induction, and mispicks happen anyway, because the induction was never testing the thing that fails. It happens in the four seconds where holding to the rule costs something and the picker already knows both the rule and the shortcut. That's not a knowledge gap. It's a decision made under pressure, and no amount of re-explaining the rule closes it. Practising the decision does.
The Strategy
So the learner picks. Before the wave starts, a shift supervisor sets the tone: rate matters, don't slow down for exceptions. The learner answers back, staking out how they'll actually run the wave, and the wave holds them to it. Then it's Wave 1138, on the gun, with a rate meter running against the dock cut-off and dispatch checking in mid-line to ask how you're tracking. The gun runs a real procedure, not a click, and sitting alongside it is a genuine shortcut: fast, safe on a clean bin, and available right up until the moment a location is engaged properly. After that, the only way through a line is to call it straight or falsify the record on purpose. The tension isn't a button versus safety. It's whether the procedure survives contact with pressure, or gets skipped before it's even started, and that's a decision, not a click.
The Methodology
Five lines, each one loading a single judgement, not a single fact. The first is a clean baseline: scan it, confirm it, nothing under it. It proves the shortcut is real and safe before anything is asked of the learner. The second puts two identical-looking bins side by side, same product, different check digit, so the only thing separating a correct pick from a wrong one is whether the gun was actually read. The third directs a quantity the shelf can't supply, and the learner has to notice the gap and raise it rather than confirm a number that isn't there. The fourth is a visibly damaged unit, so the exception has to be seen and called, not inferred from a hint. The fifth tests whether the picker fetches the right equipment for a high pick or climbs the racking to save the walk. Grading follows the same logic as the wave itself: nothing is in it that doesn't load a judgement. Each line is scored against the decision it was built to test, so the output says which judgement held and which one didn't, rather than a percentage that quietly averages a good exception call and an unsafe shortcut into a pass. Taking the shortcut on a clean bin isn't a failure. Taking it on an exception line is, and the grading knows the difference.
The Result
The learner passes by doing exactly what the procedure asks, at the exact moment doing it costs something. Scan it, read it straight, call the exception when there is one. That's the whole test, and it's a fair one: hold the standard under pressure and the wave passes you. The picker who reaches for the shortcut on an exception line doesn't fail because the test was a trick. They fail because that's what actually happens on the floor, and the wave just made it visible before the stock did. That's what this closes: the gap between reciting a rule in an induction and holding it under rate pressure, which is the actual failure mode behind the mispick numbers. Rolled out against a cohort, the per-line data says exactly which judgement is weak across the floor, not just who passed, and that's a live measure of transfer risk before it becomes a mispick. Paired with a job aid at the gun and a supervisor watching reason-code rate, this is the front half of a system that gets integrity to survive contact with rate pressure. It earns that role because the decision it trains is the same one the floor demands, at the same speed the floor demands it.