Skip to content
Improvement Kata and Coaching Kata | Part 4 of 6

Experiments Toward the Target Condition: One Change at a Time

What an experiment record holds, why the expected figure is written before the change is made, and the seven experiments that took one station from 84 to 68 seconds.

An experiment is the unit of work of the routine. It is deliberately small, so that a wrong assumption costs two days instead of a quarter, and so that nobody has to defend it in a meeting.

What the record holds

Five fields, written by hand at the process. The obstacle this experiment takes, from the list. The change: what will be different, who does it, from when. The assumption: why this should help, one only. The expected result, written before anything is touched. The actual result, measured the same day. And what was learned, which is the real output.

The expected figure is the part that gets skipped and the part that matters. Without an expectation there is no surprise, and without a surprise there is nothing to learn.

Seven experiments at station 4

The target condition asked for nine of ten cycles within 75 seconds at station 4, which averaged 84 seconds and ran between 52 and 141.

No. The change Expected Actual
01 Cable sets presented in sequence at the station 75 s 79 s
02 Picker reads the board at the line, not the daily plan 76 s 76 s
03 Hood screws in a magazine instead of a box 72 s 76 s
04 Second insert in the magazine for the longer screw 73 s 73 s
05 Test cable on a swivel arm at the station 71 s 71 s
06 Work sequence in the order the parts arrive 67 s 69 s
07 Locating pin for the hood of the difficult variant 69 s 68 s

Mean seconds per cycle at station 4. After the seventh experiment the longest of 60 cycles was 89 seconds and 92 percent were within 75, so the target condition was met two days before its date.

The four that did not go as expected

Four of the seven produced a different figure than expected, and those four carried the findings.

Experiment 1 was supposed to remove nine seconds of searching and removed five: the sequence at the station was right in seven of ten cases, because the picker worked from the daily plan rather than from the sequence the line ran. Nobody had named that obstacle before.

Experiment 3 changed nothing: the magazine jammed on the variant with the longer screw. The idea was right and the device was wrong, which is a cheap thing to find out in a day.

Experiment 6 removed one of two re-grips instead of both. The second belonged to the fixture, not to the work sequence, so it went on the obstacle list and became an experiment five weeks later.

Experiment 7 was better than expected: the locating pin took out the long tail as well as the mean.

The obstacle list

The list is the memory of the routine. One line per obstacle, in the words of the process, naming a thing rather than a solution. Added the moment it appears, never collected in a session. Never prioritized as a whole list, because the only question that matters is which one now. Closed with a date and a sentence, and carried over rather than cleared.

On L3 the list held 27 items after sixteen weeks, 19 of them closed, nine days on average from naming to closing. The eight that stayed open were not failures. Two sit outside the line, in painting and in the customer order pattern, and belong to the sponsor rather than to the pair.

Twenty-four experiments in sixteen weeks is one every 2.9 working days. That pace is not kept by willpower. It is kept by a conversation that happens every morning.

Newsletter

One Lean idea.
Every two weeks.