Back in the Olde Days, when TOF was a practicing Quality Engineer, there was a popular three-step process for implementing the techniques known as "statistical process control" -- or better yet: SPC. (The benefit of TLAs is that you don't have to think about what the words mean.) The three steps were:
- The massive training of everyone in sight
- The great control chart race
- Beating up the vendors
 |
| SPC, Science™, and other panaceae |
Of special interest are the first two steps. Statistical process control -- which meant only that one should take account of the variable nature of manufacturing processes and not make adjustments unless there are reasonably certain evidences that the process has actually shifted -- was seen by some as a panacea. Convinced of the great value of this technique, they would go hog-wild and start hanging charts everywhere. Management would mandate that everyone would have charts; departments vied to see how many charts they had hung. Yee-haw!
It was belief in magic. It was "control chart-ism." If only we hang these Shewhart charts, magic will happen. Magic did happen. Whoever hung the charts eventually disappeared.
The Allegory of the VCD Machine
Once upon a time, TOF had the occasion to visit a computer assembly plant. He was proudly shown the Great Wall of Charts. See how many charts we have! It is only a matter of time before our problems go away. Huzzah!
TOF examined one of the charts. It was replete with signals of assignable causes: that is, causes not normally part of the cause-system. What was the assignable cause of this pattern?
Blank looks. Did you investigate to discover the causes?
Investigate?
These were not foolish people. They were quite bright. But they had forgotten that the operative word in "control chart" was "control," not "chart." They had confused the instrument with the objective.
A visit was paid to the process, which was currently rejecting many circuit boards. It was a VCD machine (variable-center distancing). The jaws plucked a component off a paper strip, inserted the leads through the holes in the circuit board, and crimped them on the other side. Then the circuit board re-centered two other holes under the jaws, ready for the next component off the tape. Behold its majestic glory:
(Geez, you really can find everything on the intertubes.)
TOF noticed a peculiar thing. As each board was stuffed, the operator would remove the board, place a yellow sticky arrow on it, then inspect the board. Sometimes he saw a smashed component (or whatever) and he placed a second sticky-arrow pointing to this. The number of defects was counted and plotted on a u-chart as defects per 100,000 insertions.
A question formed in the brain of TOF. Neurons crackled. Why did the operator place the first arrow
before he inspected the board?
I don't know, the Host of TOF replied.
Shall we ask the operator? The Host was reluctant that he, an Engineer, should ask a lowly Machine Operator, but his curiosity eventually triumphed. The Operator pointed to the first marked component.
That's the wrong component, he said.
Design Engineering revised the board layout and changed the component to a different rating. But no one reprogrammed the Sequencer that tacks the components in order on the paper tape, so we're still inserting the old component here.
Every time they ran that board, 100% of them had to be reworked. Reprogramming the Sequencer was boring, and so after the excitement of board re-design and Engineering Change Orders and all the rest of the foo-foo, it fell between the cracks. The Solution? Reprogram the freaking Sequencer so it plucks the right component out of the bins. Duh? Additional solution: prepare a checklist for board re-design to flag all equipments and documentation that are to be revised or adjusted or reprogrammed as a result of the change.
Despite all the charts, they had not had statistical process control, but statistical wallpaper.
Yet their enthusiasm had been undoubted. Like the person who worships the intellect instead of using it, they admired the control charts without using them for actual, you know,
control. Like cargo-cultists, they thought that by replicating the outward forms and rituals they could achieve the same results achieved by those who had mastered the substance. In effect, although the charts were crying out to them, things the chart did not show -- the specific defects involved, the lack of configuration control at the Sequencer, etc. -- had become invisible.
They thought the answers were statistical! But a thermometer only tells you you're sick. It doesn't cure the sickness. Instruments do not act on their own.
For those who might want to noodle the substance of the SPC methodology, TOF recommends the classic work: The Western Electric Company SQC Handbook, by a writing committee headed by the revered Bonnie Small, who once received a standing ovation at a conference of the American Society for Quality Control, when a speaker noticed her in the audience and pointed her out. SQC was what SPC was called in its early days, but that word "quality" kind of spooked the turkey herd.
TOF also modestly notes that when he worked summers at Mack Printing Company, this was one of the many technical publications that were then printed there. TOF's father was pressroom superintendent, but this was not, repeat not, of any benefit to TOF, who received what we might call "character-building" work assignments, many of them involving steel wool and kerosene.
Which leads us to today's topic.