My friend John no longer drives a truck

 I ran into an old friend at one of the big-box home improvement stores today.

I asked what he was doing now and he told me that he was no longer driving a truck

John had been driving a truck and repairing/replacing refrigerated food display units for the last 20 years. Typically, he drove a large, white utility van from service-call-to-service-call.

A couple of months ago he was called into the office at the end of the shift. That is never a good thing.

The boss, an office-weenie who had never worked in the field, informed John that HQ had been informed by the sensors in his van that he had crossed the yellow-line on M-43 at 10:37 AM and had driven that way for over a quarter-mile. Then he asked "Is there anything you want to tell me before I hand you your automatic write-up for unsafe driving?

John explained that he was in an area with many warehouses and heavy semi traffic. John said that there was a semi in the left lane ahead of him. That truck was crowding his lane. John used to drive a semi. He knew that some drivers would turn from one lane out if they thought the lane next to them was empty.

Michigan. Snow. White van.

There was also the issue of wind buffet. John's van was almost empty and it was a big sail. So John went a little bit over the yellow line after checking to ensure it was empty.

"Don't let it happen again" his boss told him as he handed over the discipline form.

John complied by finding another job. That will NOT happen again.

Editorializing

I used to work in an automobile factory where car bodies were measured with laser-cameras. Roughly speaking, each vehicle was measured in 400 places through the course of it manufacture.

One of the local universities attached a "high potential" graduate student with us to teach us the errors of our ways.

The first thing she did was to pull out a copy of Shewhart's 14 Rules for Statistical Process Control.

Test	Rule	Problem indicated
1	1 point is outside the control limits.	A large shift.
2	8/9 points on the same size of the center line.	A small sustained shift.
3	6 consecutive points are steadily increasing or decreasing.	A trend or drift up or down.
4	14 consecutive points are alternating up and down.	Non-random systematic variation.
5	2 out of 3 consecutive points are more than 2 sigmas from the center line in the same direction.	A medium shift.
6	4 out of 5 consecutive points are more than 1 sigma from the center line in the same direction.	A small shift.
7	15 consecutive points are within 1 sigma of the center line.	Stratification.
8	8 consecutive points on either side of the center line with not within 1 sigma.	A mixture pattern.

Eight of Shewhart's SPC Rules

I informed her that those rules, as written, were not appropriate for an environment where you were not using "statistical sampling" but were measuring every part. I could see her eyes glaze over when I suggested that she take 24 hours worth of data, dump it into Excel and apply those rules to see how many times we would stop production.

The usual stress and conflict happened. She went to my boss's boss and accused me of being sexist and of being an obstacle to progress. Ron dodged the bullet. "Why don't you just observe for a week before you tell Joe to completely change how we run the shop?"

She came back and chewed my ass. "These rules were developed by people a LOT smarter than you. Implement them."

"I am not arguing about who is smarter. What I am arguing is that rules that are sensitive enough to detect process issues when you measure one in every sixty or one in every five-hundred parts are going to be WAY too sensitive when you measure every part in ten-times as many places" I said.

My friend John got caught by the same trap.

John's management is now getting data on a foot-by-foot basis. They are measuring position in the lane, speed, turn indicator use, seat-belt use, time at each stop, path deviations from GPS indicated and recording dash-cam footage for download by WIFI when the truck is parked in the barn.

From an automotive production standpoint, it was necessary to de-tune the rules to make them LESS sensitive because there is a limited amount of people who can respond to production shutdowns. The art of process control is to figure out which things to detune and how much less sensitive to make them.

It really comes down to matching up the number of automatic alarms and your ability to respond to those alarms. In the case of John's management, it may come down to having enough drivers/equipment maintenance people to stay in business.

In another case, in a different factory, a digital camera was used to measure the location of a robotically applied drawn-arc weld stud on the cab of a truck. It faulted out about every third truck and the team-member two stations away would look, see the stud and then "mail" the truck downline. 

This went on for months. The Union complained. The issue was put on a list. One of the high-end programming engineers determined that the tolerance of +/- 1/2mm was part of the problem. The other problem is that the software in the camera was cutting a plane through the threaded stud and it was measuring the outline of the threads. moving the plane half-a-thread could cause the measured location to jump 1.5mm by moving the valley from one side of the stud to the other.

The engineer "fixed" the problem by increasing the tolerance and by spinning the lens out-of-focus so it could not optically "see" the individual threads.

This problem is not any different than the proliferation of smartphones, video footage, body-cams and hot-mics. Good people can make poor choices of words, poor choices of associates. The rational response is to detune the system and look for patterns of behavior while the entire social-media outrage industry moves in the other direction.