What it does to the people, and what the people cost

The sentence everybody says, and what it hides

The sentence is automation replaces people. In a garment factory it is usually false as stated, and usually true in a form nobody says out loud. So it is worth doing the arithmetic rather than the argument.

Take the cell from lesson 3 and count positions rather than money.

work removed    4.32 standard minutes at 62% = 6.97 attended minutes a garment
                6.97 x 78,000 = 543,660 attended minutes a year
                543,660 / 110,450 = 4.92 operator positions
work created    0.30 attended minutes a garment x 78,000 = 23,400 = 0.21 positions
                plus half of one mechatronics technician  = 0.50
net                                                        = minus 4.21 positions

4.92 positions of sewing-room work stopped existing. That number is real and it should not be softened. What happened next is the part that is usually reported wrongly in both directions.

What actually happened at Bilecik

Nobody was made redundant. On the day the cell was commissioned, Akbulut had 14 unfilled seats on the sewing floor, and the four operators whose preparation work had gone were moved onto operations the factory had been short of for months. Headcount did not fall. Output rose, because four seats that had been empty were now filled.

This is the commonest real outcome, and it is neither the reassuring story nor the frightening one. The reassuring story says nobody loses anything, which is false: 4.92 positions of work disappeared, and if the factory had not been short-staffed, four people would have had to go somewhere. The frightening story says four people lost their jobs, which is also false here. What is true in both cases is the same sentence: the work disappeared, and where the people went depends on what else the factory had for them to do. A factory that automates while it is growing redeploys. A factory that automates while it is shrinking does not, and it should say so honestly when it asks for the money, because the two cases carry very different obligations.

The wage moves with the skill, and it is not a small movement

Half a position of the work created is not a sewing job. It is a mechatronics technician who can read an alarm log, teach a placement offset again and diagnose a vacuum fault, and he is priced by a labour market that does not overlap with the sewing floor.

The technician costs 2.90 times a sewing operator. That ratio is the single most useful line in this lesson, because a case built on the assumption that the labour created costs the same as the labour removed is wrong by a factor of nearly three on the created side.

The middle row matters too, and it is the one people forget. Sevim Duran did not give the cell to whoever was nearest. She graded the seat up. The person who tends the cell is responsible for what comes off it, has to notice a drifting placement before the auditor does, and stops the machine when it is wrong. That is a different job from the one she was doing, and it is paid 22.2% more — USD 1,988.10 a year for one seat. The skill and the wage move together. A factory that automates without moving the wage finds within a year that its best operator has stopped wanting the machine seat, and that is the beginning of every availability problem in lesson 6.

The month the technician left

Kadir Demirkol was hired for the cell. Nineteen months later he left, for a plant in another sector that wanted the certificate the cell had given him.

Replacement took 11 weeks, 55 working days. Over that period the cell's availability fell from its measured 91.3% to 61.0%, because the daily checks were done by a general mechanic who could restart it but could not teach it again.

availability lost    30.3 points x 1,880 hours x 55/235 days = 133.32 machine-hours
work displaced       133.32 x 60 / 1.16 = 6,896 garments prepared by hand
benefit forgone      6,896 x USD 0.5459 = USD 3,764.53
recruitment                                USD 2,400.00
                                           USD 6,164.53

That is before the replacement's own learning period, which Aydemir has not costed and which is not zero.

Demirkol had asked for USD 2,800 a year, eight months earlier, and had been refused on the grounds that it was out of line with the plant's grades. Losing him cost 2.20 times one year of the raise he asked for, and it cost it inside a single quarter.

A cell is not a machine plus a person. It is a machine that only works while a specific person is still there. The cost of that dependency belongs in the case at the point where the machine is bought, not in an incident report eighteen months later. Lesson 6 turns this into two things you can actually write down: a second person trained on the cell, and a written handover that survives the first one leaving.

The same machine, in a different labour market

Everything in lesson 3's arithmetic scales with the wage, because what the cell sells is labour it does not use. Hold the machine price, the cycle time, the style count and the volume exactly as they are at Bilecik, and vary only the loaded rate — remembering that the technician's rate moves with it, because he is hired in the same market.

At USD 5.72 an hour the same cell, running the same 19 new styles a year, turns positive. At Bilecik's USD 4.86 it does not. In the lowest-wage row it is not close.

That is the second axis of this course, and it deserves stating plainly, because it turns a very widely held idea upside down. Automation is not what a low-wage factory does to compete. It is what a higher-wage factory does to survive. The machine is priced in dollars and the benefit is priced in local wages, so the cheaper the labour, the worse the case. That is why the automation you actually find on a factory visit follows the wage level far more closely than it follows the size of the company or the modernity of the buyer.