The Expanded and Annotated My Life and Work

How to Run a Balanced Factory at (Almost) 100% Capacity
The next section will astonish readers of Goldratt and Cox’s (1992) The 
Goal, whose matchsticks-and-dice production control exercise shows why 
it is impossible to run a balanced factory at 100% capacity. Ford’s state-
ment “… he must have every second necessary but not a single unneces-
sary second” contradicts Goldratt by suggesting a lack of significant excess 
capacity in any operation, which makes every operation a constraint.
Subdivision of tasks, which eliminates not only the nonvalue-adding 
time of exchanging tools or searching for parts, but also the associated 
variation, was the apparent key to Ford’s success as shown by equation 5.1. 
T is the total time for a series of n process steps, while σ
2
 is their variance.
 
T
t
and
k
n
k
k
n
k
=
=
=
=
∑
∑
1
2
1
2
σ
σ
 (5.1)
The n steps include not only the value-adding actions, but also those 
that do not add value. These include setup, exchange of tools, transpor-
tation, waiting for transportation, and waiting to form transfer batches. 
Single-unit processing eliminates the latter, while conveyors and work 
slides remove most of the variation associated with transportation. Task 
subdivision eliminates considerations, such as exchange of tools.
This is why: “The man who puts in a bolt does not put on the nut; the 
man who puts on the nut does not tighten it.” If the same person did both 
tasks, he or she would have to pick up and put down a wrench each time. 
Ford later elaborates on this (Chapter 10) by pointing out that elimination 
of variation in processing and material transfer times can remove almost 
all inventory from the system.

---

66  •  The Expanded and Annotated My Life and Work
The Kingman Equation (Hopp and Spearman, 2000, p. 270) supports 
Ford’s assertion. Cycle time in queue (CTQ) is proportional to u/(1-u) 
where u is the tool utilization, and this factor does indeed approach infin-
ity as u approaches 100%. CTQ is also, however, proportional to the sum 
of the squares of the coefficients of variation for (1) arrivals at the worksta-
tion and (2) effective processing time. If single-unit flow and division of 
labor can reduce both variation terms to effectively zero, CTQ will be zero 
for any utilization of less than 100%.
The subsequent paragraph adds that the work is in continuous motion, 
and there are no batch and queue activities. Batch and queue can do noth-
ing but add cycle time and variation in material transfer time to the process.
* * *
It must not be imagined, however, that all this worked out as quickly as it 
sounds. The speed of the moving work had to be carefully tried out; in the 
fly-wheel magneto we first had a speed of sixty inches per minute. That was 
too fast. Then we tried eighteen inches per minute. That was too slow. Finally 
we settled on forty-four inches per minute. The idea is that a man must not 
be hurried in his work—he must have every second necessary but not a single 
unnecessary second. We have worked out speeds for each assembly, for the 
success of the chassis assembly caused us gradually to overhaul our entire 
method of manufacturing and to put all assembling in mechanically driven 
lines. The chassis assembling line, for instance, goes at a pace of six feet per 
minute; the front axle assembly line goes at one hundred eighty-nine inches 
per minute. In the chassis assembling are forty-five separate operations or 
stations. The first men fasten four mud-guard brackets to the chassis frame; 
the motor arrives on the tenth operation and so on in detail. Some men do 
only one or two small operations, others do more. The man who places a