Purpose: 
To demonstrate a clear linkage between Learning and making Money in Manufacturing Companies, 
Purpose: 
To demonstrate a clear linkage between Learning and Measurement in Factories, 


also: 
To present management levers to increase the rate of making Money, 


and: 
To provide a theoretical and emperical underpinning which rationalizes (and makes manageable) the above, 


by using: 
The Learning Curve as the central theme. 
If Cost is a time series, can we fit, predict, and manage
it? In 1936, Dr. T. P. Wright published his affirmative answer:
Cost(t)=Cost(0) *
Q^{B} where: Q= Cumulative quantity, and B=Slope of the Learning Curve This remained the dominant view until Dutton and Thomas in 1984 wrote: Based on our extensive research, B is not a constant, but is based on "causal factors" which are influenced [and, therefore, can be managed  ed.] by the firm’s behavior. In 1988, Bohn suggested "manageable learning:"
Cost(t)=Cost(0) *
e^{K * q / ( E *N
)} 
Cost(t)=Cost(0) * e^{K * q / ( E *N )}

where: 
K is the product of the factors influenced by the firm 

q is the square root of the cumulative number of experiments, Q 

E is the quality of the Experimental design 

N is the total of all of the sources of Noise 


note: 
E can be metricized by the average number of experiments it takes to reveal what was being sought in each experiment. 



q, K, and N can be further broken down. 
q^{2} = Q = P * D * ( t / C
) 
where: 
q is the square root of the number of cumulative experiments, Q 

P is the degree of Parallelism ( # of coincident experiments) 

D is the experimental Duty cycle ( weekly hours experimenting / 168 ) 

C is the experimental Cycle time ( hours between successive experiments ) 

t is the elapsed time ( the same t as the parameter in the Cost function ) 
K = T * V * W * S *
F 
where: 
T is the sufficiency of the Technical knowledge and knowhow 

V is the requisite Variety of possibilities ( the probability that the answer is already in the "community" ) 

W is the receptivity of the Working environment to learning 

S is the cohesiveness of the learning team’s Spirit 

F is the Fluidity of the learning environment ( the ease with which learners can use tools ) 
C 
Shorten experimental Cycle times 
Install programmable drive systems 
N 
Reduce process and experimental Noise 
Calibrate and upgrade equipment used in tests 
E 
Improve Experimental design 
Use "herringbone analysis" and/or Toyota’s "5 Whys" 
D 
Increase experimenting Duty cycle 
Snapshot events on all shifts and weekends 
P 
Increase Parallelism 
Multiplex experimental platform 
T 
Enhance Technical knowledge 
Study Fourier and Nyquest 
V 
Increase Variety of possibilities 
Go outside your regular community 
W 
Cultivate conducive Work climate 
Inform and include those who are affected 
S 
Develop cohesive team Spirit 
Bond for Trust, Respect, and Joy 
F 
Increase learning Fluidity 
Make user friendly interfaces and 24x7 access 

Maarten Meinders, Research
Engineer DuPont's Camden, South Carolina Nylon Plant, Spinning Section Photo © 1995 by National Instruments Corporation 
Place 

Instance 
Break Reductions 
Unifying Principle 
Minimize Experimental Cycle time 
Reconfiguration 
Work on "floor," not in offices 
Performance Parameter 
Breaks per pound 
Gain in Performance 
3:1 
Cost Reduction 
> $25,000 / day 
