The learning effect is obvious to all of us, yet still very powerful. The more human factor is embedded in the system, the more impact it has on the project outcome. On the other hand, when we start a new initiative or projects, we tend to forget about this element and treat the effects we plan to achieve in the stable state as granted from the beginning – which is simply not possible.
This is the first part of a learning curve discussion, where I will give you few hints on what learning curve is, how it originated and was applied in manufacturing and what are the assumptions, which stay behind correct application of a learning curve calculations.
The story of a learning curve starts in 1885, when psychologist Hermann Ebbinghaus tested memorizing of nonsense syllables and the results presented in a graphical form showed the output that we call today “learning curve”. In 1920s and 1930s Theodore Paul Wright, an aeronautical engineer, observed that more planes manufacturers build, the better they are and the faster work goes. He described the effects of learning and created a mathematical model for it in 1936. The principles and assumptions of this model are used still as of today.(1)
The learning curve assumptions, which have been formulated in early 20th century, apply mostly to manufacturing, but they’ve been cunsumed and adjusted by knowledge based work as learning process is present in any kind of human activity.
The learning is the most effective, when the following factors are met:
– The process which is a subject for the learning is new (or presented to the new people).
– There is significant repeatability of the tasks and they are homogenous.
– There is still a space for standardization and automation.
– People, who learn, do not rotate frequently.
There are few curves that can describe the learning effect. For our needs the basic one is sufficient (2):
The learning effect is the most significantly visible at the beginning of the process and then the curve starts to flatten.
Learning Curve Equation
The equation behind it says that the learning effect is achievable after each doubled production batch.
Y = axb
Where Y = cumulative average time per unit to produce x units
a = the time taken for the first unit of output
x = the cumulative number of units produced
b = the index of learning (log LR/log2)
LR = the learning rate as a decimal
Learning rate should be established separately for each project. E.g. for his aircraft industry T.P. Wright identified LR equal to 80%.
Importance of a learning curve
From manufacturing perspetive two important questions can be answered by this:
– What’s the time when my process becomes the most optimal (when no further learning can be applied to it)?
– How many batches of a product will be affected until learning phase ends?
At the end it all sums up to the question about product cost in relation to working hours and potential errors or rework needed.
As each of the models, also learning curve is simplified and has its limitations, especially when we would try to simply transpone it to the knowledge based work:
– First of all we never assume that learning process stops. Continuous learning and improvement are embeded in the very basic principles and practices that we follow.
– The work is rarely homogenous and tasks are not 100% repeatable between the projects. We rather observe the type of work, which shows some similarities – but not the particular tasks.
– The learning process applies also before the tasks are finished (we don’t need to complete the batch of work to be able to observe this).
Although the learning effect is something we cannot avoid, I often see that this element is skipped when planning a project or forecasting the work. We tend to translate the metrics and statistics from the stable state to the initial phases, when team is still formulating. Similar situation applies to the new person in a role, who needs to learn not only her place in the project, but very often – new responsibilities.
Let’s keep it in mind, when planning the work or identifying the impediments resons.
Having the theoretical introduction to the learning curve effect, we can move to the second part of a discussion, which will be focused on what to look for when thinking about learning process and how to apply it to your current and future projects. I will also show you how the learning curve is related to Tuckman model, PDCA and continuous learning.
(1) & (2) Learning curve history and graphical representation by: https://en.wikipedia.org/wiki/Learning_curve.