Technical Supplement 3
Learning Curves

LEARNING OBJECTIVES
After studying this technical supplement, you should be able to: 1. Explain the learning curve concept 2. Identify different uses of learning curves in operations management 3. Calculate the estimated time required to do a task for a given learning curve

LEARNING CURVES AND OPERATIONS MANAGEMENT

As people gain experience in doing a task, they usually can do the task more quickly. For example, consider the time it might take someone to wash a car for the first time. Then imagine how that person might be able to wash his car in less time as through repetitions he learns to sequence the tasks more efficiently or perhaps as he uses better tools to do the tasks. The same learning effect occurs in many different operational settings.

The learning curve is an analytical tool that can be used to estimate the rate at which cumulative experience allows workers to do tasks faster and with less cost. Operations managers use learning curves to estimate how much the repetitions of a task will enable them to reduce the amount of resources required to accomplish the ask. A learning curve is defined by an equation that contains the rate of improvement (i.e., reduction in costs or reduction in time taken) in performing a task as a function of the cumulative repetitions of the task.

As early as 1925, managers began developing learning curve concepts. The commander at Wright-Patterson Air Force Base in Dayton, Ohio, observed that workers performing manufacturing operations at the base exhibited a definite learning pattern. He noted that most aircraft manufacturing tasks experienced what he called an 80 percent learning rate, meaning that workers need 20 percent fewer hours to make a part each time their cumulative experience making that part doubled. Thus, if the first part took 100 minutes, the second would require 80 minutes, the fourth would require 64 minutes, and so on. Exhibit 1 graphs the reduction in unit time required to complete a task as a function of the number of times that the task is repeated when the organization has an 80 percent learning rate.

Exhibit 1
Learning Curve – 80 Percent Learning Rate
[pic]

The learning curve is also sometimes referred to as an experience curve, a progress function, or an improvement function. Essentially, the learning curve is a mathematical function that can be used to chart the progress of workers as they learn to do their work faster. An operations manager can express the relationship between the amount of time it takes an organization with a learning rate percentage of r to produce the nth item as an equation:

Tn = T1 (nb)

Where:
Tn = time required to complete the nth task r = learning rate percentage b = ln(r)/ln(2)

Example:

Consider the information given in the aircraft manufacturing example above:

T1 = 100 minutes
T2 = 80 minutes
T4 = 64 minutes

What would be the time required to produce the eighth part?

T8 = (100)(8-0.322) = 51.2 minutes since b= ln(0.80)/ln(2) = -0.322

We can verify that this is the correct answer by remembering that the fourth unit required 64 minutes. Since the eighth unit represents a doubling of output beyond the fourth unit, we would expect its task time to be 80 percent of the time required for the fourth unit. Thus, T8 = 64(0.80) = 51.2. This is the same answer given by the learning curve equation above.

Appendix A at the end of this supplement presents a table giving task time values for selected learning rates. The appendix shows that the estimated time for the eighth unit produced on an 80% learning curve is 0.512 times the task time required for the first unit. Again this confirms the result we calculated above.

How much time would be required to produce all eight parts?

Note that the table also gives the total time required to produce a cumulative number of units. In our example, the total time required to produce the first eight parts would be 5.346 X 100 = 534.6 minutes. The average time per part would be 534.6 / 8 = 66.8 minutes.

HOW OPERATIONS MANAGERS USE LEARNING CURVES

The learning curve helps operations managers make task time estimates by accounting for the fact that initial labor requirements usually do not accurately represent future requirements. Learning curves express the effects of improvements in task procedures over time that serve to reduce the time needed to execute a task. Using these curves managers can develop resource requirement estimates that may be used to financially justify the development of a new product or process. However, operations managers must evaluate learning curve expectations carefully. Expectations are based on the skill levels of workers, incentives and rewards for improvement, performance measurements, and other factors. Importantly, an organization’s culture must support learning in order to derive any benefits from the learning curve.

Selecting the proper learning rate value sometimes poses a problem. How should managers go about setting an expectation for learning? To estimate this rate for a new product, industrial engineers often try to identify similar existing products and gather historical data to compute their observed learning rates. They may also evaluate the complexity of the process, since a more complex process offers greater potential learning. In addition, they may look for outside effects on the pace of a process, such as the technology being used, the details of a labor contract, etc.

In addition to estimating what a learning rate might be, operations managers also consider what methods they might use in order to achieve a targeted learning rate. An unusually slow learning rate, perhaps around 95 percent, may indicate that the organization’s culture inhibits learning, or it could indicate that excellent production methods and training programs maximized productivity early on in the project. An unusually fast learning rate, perhaps around 75 percent, may indicate that a project began with insufficient planning and worker training, or it could indicate that highly motivated workers are straining to improve.

In general, the majority of learning rates range between 70% to 90%. Rodney Stewart and Richard Wyskida (1987)[1] give the following guidelines for expected learning rates in different situations:

• 75% hand assembly/25% machining – 80% learning • 50% hand assembly/50% machining – 85% learning • 25% hand assembly/75% machining – 90% learning or • Aerospace – 85% learning • Shipbuilding – 80-85% • Complex machine tools for new models – 75-85% • Repetitive electronics manufacturing – 90-95% • Repetitive machining or punch-press operations – 90-95% • Repetitive electronic operations – 75-85% • Repetitive welding operations – 90% • Raw materials – 93-96% • Purchased parts – 85-88%

Using learning curves creates several key benefits for operations managers:

• Improved capacity planning: Learning curves help managers know how to manage capacity throughout the life of a production project. Managers typically want to have more capacity at the start of a project. As workers become more skilled and experienced, with the same amount of capacity they can make more. Alternatively, fewer workers are needed to achieve the same level of output. This knowledge can be built into a schedule for hiring and firing personnel as well as a schedule for material usage requirements. • Improved Costing: The learning curve also tells managers how costs can be expected to fall over time. This realization has been used by some companies as a competitive weapon. Texas Instruments, for example, during the 1970s, used this knowledge to drive out many of its competitors. Managers at Texas Instruments priced products lower than the initial costs to produce them with the expectation that the lower prices would generate more demand, which, in turn, would drive down the actual costs to the point that products then became profitable. By planning on their ability to achieve a certain learning rate, Texas Instruments was able to undercut its competitors’ prices. In the same way managers can use learning curve estimates to negotiate prices with their suppliers for purchased items. • Changes in Product/Process Design: A company that understands the causes of learning effects can design its products and processes in ways that maximize the potential for learning. Some companies have been very successful at fostering learning through product design, and more importantly by developing processes that motivate and reward learning. .

SUMMARY

Almost all operational tasks involve some level of learning. In this supplement, we define learning specifically as reductions in task time that occur as a result of repetitions of the task. Such learning has been shown to be fairly steady and predictable in many operational settings. Consequently, learning curves can be used to estimate task times and associated resource requirements. Learning curves and learning analysis are important tools that help operations managers to plan for capacity needs, product pricing, and product/process design.

SOLVED PROBLEM

A firm must make a bid on a contract to make 12 units of a new product. Engineering analysis indicates that the nature of this product and its manufacturing processes resemble those for the firm’s current Model 206, so the firm decides to base its bid on the 85 percent learning rate it experienced on that model. If engineers estimate that the first unit will require 10 hours of labor, how many hours will the 12 unit require? How many hours will the firm take to make all 12 units?

Solution:
Find the unit times for an 85 percent learning rate in Appendix A.

Unit Produced Unit Time Hours per Unit Cumulative Hours Required

| 1 |1.0000 |10.0 hrs. |10.00 hrs. |
|2 |0.8500 |8.50 |18.50 |
|3 |0.7729 |7.73 |26.23 |
|4 |0.7225 |7.23 |33.45 |
|5 |0.6857 |6.86 |40.31 |
|6 |0.6570 |6.57 |46.88 |
|7 |0.6337 |6.34 |53.22 |
|8 |0.6141 |6.14 |59.36 |
|9 |0.5974 |5.97 |65.33 |
|10 |0.5828 |5.83 |71.16 |
|11 |0.5699 |5.70 |76.86 |
|12 |0.5584 |5.58 |82.44 |

This table indicates that the 12th unit will require 5.58 hours, and the total number of hours needed to make all 12 units to 82.44 hours.

We can also calculate the time required for the 12th unit directly using the learning curve equation:

b = ln(0.85)/ln(2) = -0.234

T12 = 10 (12)-0.234 = 5.58 hours

PPROBLEMS

1. A home builder needs to schedule labor for the construction of 24 homes in a subdivision. In the past the builder has noted a 90% learning rate. If the first home requires 2000 labor hours to build, estimate the time required to build: a. The 4th house b. The 15th house c. All 24 houses

2. What would the labor estimate for all 24 homes in problem 1 be if the learning rate were: a. 85% b. 80% c. 75%

3. A new worker requires 20 minutes to complete a task. If he experiences a 50% learning rate, how much time will it take him to complete the 4th task?

4. If a production team experiencing a 75% learning rate takes 10 hours to produce the 4th unit, how many hours would it be expected to take to produce the 12th unit?

5. Suppose that a consulting company typically sees an 87% learning rate for software installations at its clients’ operating sites. If the first installation required 100 labor hours, what is your estimate of the labor required for: a. The 3rd installation b. The 10th installation c. The 20th installation

6. New workers at company ABC typically learn at a 77% rate. If a new worker requires 30 minutes to produce his first unit of output, how many units will he need to produce before his output rate exceeds 10 units per hour?

7. A new product requires 50 labor hours for the first unit. If the production line operates at an 82% learning rate, what will be the average labor hours per unit required to produce 4 units?

8. New technicians performing MRI procedures in a radiology department are processing patients at rates shown below.

Patient number: 1 2 3 4 5 6 7 8 Process time (min): 75 61 52 50 45 43 40 39

a. What is the approximate learning rate? b. What is the approximate total time required to process 15 patients?

Appendix LC-A
Learning Curve Coefficients

[pic]
-----------------------
[1] Rodney D. Stewart and Richard M. Wyskida. 1987. Cost Estimator’s Reference Manual. New York, NY: Wiley.