Learning curves graphically portray the costs and benefits of experience when performing routine or repetitive tasks. Also known as experience curves, cost curves, efficiency curves, and productivity curves, they illustrate how the cost per unit of output decreases over time as the result of learning and experience. That is, as cumulative output increases, learning and experience cause the cost per unit to decrease. Experience and learning curves are used by businesses in production planning, cost forecasting, and setting delivery schedules, among other applications.

Learning curves are geometric curves that can be graphed on the basis of a formula. Typically the X (horizontal) axis measures cumulative output, and the Y (vertical) axis measures the cost per unit. The curve starts with a high cost per unit at the beginning of output, decreases quickly at first, then levels out as cumulative output increases. The slope of the learning curve is an indication of the rate at which learning becomes transformed into cost savings.

An 80 percent learning curve is standard for many activities and is sometimes used as an average in cost forecasting and production planning. An 80 percent learning curve means that, for every doubling of output, the cost of new output is 80 percent of prior output. As output doubles from one unit to two units to four units, etc., the learning curve descends quite sharply as costs decrease dramatically. As output increases, it takes longer to double previous output, and the learning curve flattens out. Thus, costs decrease at a slower pace when cumulative output is higher.

One can explain the shape of learning curves another way. When a new task or production operation begins, a person or system learns quickly, and the learning curve is steep. With each additional repetition, less learning occurs and the curve flattens out. At the beginning of production or learning, individuals or systems are said to be "high" on the learning curve. That means that costs per unit are high, and cumulative output is low. Individuals and systems "move down" the experience or learning curve by learning to complete repetitive tasks more efficiently, eliminating hesitation and mistakes, automating certain tasks, and making adjustments to procedures or systems.

Some theorists believe that learning curves are not actually curves, but more like jagged lines that follow a curving pattern. They assert that learning occurs in brief spurts of progress, followed by small fallbacks to previous levels, rather than in a smooth progressive curve. Such a model of learning, however, does not affect the usefulness of learning curves in business and production applications.

System changes that affect how tasks are accomplished cause disruptions in learning curves. Every change has costs and, presumably, benefits that can be graphed by overlaying the learning curve based on the new way of doing things with the old learning curve. When the change is introduced, a new learning curve starts at a point above the current learning curve. The new learning curve soon intersects the old learning curve as cumulative output increases. The graphic area above the old learning curve and under the new learning curve represents costs associated with the change. Once the new learning curve intersects the old learning curve, the area under the old learning curve and above the new learning curve represents benefits associated with the change. Optimally, no new changes would be introduced until change benefits exceed change costs.

[ David P. Bianco ]


Argote, Linda. "Organizational Learning Curves: Persistence, Transfer, and Turnover." International Journal of Technology Management, July 1996, 759.

Brisco, Nat R., and Stephen Roark. 'The Learning Curve and Production Standards: Learning Implications." Review of Business, spring 1991, 31-35.

Chang, Tung-lung. "Cultivating Global Experience Curve Advantage on Technology and Marketing Capabilities." International Marketing Review, December 1996, 22-42.

User Contributions:

Comment about this article, ask questions, or add new information about this topic: