Learning Curves and Innovation

Anne Marie Knott

The Wharton School
University of Pennsylvania


One of the central tensions in firms is that between optimization for the static environment (exploitation of existing capabilities) versus survival in a dynamic environment (exploration for new capabilities). The earliest proposed solution to this tension was to choose a position along a continuum from exploration to exploitation based on the underlying rate of change in the environment. The main challenge was characterizing the "rate of change". More recent work has shown that even with knowledge of the rate of change, the optimal solutions are still the extrema of static optimization and dynamic optimization. Thus balance is infeasible, and the tension remains.

We find the infeasibility argument compelling, thus are intrigued by the existence of "intermediate strategies" that appear to combine exploration and exploitation. We view product development to be such an intermediate strategy. Product development refers to a wide spectrum of efforts that convert knowledge/technology into products. Product development inherently exhibits exploration through the addition of new characteristics or higher performance with each new product introduction. Product development may also exhibit exploitation if there are regular reductions in product cost with each new product introduction (learning curves). We assert that if product development exhibits both exploration and exploitation, we have evidence that the two seemingly dichotomous strategies can in fact coexist, and that product development itself may be a viable integrative strategy.

To test the hypothesis we build a qualitative model of learning across sequential product developments. From this qualitative model we derive a quantitative model of product development learning adapted from that for manufacturing learning. Next we conduct an empirical test that jointly solves for manufacturing and product development learning over a 28 year period in a single firm (Toyota Motor Company).

We find evidence of substantial exploration and exploitation in product development -- demonstrating not only that the two processes can coexist, but that they can coexist in the same function. Moreover, we find that most of the learning normally attributed to manufacturing processes, actually is attributable to product development.

To appear in "Strategic Management of Intellectual Capital and Organizational Knowledge" edited by Nick Bontis & Chun Wei Choo (Oxford University Press).