Learning Curves and Knowledge Decay
Organizational learning is commonly defined as modifications in the knowledge base of an organization induced primarily, but not exclusively, by the organization’s experience (Huber, 1991). It is perceived as essential for sustained competitive advantage (Aldrich, 1999; Argote and Ingram, 2000; Eisenhardt and Martin, 2000; Kogut and Zander, 1992; Prahalad and Hamel, 1990; Rumelt, Schendel, and Teece, 1994; Teece, Pisano, and Shuen, 1997), as well as for strategic renewal (Crossan and Bedrow, 2003). Unsurprisingly, the concept of organizational learning has been of enduring interest to researchers and practitioners alike, and the chapters of this volume are a testament to the progress that has been made in understanding organizational learning.
At a practical level, organizational learning is the ability of an organization to create answers to problems that need to be solved, or to create a better answer to existing problems (Argote, 1999; Miner and Anderson, 1999; Miner and Mezias, 1996). It refers to the different processes by which organizations add to their stock of knowledge and to their repertoire of capabilities. Learning occurs most obviously when an organization encounters unusual events for which it has no solution in its collection of standard operating procedures (e.g. Cyert and March, 1963) and is therefore forced to do things in novel ways. More specifically, learning happens when the solution that has been found to that particular problem for which there was no solution is formalized in routines (Cyert and March, 1963; Nelson and Winter, 1982), rules (Levitt and March, 1988; Schulz, 1998; Schulz, 2001), operating procedures (Cyert and March, 1963), mental models, and dominant logics (Bettis and Prahalad, 1995) that enable the organization to make further decisions and to take more actions in the future.
Early studies of learning in organizations were mostly based on learning curve models, which established a positive relationship between the experience gained by performing similar activities and gains in productivity. The theoretical details of gains-through-experience were first suggested by Wright, who showed a non-linear relationship between the average direct labor needed to produce a unit, and the cumulative number of units produced (Wright, 1936).
Thereafter, a large number of studies documented this ‘learning-by-doing’ and ‘learning-through-experience’ phenomenon such as Rapping (1965) who observed large gains in productivity in the production of war vessels, with average annual increases of ‘about 40 per cent.’ The same phenomenon was found in the production of rayon (Hollander, 1965), nuclear plants (Joskow and Rozanski, 1979), and ships (Searle and Goody, 1945), and also, later, in service organizations such as pizza parlors and hotels (Darr, Argote, and Epple, 1995; Martin de Holan and Phillips, 2004b).
At their core, learning curve models establish a positive relationship between experience and productivity: as organizations repeat a similar task, they learn, leading eventually to an accumulation of knowledge that translates into improved outcomes. Experience in these studies was usually assessed by cumulative production of similar units with relatively stable technology, and efficiency is seen to increase at some predictable rate with cumulative production. Several authors provide useful reviews of this literature (for a detailed review, see Adler and Clark, 1991; Argote and Epple, 1990).
Because of their apparent simplicity and intuitive appeal, learning curves quickly became popular among practitioners as useful instruments for cost estimation and control (Dutton, Thomas, and Butler, 1984). At the same time, they were a kind of ‘black box’ and suffered from several limitations. In particular, there was a lack of understanding of how learning from experience occurred in organizations and a lack of clear boundary conditions that would allow practitioners to understand under what circumstances learning through repetition would occur and when it would not (Abernathy and Wayne, 1974; Yelle, 1979, 1985).
Furthermore, one important conundrum rapidly became evident: because cumulative production can only increase over time, productivity should only rise over time also (Sterman, 2000). But this contradicts both intuition and what is observed in practice: often, performance in organizations deteriorates even in the presence of increasing cumulative volumes of production. Accordingly, the idea that knowledge accumulated by organizations can also deteriorate, decay, or otherwise disappear began to appear in the literature along with concerns with the ways it accumulates. These early learning curve studies led to both the blossoming of learning research and the beginnings of a concern with forgetting. This concern with forgetting was also encouraged by a small number of early theoretical discussions (Carlson and Rowe, 1976; Hedberg, 1981; Nystrom and Starbuck, 1984; Starbuck, 1996; Wickelgren, 1976).
However, in spite of this trickle of work and the general interest surrounding the concept (Besanko et al., 2010; Easterby-Smith, Antonacopoulou, Simm, and Lyles, 2004) organizational forgetting failed to gain traction among mainstream researchers. This is not, of course, to say that no research on forgetting has followed these early discussions. In early research, processes of forgetting or unlearning were mentioned explicitly by a few authors (Bettis and Prahalad, 1995; Day, 1994a, 1994b; Hedberg, 1981; Lyles and Schwenk, 1992; Nystrom and Starbuck, 1984; Starbuck, 1996), but seldom as the main object of the research, and often as a serendipitous finding that did not fit quite well with theory. Anand and associates, for example, present a clear example of forgetting as part of their argument, although the notion itself is not subsequently developed in their discussion.
Managers at the propulsion systems division of a major aerospace company selected an engineer to become the in-house expert in a new technology. In a wave of management changes, the champions of the technology all moved out of the division. The expert engineer was reassigned to normal duties. After another wave of change management, it became apparent that the technology was critical, but no one remembered that there was an expert on staff, and the process was repeated.
(Anand, Manz, and Glick, 1998: 798)
Similarly, but from a more practitioner-oriented perspective, Peters (1994: 128) argues that forgetting (or unlearning) is important for the performance of the organization, but doesn’t take it much further than this vivid vignette:
The issue for Ford Motor Co., home of the original whiz kids, and others in the 1980s was forgetting—that is, unlearning the habits attached to a once-viable way of life. . . . In 1938 the company tried to build a small car and failed miserably. ‘Small car’ was translated by Ford’s engineers into ‘shrunken big car’. A stubby, expensive, over-engineered product emerged. Not only was Ford snared by yesterday’s routines (big-car design was the only variety the firm’s engineers knew), but it then overlearned from its 1938 error—that is, ‘We don’t know how to build small cars.’
More recently, however, a number of researchers have begun to explore the nature and consequences of organizational forgetting, and some empirical evidence (Argote and Epple, 1990; Benkard, 2000; Darr et al., 1995; Epple, Argote, and Devadas, 1991; Martin de Holan and Phillips, 2004b) has begun to emerge, showing that the knowledge base of an organization can deteriorate, sometimes quite rapidly, which suggests that knowledge retention should be seen as a process requiring managerial effort and attention (Ocasio, 1997), and one that is prone to errors and mishaps.
From almost the beginning, then, both theoretical work and empirical findings gave rise to the concept of ‘organizational forgetting,’ highlighting that an organization not only acquires new knowledge, but also loses knowledge, either purposely or voluntarily. We will explore the literature on involuntary and purposeful forgetting in the next section.