Category:Know-How

Table of Contents

• Introduction
• Structure of Artifacts
• Radical and Incremental Innovations
• Operational principle
• Granularity of Analysis
• References

Introduction

Technological innovations have a profound impact on the development of industries. New technologies such as the steam engine, the internal combustion engine, the telegraph, the telephone, synthetic chemicals, semiconductors, the internet and the like have spawned entire new industries. Typically technologies are improved dramatically over their course of the existence. Writers on technological change distinguish between radical and incremental innovations, product and process innovations, etc. This article gives a primer on how to analyze different types of innovations.

Structure of Artifacts

In his famous article on the architecture of complexity, Simon (1962) pointed out that artifacts, like other complex systems, are not just made up of elementary components, all directly interacting with one another, but rather consist of a nested hierarchy of subsystems. Simon and more recent scholars of modularity in design (Langlois and Robertson, 1992; Ulrich, 1995; Sanchez and Mahoney, 1996; Frenken, Marengo, and Valente, 1999; Schilling, 2000; Baldwin and Clark, 2000; Simon, 2002; Ehiraj and Levinthal, 2004) highlight that the vast majority of technical systems are nearly decomposable in the sense that the high-frequency interactions tend to occur within subsystems and the low-frequency interactions occur across subsystems. The crucial conceptual point that empirical research on dominant designs has not sufficiently paid attention to in collecting empirical evidence is that most technologies have multiple levels of subsystems that are organized in a hierarchical fashion (Tushman and Murmann, 1998). In terms of Henderson and Clark’s innovation framework (see below), which distinguishes between incremental, modular, architectural, and radical innovations, the hierarchical structure of systems has an important implication: A modular innovation at one level in the hierarchy can clearly be an architectural or radical innovation at a lower level in the hierarchy.

First, artifacts are structured in terms of a hierarchy of nested parts (see Figure on the right). An entire airplane is made up of a fuselage, wings, propelling device, and landing gear, which can be represented as first-order subsystems. Each of these first-order subsystems has potentially included within it smaller second-order subsystems, and potentially many further levels of ever-smaller subsystems until the level of the fundamental or basic components is reached. In the case of an airplane, its wing subsystems contain within flaps, fuel tanks, lights, etc. The flaps, in turn, are composed of steering flaps and breaking flaps. Flaps have a lattice and a skin and are held together by flush rivets (for details of this example, see Tushman and Murmann, 1998). Each level in the artifact hierarchy can go through its own technology cycle marked by the processes of variation, selection, and retention.
Second, complex artifacts are also structured in terms of a hierarchy of control. The different functional parts of an airplane are subordinated to one or few subsystems that control how all other the first-order subsystems interact to form a well-functioning system. Operating systems in computers and the brain in the human body are examples of subsystems that control all other subsystems in their respective systems. Each subsystem, in turn, tends to have a component that controls the other components within the subsystem. At any level, a discontinuity in such a “core” component starts a new technology cycle.
Finally, when mapping out the structure of complex systems, it is useful to keep in mind that systems can be small or large in two senses. They can have few or many levels of subsystems (flat hierarchies versus tall hierarchies). At the same time, within each level they can be homogeneous (the first-order subsystems are of the same kind, e.g., many computers connected together into a network) or heterogeneous (e.g., the many different pieces of equipment in a plant that form a long assembly line.)

Radical and Incremental Innovations

Writers on technology have used the notion of a radical innovation in at least two distinct ways. The only common feature of these two uses is that in both cases something big and exceptional happens. Radical innovations have been defined either in terms of their antecedents (the scope of new knowledge required) or in terms of their consequences (the increased performance they make possible).  Given these two different dimensions of radicalness, an innovation could be incremental in terms of the new knowledge required but radical in terms of the additional performance achieved, and vice versa (see Table 4). Most scholars do not distinguish between these two meanings, which makes it difficult to interpret in what sense they see an innovation as being radical. When one becomes aware of these two dimensions of radicalness, innovations that require large amounts of new knowledge and create large performance improvements clearly have a particular potential to transform industrial structures. To distinguish this kind of innovation from the others, we propose to call it a radical-square (r2) innovation.

How does this innovation typology relate to the concept of a systems hierarchy? In terms of innovation in the new knowledge dimension, moving up the systems hierarchy (i.e., encompassing more and more components) by definition means that an innovation is becoming more radical because the design of more and more components is being based on new principles. This is not true, however, for the performance dimension of innovations. Here, innovations that occur at lower levels can have more radical consequences than innovations that involve the entire system. In fact, most rapid improvements in technological history have typically taken place along well-defined technological trajectories (Dosi, 1982; Sahal, 1985). This means that radical innovations (in terms of both dimensions) can occur at the individual component, individual subsystem, or a higher level of aggregation. Empirical research needs to track the exact location where the innovations takes place and examine what impact a particular location of innovation has on the industrial organization of the production system.

Operational principle

We saw earlier that empirical researchers on dominant designs always face the question of whether two designs are the same or different. A similar question that arises in defining technologies as complex systems is how one decides when two artifacts belong to the same class of complex artifacts. Obviously, both aircraft and helicopters are prime examples of complex artifacts, but should their evolution (e.g., in terms of the emergence of a dominant design) be analyzed together? Without delineation, any conceptualization of a dominant design is impossible, because dominance can only be established with reference to a distribution of artifacts that belong to a technological class. We believe the concept of an operational principle is a very useful tool for classifying artifacts into classes. This concept, originally developed by Polanyi (1962) in the context of developing a theory of how human beings know things, was later used in Vincenti’s writings on the history of airplanes (1990, 1991, 1994). For Polanyi, an operational principle captures the kind of knowledge a human designer must have in order to build a technological device that works on the physical world in a desired way. ‘In a desired way’ means fulfilling a basic user need such as “transporting goods and people through air.” To put it differently, an operational principle defines how the parts interact with one another to implement the goal of overall technology.
Consider the example of the principle underlying the first successful human flight. Instead of trying to design a flying machine with flapping wings to provide both the counterforce to gravity and forward thrust, Cawley in 1809 (Vincenti, 1990) proposed to separate lift from propulsion by using a fixed wing and propelling it forward with motor power. The central idea was that moving a rigid surface through resisting air would provide the upward force countering gravity. As Vincenti has noted, this was a radically different way to conceptualize the design of an airplane because it freed designers from the impractical idea of flapping wings. Subsequently, the fixed-wing and forward propulsion idea became the operational principle underlying all airplane designs.
When human beings have grasped the operational principle of a technology, they know how an artifact can act on nature in a special beneficial way. Because an operational principle essentially specifies how components need to be arranged in order to create a successful artifact, operational principles reveal the abstract logic of how an artifact works and thus provide the starting point for understanding what the essential aspects of a particular technology are. Saviotti and Metcalfe (1984) have argued that to analyze technological evolution it is important to ascertain the technical characteristics of an artifact, which they define as the key dimensions of a technology. In fact, the operational principle of an artifact sets out the relevant dimensions of what we will later call the design space of an artifact. For the researchers of technological change this means that once the operational principle of an artifact has been determined, this automatically decides the key technical dimensions of an artifact and thus determines in what dimensions two artifacts can differ technically without belonging to different classes of technology.
In using the concept of an operational principle, one is able to compare different technologies by probing whether they work according to the same the operational principle or not. For instance, planes and helicopters, both devices for air travel, differ in terms of how they achieve the general task of transporting humans in the air. A plane accomplishes flight by separating the propelling function and the lifting function into two separate components (the propeller or jet and the wings), whereas the helicopter realizes movement in the air by implementing the lifting and propelling function in one and the same component, the large vertical rotor. Rockets, another class of devices for traveling, make air travel possible by allowing an expanding air–fuel mix to escape only through the rear of the device and thus propel it forward. Rocket propulsion requires neither wings nor propellers.
Operational principles allow the student of technology to categorize a set of artifacts into general product classes. This is useful for research on dominant designs by making it possible to distinguish between variation within a product class that shares the same operational principle from variation between product classes that are characterized by different operational principles.

Within a product class, subclasses can be defined by distinguishing between the different operational principles used in different subsystems. More generally, designers distinguish between different design dimensions of an artifact, or its “technical characteristics”. Because different solutions are possible along each dimension, many combinations of solutions are possible to construct one and the same type of artifact. The total set of designs that can be constructed out of all possible combinations of alternative choices along its dimensions is called an artifact’s design space (Dennett, 1995; Baldwin and Clark, 2000). Consider, for example, Bradshaw’s (1992) reconstruction of the design space for gliders alone, as faced by the Wright Brothers at the turn of the nineteenth century (see Table 3). Though only a relatively small number of dimensions/components are distinguished (9), the number of possible designs is enormous (12,960,000).

Design space of early aircraft technology

The example of the three technology classes enabling air travel also serves to demonstrate a second feature of complex artifacts that we discussed earlier: hierarchical structure. Devices to travel through air are classified as aircraft, helicopters, rockets, and others. Within the class of aircraft, one can distinguish between propeller aircraft and jet aircraft. In propeller aircraft, one can distinguish between piston propeller and turbopropeller, and so forth. Complex systems typically consist of components that are themselves complex systems. As a result, one can conceptualize complex artifacts as a nested hierarchy of design spaces. This hierarchical view of dominant designs implies that there can, in fact, be dominant designs at a higher (more encompassing) level without there being any dominant design at the lower level. All that is required is that the operational principles—i.e., the way lower-level subsystems are combined into a system through a set of standard design rules—be dominant across the industry (see Baldwin and Clark, 2000, for a superb articulation of the idea of design rules).

Granularity of Analysis

Empirical research on technological change requires a judgment about whether two designs are different or the same. The outcome of these judgments depends crucially on the level of resolution or granularity one brings to the analysis. Two distinct dimensions of granularity are relevant in this context: the level of detail at which the artifact is examined and the granularity of the time interval used for measuring the dynamics of technical change. At the most detailed level of analysis, no two artifacts are the same; at the coarsest level of analysis, every two artifacts are the same. In addition, using a very small time interval for recording observations (e.g., milliseconds) a technology will never display any change in its design from one interval to the next because human designers operate at longer time scales. Using a very long interval for recording observations (e.g., 1,000 years) a technology is likely to display so much change that one can never speak meaningfully about the emergence of any kind of standardization. Moreover, such a coarse time resolution misses entire classes of technology for which the full life span is shorter than 1,000 years. Competitive dynamics in industries play themselves out on the time scale of days, weeks, months, quarters, years, and decades. And in many instances, judgments about whether a dominant design has emerged within a product class will be sensitive to whether one has picked one time scale in the range or another one. To date, research on dominant designs has not fully recognized this problem. We believe that, to make studies comparable, it is important to be explicit about the time interval used for measurement.

References

Murmann, Johann Peter and Koen Frenken. 2006. Toward a Systematic Framework for Research on Dominant Designs, Technological Innovations, and Industrial Change. Research Policy. Vol. 35. In Press.

Tushman, Michael L., and Johann Peter Murmann. 1998. “Dominant Designs, Technology Cycles and Organizational Outcomes.” In B. Staw, and L. Cummings (eds.), Research in Organizational Behavior: 231-266. Greenwich, CT: JAI Press.

Categories:

• Know-How