Working paper: 02-043
Just-in-Time in practice at Toyota:
Rules-in-Use for building self-
diagnostic, adaptive work-systems
Steven J. Spear
Last revised: April 12, 2010
Copyright ? 2001 Steven J. Spear
Working papers are in draft form. This working paper is distributed for purposes of comment and discussion only. It
may not be reproduced without permission of the copyright holder. Copies are available from the author.
Just-in-Time in practice at Toyota:
Rules-in-Use for building self-diagnostic, adaptive work-systems
Working paper: 02-043
Harvard Business School
1 617 495 6741 (p)
1 617 496 4059 (f)
Operations Management Division nominee for
the William H. Newman Award for outstanding paper based on a recent dissertation
at the Academy of Management conference,
Denver, Colorado, August 2002
This paper asserts that problem identification and problem solving processes can be
integrated into work processes by imbedding tests that evaluate system-performance. These tests
are imbedded in individual work activities, in the connections that link those who provide a
product, service, or information with those who receive it, and in the overall construction of
pathways over which products, services, and information take their final form. These tests make
it unambiguous when, where, and by whom problem solving is necessary, and, as an integral part
of collaborative work, these tests help improve processes and deepens process knowledge,
allowing an organization to be increasingly adaptive, both when it experiences operating
difficulties and in determining how to exploit best market opportunities. These immediate tests
are possible if work designs are specified before work is performed, and these immediate tests
have most value if each indication that a problem has occurred is followed immediately by root-
cause analysis and structured problem solving.
This paper builds upon observations made in the manufacturing sector to draw lessons
applicable to more general management concerns of delegating/task partitioning, coordinating,
and task execution. This paper shows how the specific tools of the Toyota Production System
(„TPS‟) such as pull-systems, kanban cards, and andon cords are artifacts of a general,
comprehensive approach to managing collaborative work systems that allows frequent, fine-
grained problem identification and improvement in overall organizational structure, coordinative
mechanisms, and task-performance. Therefore, this paper phrases Toyota‟s practices in terms of
solving problems of work delegation, coordination, and execution.
Toyota Production System, Rules-in-Use, organizational design, process improvement
- 1 -
In reaction to a race to „best practice‟ -- as reflected in initiatives such as TQM, JIT, re-
engineering, and „lean manufacturing‟ -- Hayes and Pisano (1994) encouraged managers to re-
focus on achieving strategic fit by configuring production systems „through a series of
interrelated and internally consistent choices [that reflect] the priorities and trade-offs in its
competitive situation and strategy‟. This had to be grounded in „a collection of evolving
capabilities … which provide the flexibility needed to embark in new directions‟. This
admonition fit well in the organizational theory, i.e., Lawrence and Lorsch (1967), and
operations management literature, i.e., Skinner (1974), which had encouraged „contingent‟,
„focused‟ organizational forms. Nelson and Winter (1982) offered that the structures and
dynamics of organizational „routines‟ are discovered iteratively, and writers such as Clark, Hayes, and Wheelwright (1988) and Jaikumar and Bohn (1991) emphasized that improvement and
development occurs through problem solving, which, von Hippel (1994), Leonard-Barton (1994),
von Hippel and Tyre (1995), and MacDuffie (1997) reminded, was situated with problem solving
information localized in terms of time, place, process, and person.
This paper asserts that organizations can develop the capabilities to be highly adaptive -- able to address operational problems as they occur and capitalize on market opportunities as they
develop -- by putting in place mechanisms that allow highly situated learning that is both broadly
distributed throughout the organization but which also works towards common purpose. A
critical element in achieving this capability is designing work -- both that done by individuals
and that done by groups, collaboratively -- so that problem-solving based, improvement
opportunities are evident quickly and so that these opportunities are exploited rapidly.
This paper is organized in the following fashion. Section 2 explains why Toyota was chosen as a research site, and Section 3 explains the ethnographic approach I used. Section 4 introduces
a framework for characterizing collaborative work as complex systems with „hierarchical designs
levels‟, akin to characterizing technical systems as having an overall architecture in which
interfaces link components. Section 5 provides illustrations of a Rules-in-Use approach for
managing the design, testing-in-operation, and improvement of work-systems. Section 6
- 2 -discusses the implications of using Rules-in-Use to managing collaborative work.
2 RESEARCH CONTEXT
To study first hand and thereby gain an understanding of the micro-dynamics of process
improvement, Toyota and its affiliates were chosen as research sites. Existing links among
Toyota‟s quality, cost, and variety advantages and its workforce management and problem-solving processes -- collectively referred to as the Toyota Production System („TPS‟) -- supported this decision. Since the 1960s, Toyota has been more productive than its competitors
[Cusumano (1988, 1989)]. Its „TPS‟ factories have been operationally different from „Fordist‟
and „pre-Fordist‟ competitors [Krafcik (1988)]. Indeed, Toyota and its Takaoka plant epitomize
„lean manufacturing‟ [Womack, Jones, and Roos (1990)].
In 2001, Toyota continued to maintain industry leadership. Consumer Reports rated Toyota
models first in four of ten product categories. In a separate 2001 study of initial quality, J.D.
Power and Associates rated Toyota and Lexus products first in 7 of 16 product categories.
Toyota‟s Kyushu car plant was rated the best in the world, with Toyota‟s Tahara car plant second
in Asia and the Kyushu truck plant third. In North America, Toyota‟s Cambridge Ontario plant
was first, and the Georgetown Kentucky truck plant tied for second. Despite industry-wide
difficulties, Toyota‟s market share and capitalization continued to grow [Burt and Ibison (2001a, b, c)]. Worker involving, problem-solving processes at the NUMMI plant, specifically, a TPS-
managed joint venture with General Motors, were a source of performance superiority [Adler
and co-authors (1993a, 1993b, 1997)].
2.1 THE LITERATURE’S EXPLANATIONS OF TOYOTA’S OPERATIONS BASED ADVANTAGE
To explain Toyota‟s performance advantages, much focus has been on Toyota‟s Just-in-Time tools such as kanban-card paced pull systems, frequent, small batch production and delivery, and
reduced inventories. For instance, Hopp and Spearman (2000) have contrasted ConWIP and
kanban control of production flows. Deleersnyder et al (1989) and Lee (1989) have compared
the relative efficacy of push and pull approaches for production.
In contrast, Adler (1993), Adler and Cole (1993), and Adler, Goldoftas, and Levine (1997)
have focused on Toyota‟s work practices and have emphasized the role of workers in solving
production-related problems. Thus they are positioned more closely to writers such as Hayes - 3 -
and Wheelwright (1984) and Clark, Hayes, and Wheelwright (1988) who emphasize the „micro-
infrastructural‟ elements of operations management such as measurement and control systems,
workforce policies, management selection and development policies, and organizational
structure, and MacDuffie (1997), who has focused on human resource practices and problem
solving. Thus, this latter group is more aligned with administrative theory that is concerned with
delegation, coordination, and problem solving. This literature has a history tracing back to
Weber, Taylor, Barnard, and Drucker and that more recently describes organizations in terms of
routines, such as Nelson and Winter (1982) or dynamic capabilities, such as Teece and Pisano
2.2 RULES-BASED, ADAPTIVE ROUTINES AS THE SOURCE OF TOYOTA’S ADVANTAGE
The field research reported in this paper leads to the conclusion that Toyota has developed a
powerful „dynamic capability‟ in the form of consistently practiced „Rules-in-Use‟ for
organizational design, improvement, and adaptation. I discovered that in TPS-managed
organizations, [tending to] all work is executed as hypothesis testing experiments that contribute
to accelerated generation and accumulation of individual and organizational learning about
delegating, coordinating, and performing work done collaboratively. This includes work that is
done repetitively and that which is done a few times only.
This discovery adds to the literature most immediately by explaining that the production
tools that have received so much attention in the operations research literature are artifacts of
these fundamental „Rules-in-Use‟ routines, and by explaining the organizational structure and
dynamics in which Toyota‟s continuous improvement occur. More broadly, it provides
actionable principles for designing systems that are quick to detect problems in their design as a
precursor to rapid improvement and adaptation.
- 4 -
Many scholars argue that, as a prerequisite to building inductive theories of how processes truly operate, observation and participation must be used to study complex social systems.
Ethnographic methods, for example, have been used to articulate social structure and dynamics
in situations such as an immigrant Boston neighborhood [Whyte (1993)], religious communities
[Heilman (1984, 1992)], and medical practices [Barley (1986, 1990)]. Classic works, such as
those by Barnard (1938), Roethlisberger (1942), and Parker-Follet (1940), were deeply grounded
in each author‟s self-reflective participation or intense, close-hand, sustained observation.
As previously discussed, Toyota consistently outperformed competitors, even though it had
1been open to them, and they had tried to emulate Toyota. This not only suggested that Toyota‟s
management processes had not yet been fully characterized but made ethnographic methods
appropriate for understanding the phenomenon of work-system management in greater detail.
To learn how various work systems actually operated, for 176 days during June 1995 to May 1999, I gathered data by doing or observing work across functional specialties at several
different organizational levels. My involvement covered a variety of technical processes at
different supply-chain stages and in different product-markets across 33 sites in North America
and Japan. For five months, I was one of a four-member Toyota team implementing TPS on the
shop floor at a supplier. I gathered additional data at Toyota‟s Tsutsumi, Takaoka, Kyushu,
Georgetown Kentucky, Princeton Indiana, and NUMMI assembly plants, and the Kamigo engine
plant. Others sites included six Toyota suppliers in Japan and six in North America at differing
stages of TPS mastery. To further avoid „sampling on a dependent variable‟, I gathered data at
non-Toyota sites as well. This included actually working on the assembly line at a non-Toyota
plant for one week and observing work at several other plants not affiliated with Toyota.
1 Toyota‟s Georgetown plant has had hundreds of thousands of visitors, and competitors have done major benchmarking studies (Source: Toyota). Chrysler‟s Operating System (COS) was meant to emulate
TPS (Source: Chrysler manager who helped develop COS and deployed it at two plants). General Motors
has had the NUMMI joint venture with Toyota since 1984. (Source: http://www.nummi.com.), and - 5 -modeled its Global Manufacturing System („GMS‟) on TPS, according to an authority deeply involved in
For validity, I followed the guidelines for grounded, theory-building research developed by
Strauss and Corbin (1990) and Yin (1994). I visited work sites with and worked under the supervision of members of Toyota‟s Operations Management Consulting Division („OMCD‟) and the Toyota Supplier Support Center („TSSC‟). These groups are tasked with developing TPS expertise at Toyota and supplier plants in Japan and North America, respectively. I kept journals that ultimately totaled thousands of pages of daily narratives, material and information flows, ethnographs, and other diagrams and illustrations.
Highly detailed documentation of how systems actually operated, across the multiple dimensions of product, process, function, etc., mentioned above, protected the data and analysis from becoming overly subjective. Analysis was not based on recalling impressions that had faded with time. Rather, detailed, written documentation allowed me to determine what features were context-specific and what were generally characteristic of high-performing systems. Furthermore, I sought to discern consistent patterns in my data, such as what were or were not good applications of „TPS thinking‟. I made these determinations in conjunction with the Toyota staff with whom I worked and who are mentioned above, and through frequent reviews with colleagues. I did this to ensure that both the Toyota staff and my colleagues drew the same conclusions from my data, either on-site while the data were being collected or after the fact with reference to the detailed descriptions I was creating. As my formulations progressed, I predicted work designs before I arrived at plants and used discrepancies between my predictions and actual practice to make refinements. These cycles were methodologically important as deductive tests of the inductively generated frameworks I was developing.
- 6 -
4 CHARACTERIZING COMPLEX WORK SYSTEMS
This paper characterizes complex work systems with a framework of hierarchical design
levels similar to one used to characterize complex technical systems. Consider a familiar
example. A typical personal computer performs a number of functions. Most simply, the system
accepts data as input, stores information, performs computations on that data, and generates
output in a form valued by the user. In order to create a system capable of performing these
basic functions, designers had to make architectural decisions as how to map those functions
onto different parts of the system. With early personal computers, the input function was
assigned to the keyboard and a disk drive, for instance, data storage was assigned to a hard disk
drive, computation to a CPU on a motherboard, and the output function was assigned to monitors
and dot-matrix printers. Designers then faced the issue of coupling the various pieces together,
so had to make interface decisions, for example creating the serial and parallel port formats for
linking computers to printers. Finally, they had to design the individual components of the system, determining how the various printers, monitors, and keyboards would operate. In
contrast, modern-day, hand-held computers represent a different set of hierarchical design
decisions. The system‟s value proposition is primarily around portability and convenience, not
processing capacity and power. This led to the architectural decision to assign input functions to
a touch screen and stylus rather than a higher speed keyboard, storage to flash memory rather
than a higher capacity hard disk drive, and computation to a small, relatively inexpensive chip
rather than a high end micro-processor. These pieces are linked through interfaces different from
the ports characteristic of desktop machines, and the design of the hand-held‟s components comply with the system, architecture, and interface decisions that have been made previously.
Organizations too can be characterized in terms of hierarchical design levels. At the system
level, we ask, what mix and volume of products or services does the organization provide to
whom, when? In asking about delegation of responsibility, we are asking an architectural or
pathway question by inquiring who provides what intermediate product, service, or information
to whom? In asking how those who need a product, service, or information -- whom we will
refer to from here as „customers‟ -- will receive what they need, and in asking how those who
will provide a product, service, or information -- whom we will refer to as „suppliers‟ -- will
- 7 -deliver the item, we are asking an interface or connection question. Finally, in asking how individual „suppliers‟ will do the work assigned to them, given the way they connect to the rest
of the work system, we are asking an work-activity question about the functional components of
Table 1: Product and Process Design Hierarchies and related questions
Products Processes/Organizations Design Level Critical questions Design Level Critical questions
What does the organization What functions does the System System produce and deliver (mix, system provide for whom? volume, timing) for whom?
How is functionality assigned Who creates what output to „spaces‟ in the system and Architecture Pathway (product, service, or how do the spaces relate to information) for whom? each other?
How are the spaces joined How do customers and together? Interfaces Connections suppliers communicate How do material, information, requests and responses? and energy flow?
How do people or machines
How do components perform produce and deliver outputs for
the functions assigned to them which they are responsible Components Activities given the interfaces they have given the connections they
with the system? have with immediate
customers and suppliers?
Figure 1: Pathways of connected activities
Person doing a
of material and
- 8 -