Automation, motivation and lean production reconsidered
The Authors
Dan Coffey, Leeds University Business School, University of Leeds, Leeds, UK
Carole Thornley, School of Economic and Management Studies, University of Keele, Keele, UK
Abstract
Purpose – Aims to present an alternative way of interpreting unfolding events as these pertain to the organisation of manufacturing practices in the assembly plants of the leading Japanese car assembler, Toyota.
Design/methodology/approach – This is an analysis of assembly plant automation in the automotive industry.
Findings – Fifteen years ago, it was argued that the lean car assembly plants of the future would be comprehensively automated, but that in the meantime organization rather than automation was the watch-word for efficient plants. Today it is possible to invert this prognosis as it applies to the leading “lean” car assembler, Toyota. Automation certainly played a much larger role in accounting for high labour productivity in the late 1980s than has generally been understood; but in the subsequent years priority has been given to managing the manual component in car assembly, and aggressive automation as a preferred strategy has been put on ice.
Originality/value – The findings raise new questions about future trends in the world automotive industry.
Article Type:
Viewpoint
Keyword(s):
Assembly; Automation; Lean production; Automotive industry.
Journal:
Assembly Automation
Volume:
26
Number:
2
Year:
2006
pp:
98-103
Copyright ©
Emerald Group Publishing Limited
ISSN:
0144-5154
Introduction
The way in which we interpret the present is inevitably shaped by the way in which we interpret the past. In this paper, we suggest an alternative way of interpreting unfolding events as these pertain to the organisation of manufacturing practices in the assembly plants of the leading Japanese car assembler, Toyota. This obviously neither denies the manifest success of Toyota as a business nor its individually unique characteristics as a manufacturer and seller of cars; similarly, there is no question as to its intrinsic interest both as a producer and as a corporation. But it is possible to frame the question of the role of automation at Toyota in a way which differs from what has become a conventional wisdom, while retaining as valuable much of the existing body of evidence.
The enormously influential and best-selling 1990 book – The Machine that Changed the World, by Womack et al. – which popularised the concept of lean production in the world car industry set out a somewhat paradoxical position on the role of automation in the manufacture and assembly of cars. A prediction was made that by the end of the century automation in lean factories would have advanced to the point where car assembly plants no longer relied on manual labour for the more mundane tasks involved in putting cars together:
We believe that once lean production principles are fully instituted, companies will be able to move rapidly in the 1990s to automate most of the remaining tasks in auto assembly – and more. Thus, by the end of the century we expect that team-assembly plants will be populated almost entirely by highly skilled problem-solvers whose task will be to think continually of ways and means to make the system run more smoothly and productively (Womack et al., 1990, p. 102).
The future, for lean plants at least, would therefore be one of the increasingly automated factories, especially in car assembly.But at the same time, when reporting their views on the then existing state of affairs in the world car industry, Womack et al. chose to emphasize the immediate importance of what they called lean organization, rather than automation, in explaining why Japanese car assemblers appeared to possess such large productivity advantages in car assembly over western rivals. This view was based on the results of the famous MIT survey of international productivity differentials in the world car industry, carried out in the late 1980s under the auspices of the largely industry funded International Motor Vehicle Programme. This was the largest international survey in the history of the industry, and it collected data both on the hours of labour used to build cars in assembly plants around the world, and on differences in prevailing levels of assembly plant automation. And as Womack et al. were able to note, this data seemed to show a very considerable variation in the productivity scores for different assembly plants at each and every level of automation: in particular, the Japanese assembly plant listed as most productive in terms of its savings on manual labour by the MIT survey – “the most efficient plant in the world” – was also the least automated of all of the Japanese sites visited (Womack et al., 1990, p. 94). From this they evolved what they called the “simple axiom”: lean organization should precede automation.
And this view is one which is certainly now embedded in the western literature. On the specific subject of Toyota and investments in automation, for example, a just published guide to the Japanese assembler's production system offers this illustration (its authors are notable figures on the contemporary North American scene in car manufacture):
[T]he competitive advantage that many Japanese firms had gained in their respective industries came not from advantages in “hard” technology … but from the way they manned the same technology … Toyota attained holistic integration of technology with people, organization, product and strategy … the difference lay in their socio-technical system (Liker and Meier, 2006, pp. 194-8).
This refers to views advanced on the reasons for Toyota's initial advance against its North American competitors and the subsequent basis for its enduring success, and similar examples can be found in any number of recently published books or articles. The interpretative template initially laid down for Toyota is one which has stuck.Paradoxically, therefore, the vision advanced by Womack et al. even while predicting a steady move towards more fully automated factories, also served to downplay the immediate significance of differences in automation in accounting for differences in measured productivity. This instead was attributed to the “socio-technical” dimensions of production, if we borrow this phrase. Our first objective in this paper is to suggest that automation was far a more significant variable in accounting for Toyota's initial (labour) productivity advantages than is now commonly supposed.
Subsequent developments, however, have also pointed to abiding complications in this complicated industry. Just when a wider world was learning about lean production, it became evident that Toyota MC, the lean thinker's lean producer, was experiencing a crisis in recruiting and retaining operatives to work in its assembly plants. This has generally been attributed by commentators since to tough working conditions:
Whilst Toyota Motor's economic performance is widely regarded as superior, its employees are also believed to be operating under high pressure. As at virtually all assembly lines, the physical working conditions are tough. Assembly workers operate under a tight time schedule on which they have no influence, and perform repetitive tasks at a high pace (Benders and Morita, 2004, p. 435).
In itself, a comment of this type will not surprise anyone with some longevity of experience in the management of assembly plant operations: all assembly systems with a sizable manual component encounter issues of worker motivation. If the oxymoron be permitted, what Toyota learned in the 1990s was that the ghost in its machine was alive and well. But the irony in this instance is that while this has since led Toyota to experiment with new forms of work organization – and Toyota has never been a static organization – there is evidence that these new forms may be less conducive that the old to more extended automation. Problems experienced in managing a still sizeable manual component in partially automated car assembly plants in the late 1980s and early 1990s may have encouraged solutions that are less obviously compatible with more completely automated factories in the near future. Our second objective in this paper is to complete the transformation in perspective as follows: if the importance of automation in accounting for Toyota's performance in the 1980s has been understated, unforeseen problems with the “socio-technical” aspects of its organisation may in turn have served to forestall the longer-run move predicted towards more fully automated systems.Automation in the world car industry: the MIT survey revisited
It is clear that the prediction of more or less fully automated lean factories by the end of the century has proved premature. It is not difficult to see, however, why observers looking at Japanese car assembly plants in the late 1980s might have supposed this to be a discernible trend in the industry. Indeed, the significance of extensive automation in explaining Japanese performance at that time may have been understated.
This point is best established by briefly reconsidering the original findings of the MIT survey, alluded to above. The design of the automation index employed for this survey was undoubtedly clever, given the number of steps involved in assembling a car. A “total automation index” was constructed using a weighted average of individual measures of automation calculated for the different main stages of the car assembly process – in the body shop, in the paint shop, and on the trim and final assembly lines. The weights applied to each stage were based on the proportionate distribution of employment in a “representative” plant. For each individual stage, considerable ingenuity was displayed in contriving reasonable estimates of the extent of automation. For example, for the installation of trim and final assembly 44 separate process steps were identified, and dummy variables used to signify automation at each point: each automated step was assigned a value of 1, each manual step a value of 0, and the results were summed. In this way, a percentage figure could be obtained for the number of automated steps. The index itself was designed by Krafcik (1989), an MIT-based engineer.
As with any measure of automation, it is always possible to quibble. In particular, the use of constant weights when judging the contribution of automation, in individual processes or process-steps, to labour productivity in a factory system as a whole, implies that the only effect automation has is to displace labour from that immediate point: the tacit assumption would be that automation never clears (or causes) a “bottleneck” – there are no secondary effects in the system as a whole vis-à-vis labour productivity. A much more important consideration in this instance, however, was the subsequent analysis of data.
When plotted on a graph with labour hours per car on the vertical axis and estimated levels of factory automation on the horizontal, the data obtained by the MIT survey was suggestive of an overall relationship between the two variables. But at the same time, assembly plants in Japan were typically highly automated when compared to plants in other regions covered by the survey: North America, Europe, Australia, and the newly industrialising countries (NICs), Brazil, Mexico, South Korea, and Taiwan. Even the least automated of the Japanese assembly plants surveyed – the one selected by Womack et al. as the most efficient plant in the world – still achieved a very high score on automation compared to most plants elsewhere in the world. In stylised terms, the basic shape of the data could be visualised in terms of Figure 1. While observations as a whole distributed across a wide area, represented here by W, Japanese observations clustered in the bottom right hand corner of the scatter – in the region J. (Readers interested in the details of the labour productivity index used in the MIT survey, based on a rather complicated formula for the calculation of labour hours per assembled car, should consult Krafcik (1988); readers interested in inspecting the actual scatter of observations for comparison's sake will find it in Womack et al. (1990, p. 95)).
All else being equal, this would suggest a very important role for automation in accounting for Japanese performance in the world car industry of the late 1980s. But statistical analysis carried out at the time found that while passing a “best-fitting” line through the sample of observations did indeed establish a statistically significant relationship between productivity and automation, variation in automation seemed to account for only one-third of the total variation in the hours of assembly plant labour used to build cars in plants around the world, as measured by the R 2-statistic.
It is, however, possible to draw a bolder inference. Because what was also immediately evident from the data was that the “poor” performers at each and every level of assembly plant automation were all plants based in Europe: regardless of ownership, these sites all seemed to perform less well than plants elsewhere. Car assembly plants in Europe with high levels of automation certainly seemed less productive than plants with similar levels of automation in Japan; but they also performed less well than similarly automated sites in North America, Australia, and even most of the NICs. Judged solely on the basis of the MIT survey report, Europe was distinguished, as a region, by “weak” results.
But if due account were taken of this when interpreting the MIT data, a much bigger role could be ascribed to automation in accounting for the relative performance of assembly plants in Japan vis-à-vis the rest of the world, with respect to productivity. The idea can again be captured by means of a simple diagram, as shown in Figure 2. In order to separate out a possible bias in the statistical analysis arising from one differently performing region, the observations on productivity and automation could have been quite reasonably divided into two samples, one for Europe, and one for the rest of the world. If a “best-fitting” line had then been passed through each, the result would be as shown: the regression line for the European subset would lie above the regression line for sites elsewhere. To gauge the likely importance of automation after making allowance in this way for the possibly distorting effects of Europe, one would simply look again at the R
2-statistics. On the basis of some experimentation with data similar to the data published in the MIT survey report – as set out in Womack et al. (1990, pp. 94-5) – automation seems to have been a much more important influence in accounting for Japanese performance than was generally allowed at the time, accounting for about three quarters of the sampled variation in labour productivity in all regions of the world, excluding Europe
This view is consistent with other sources of evidence at this time. It has long been recognised that Toyota investments in facilities and equipment was relatively large compared to other car assemblers in the 1980s, a spending pattern which accelerated not only for Toyota in the later 1980s, but also for Nissan and Mazda (for a summary of work carried out by Japanese researchers, see, for example, Gronning (1995, pp. 411-2)). So far as the relative merits of superior organization versus higher investment in automation in the world car industry of the late 1980s is concerned, Japanese performance may in fact have owed a great deal more to the second factor than has been supposed. Why assembly sites in Europe should have performed so badly in the MIT survey is obviously an interesting question in its own right, but so far as the rest of the world was concerned, a much higher explanatory value could have been attached to plant automation. But in this context, subsequent departures at Toyota (in Japan) appear all the more striking.
Problems at Toyota: automation versus motivation
With the tightening of labour markets in Japan in the 1980s Toyota suffered what some writers refer to as a “crisis of work”, that struck just before the company experienced a second type of difficulty caused by the collapse in the 1990s of the Japanese economic “bubble” (a good review is given in Shimuzu (1995), with a helpful update in Benders and Morita (2004)). The great Japanese assembler experienced difficulties in recruiting assembly line operatives sufficiently motivated to work in its factories. What has perhaps occasioned most comment about the initiatives since taken by the company – and initiatives that have continued past the point where recession in Japan has greatly improved recruiting conditions for the company – has been the introduction of segmented assembly lines, connected by in-process buffers of partially assembled cars, in the latter stages of production in a number of Toyota's car assembly plants.
Attention has perhaps focused most of all on the Toyota “laboratory” site at Kyushu, on the Aichi prefecture, although there has been experimentation elsewhere. At Kyushu, the most developed case, and a site at which production commenced only as recently as December 1992, the final assembly process has been divided into what Yasuhiro Monden has described as 11 functionally-divided “mini-lines”. Each is 100 yard (or over 90 m) or so in length: three trim-lines for electrical parts, two chassis lines, five “further” assembly lines, and an inspection line, each separated by an in-process buffer. The perceived advantages to workers have typically been described as a better identification with the “job” that follows from working in reorganized teams on discrete sections involving a fine grouping of functionally similar parts. For example, electrical parts at Kyushu are grouped into three sub-categories – wire related, instrument panel related, and pipe related – each with its own mini-line. The hoped for outcome is that this facilitates the job while enabling workers to better understand the role of their own activity in the process as a whole, thereby improving morale on each section, a step change supported by a series of ergonomic measures in the factory (for a full description see Monden (1998, pp. 350-2); see also Shizumu (1995), cited above). At the same time, it is evident too that Toyota has hoped for some improvements in line efficiency.
On a conventional assembly line, an unplanned stoppage at any single point on the line will bring the entire sequence of activities to a halt. But on a segmented-line, provided that the temporary effects of stoppages on any single line-segment (or mini-line) can be successfully absorbed via adjustments in the connecting buffers, the factory process as a whole can continue unimpeded. The significance of this development at Toyota as a means of better managing problems that might otherwise stop the entire assembly process is highlighted by Monden (1998, pp. 357-62), who provides a careful treatment of this topic, with illustrative data for the company. This, of course, is basic theory in any course on assembly lines processes: but it is a new development at Toyota (for discussions more generally of buffered assembly processes, see, for example, Groover (1987)).
It is commonplace in industrial engineering to express problems with line stoppage rates in proportionate terms. However, presented, the ratio of the average total unplanned stoppage time (say S
T) to the total processing time (say P
T) for cars built in a conventionally arranged assembly process will be a key variable for plant managers. The essence of partial automation in car assembly for the (nowadays) more labour intensive stages in trim and final assembly is that automated cells intersperse with manual cells as the car under assembly progresses to completion: this was the basis, for example, upon which the estimation was made in the MIT survey described above of comparative degrees of automation in this stage of the production process. Automation is also one way of broaching the issue of line-stoppages in a car assembly process. First, provided that the problems arise on the manual segments of the line, progressively automating the assembly sequence may effect a reduction in the ratio S
T/P
T. Second, the absolute cost that can be imputed to line stoppages will be reduced (all else being equal) by the simple fact that there are fewer workers employed in assembling the car. If direct labour costs are still a significant portion of the costs of car assembly, then employing fewer workers will reduce the incremental labour cost accruing to stoppages on the line
Figure 3 shows the basic idea. As the percentage of assembly steps, which are automated increases, the incremental direct labour cost of stoppages will fall more or less proportionately, as employment falls (X); if in addition automation reduces the frequency with which stoppages occur, the decline will be steeper (Y). Viewed in this way, a strategy of progressive automation is one way of dealing with line-stoppages.
The strategy since adopted by Toyota, however, in some of its assembly plants at least, points in a quite different direction. The introduction of buffers to an assembly process in itself need not mean that all attempts at further reducing the manual component in assembly have been abandoned. Toyota has tried to mechanize selectively, by placing mechanized workstations at either end of line-segments, “so that workers do not feel alienated through the presence of complex machinery” (Benders and Morita, 2004, p. 437). But the high-tech automation route seems to have been abandoned. Instead, the onus instead has been placed on segmentation of the assembly line from the viewpoint of better managing the manual component, with in-process buffers absorbing the temporary effects of stoppages on any single segment without stopping the entire process.
The reason for this change in direction involves other considerations, of course. Toyota's most thorough-going experiment with automation in the late 1980s proved disappointing. The “technically sophisticated” line at Toyota Tahara, which became operational in October 1991, and which required a very substantial investment, struggled to achieve the production volumes needed to justify this capital expenditure with the collapse in the Japanese domestic car market after 1992 (Benders and Morita, 2004, p. 435). And subsequently, where space constraints have meant a conflict between maintaining line segmentation with in-process buffers, and introducing large mechanized workstations, the new imperative at Toyota and its subsidiaries appears to favour the former strategy over the latter. More generally, less expensive investments in simpler labour-saving devices are being systemically favoured over complex machinery (Benders and Morita, 2004, p. 439).
On this basis, it is tempting to invert the paradoxical proposition that was first set out 15 years ago. Then, it was suggested that fully automated lean car assembly plants were imminent, but that in the meantime organization rather than automation would be the watch-ward for pioneering facilities: all eyes were on Toyota. Now, and in retrospect, it is possible to suggest a different paradox. Fifteen year ago, Japanese car assemblers were leading the field in introducing aggressively automated factories, and automation could have been quite reasonably invoked as a very major part of the explanation for comparative labour productivities in the world car industry. But the subsequent trend has been in a quite different and unexpected direction: the watch-word now, for Toyota at least, is how best to manage the manual component in the factory.
Future priorities
This is not the only step-change for Toyota, and it is perhaps worth concluding by highlighting another. This refers to the recently popular topic of mass-customization strategies in the car assembly sector. It has been widely surmised, and for a long-time, that the medium term strategy of Toyota was to expand its provision for the width of selection permitted prospective customers vis-à-vis “minor” variations on car-line derivatives achieved by increasingly generous options lists. Pine (1999, p. 137), for example, could quite recently argue that the ultimate goal of the company was to realise a strategy of mass-customization, and similarly Michael Porter, writing with Japanese collaborators, could encapsulate the lean manufacturing strategy as being one which comprised (amongst other things) “a wide line of models offering multiple features”, based on “standard products” with a “wide range of options” (Porter et al., 2000, p. 70). It is generally understood that the “bursting of the bubble” in Japan caused Toyota, like other Japanese assemblers, to draw back the number of separate car-lines that it was building in the peak years of the late 1980s. What is perhaps less obvious is that any intent to move towards more “customized” cars has been put on ice.
The effect of this can be presented in a more or less dramatic form by comparing the number of alternate model-specifications that can be constructed from representative car-lines marketed by Toyota and other major car assemblers. Table I gives a small selection from some calculations made by one of the authors of this paper. The differences in the model-specification counts shown simply illustrate how many different varieties of car can be defined based on all the choices offered customers in each case. The very large numbers for the European luxury marque BMW reflect the very long options lists that form part of the German firm's marketing strategy: but at the other extreme, Toyota, as with other Japanese assemblers, has severely curtailed choices of this type – for which reason its model-specification counts are now typically small. In this sense, there has been a definite move away from any hint of-custom building
The point is mentioned in the context of the topic, in this paper, because it is not unconnected – or at least not entirely unconnected – with the question of the re-organization of Toyota's car assembly sites. If automation has been losing out to the management of the manual component in Toyota car assembly plants, there is also evidence that one reason (no doubt amongst others) for the abandonment of any attempt to expand options lists for customers is to conserve space in the factory: the priority here has been to make space for the in-process buffers connecting the line segments. The significance of space considerations, arising from and contributing to Toyota's experiments with the re-organization of its assembly processes, have been emphasized by practically every factory visitor (Shimuzu, 1995, p. 401; Monden, 1998, pp. 350-1; Benders and Morita, 2004, pp. 439-41). While Toyota's provision for customer options was never particularly large, this may be one reason why modest increases that were evident in the 1980s in the size of this provision have been sharply reversed since. Here, as elsewhere, there has been a step-change in priorities.
What is this likely to entail for future trends in the industry as a whole? Toyota has been at the forefront of industry debates for more than 25 years, so much so that it has almost come to seem axiomatic that where the Japanese firm leads, others follow. And if the step-change in Toyota's attitude towards assembly plant automation were emulated, this would certainly be a significant development, because the general assumption remains that car assemblers will continue to favour an extension of automation as techniques and finances allow (see, for example, Clementi et al. (2005), who provide a clever cost simulation that touches on this issue). But if this assumption remains justified, Toyota will in future be distinguished not by its lead, but by its departure.
Figure 1
Figure 2
Figure 3
Table IComparing overall width of selection for selected ranges
References
Benders, J., Morita, M. (2004), "Changes in Toyota Motors' operations management", International Journal of Production Research, Vol. 42 No.3, pp.433-45.
Clementi, E.R., Piazza, P.A., Volpato, G. (2005), "Competitive assembly: brown or green field site – conventional or flexible plant – what's the best?", International Journal of Automotive Technology and Management, Vol. 5 No.3, .
Coffey, D. (2005), "Delineating comparative flexibility in car assembly: the problem of wide selection", International Journal of Automotive Technology and Management, Vol. 5 No.1, pp.18-31.
Coffey, D. (2006), The Myth of Japanese Efficiency: The World Car Industry in a Globalising Age, Edward Elgar, Cheltenham, .
Gronning, T. (1995), "Recent developments at Toyota Motor Co.", in Sandberg, A. (Eds),Enriching Production: Perspectives on Volvo's Uddevalla Plant as an Alternative to Lean Production, Avebury, Aldershot, .
Groover, M.P. (1987), Automation, Production Systems, and Computer Integrated Manufacturing, Prentice-Hall, Englewood Cliffs, NJ, .
Krafcik, J.F. (1988), "A methodology for assembly plant performance determination", IMVP Research Affiliates, MIT Press, Cambridge, MA, .
Krafcik, J.F. (1989), "A comparative analysis of assembly plant automation", IMVP International Policy Forum, MIT Press, Cambridge, MA, .
Liker, J.F., Meier, D. (2006), The Toyota Fieldbook: A Practical Guide for Implementing Toyota's 4Ps, McGraw-Hill, London, .
Monden, Y. (1998), Toyota Production System: An Integrated Approach to Just-in-Time, 3rd ed., Engineering and Management Press, Norcross, GA, .
Pine, B.J. (1999), Mass Customization: The New Frontier in Business Competition, Harvard Business School Press, Boston, MA, .
Porter, M.E., Takeuchi, H., Sakakibara, M. (2000), Can Japan Compete?, Macmillan Press Ltd, London, .
Shimuzu, K. (1995), "Humanization of the production system and work at Toyota", in Sandberg, A. (Eds),Enriching Production: Perspectives on Volvo's Uddevalla Plant as an Alternative to Lean Production, Avebury, Aldershot, .
Womack, J.P., Jones, D.T., Roos, D. (1990), The Machine that Changed the World, Harper Collins, New York, NY, .
About the authors
Dan Coffey
holds an MPhil from the University of Cambridge and a PhD from the University of Warwick. He works at Leeds University Business School, where until recently he was Director of all economics-related MA programmes. His research interests are industrial organisation and operations research, and he has extensive field-based experience in the area of car assembly scheduling and logistics. His current projects include a forthcoming book on the world car industry (The Myth of Japanese Efficiency: The World Car Industry in a Globalising Age), and a co-edited volume on the industrial crisis in Japan, both commissioned by Edward Elgar. He has recently been Guest Editor for the International Journal of Automotive Technology and Management for the Special Issue (Vol. 5, No. 3) entitled “Marketing strategies and product supply in world-wide car manufacture”.