Plant Capabilities for Quick Response Manufacturing
A Working Paper
David M. Upton, Harvard Business School
Copyright (c) D. M. Upton 1994
Table of Contents
This paper examines the relationship between one form of manufacturing flexibility---operational mobility (or the ability to change quickly between products)---and structure, infrastructure and managerial policy at the plant level. Using data from a broader study aimed at exploring the general sources of manufacturing flexibility, the paper provides evidence of the strength of the links between manufacturing flexibility and such factors as scale, technology vintage, computer integration and workforce management. There has been little empirical work on this subject, partly as a result of the difficulties of defining and measuring flexibility. The type of flexibility explored in this paper is specifically the capability of a plant to change between process states quickly.
I find that most of the variance in flexibility across plants may be explained by a combination of structural factors, such as the scale of the plant; infrastructural factors, such as the length of service of the operators; and measures of managerial emphasis, such as the perceived importance of making quick changeovers. Of these factors, I find that the scale of a plant does not strongly inhibit its flexibility and that computer integration is either insignificant or detrimental to the flexibility of a plant. Workforce characteristics are also important determinants of flexibility, and the results suggest that less experienced operators may be more flexible in their ability to change quickly between products. The strongest determinant of the flexibility of a plant is the extent to which managers communicate the importance of various performance dimensions.
Finally, I show that different forms of flexibility are not necessarily related to each other. This emphasizes the importance of identifying precisely the particular type of flexibility that is to be developed when building manufacturing capabilities.
Keywords: manufacturing flexibility, competitive capabilities, plant management, plant network management.
The improvement of flexibility has become increasingly important as a method to achieve competitive advantage in manufacturing(1). One of the most serious impediments to its improvement however, has been the vagueness of the term. The word has come to be used for many purposes, each of which characterizes a different quality or capability of a system. The confusion and ambiguity about a concept that often represents a critical competitive capability seriously inhibits its effective management.
A key source of confusion is that flexibility may be seen as both a set of capabilities (internal) and a source of competitive advantage in a particular environment (external). For the purposes of both manufacturing improvement and the development of manufacturing strategies, it is important to distinguish the capability of being flexible from the competitive need it is intended to match or the advantage derived from it. For example, a company aiming to compete through its flexibility in providing a broad product range may pursue a manufacturing strategy in which production is distributed among a number of focused plants or sub-contractors. Alternatively, it may develop the ability to supply a range of products in one plant ---and build the internal capabilities for producing a broad range in one location. Similarly, a firm may choose to compete on the basis of its ability to provide quick-response to customer orders. It may avoid the need to build manufacturing capabilities to support this external need simply by keeping inventory. An alternative is to build capabilities which allow the manufacturing system to switch effortlessly and quickly between products, avoiding the carrying cost of the inventory and facilitating "just-in-time" production. This internal form of flexibility has been termed mobility(2). The ability to change the product being manufactured quickly, on an on-going basis is the capability which most frequently supports the ability to provide quick response. It is this capability which is the subject of this paper. Specifically, the paper explores the characteristics of plant operation policies, structure, infrastructure and managerial emphasis which support the ability to change quickly among a group of (known) products.(3)
Sixty-one plants in the North American uncoated fine paper industry were studied in depth over a period of two years. This industry is characterized by a growing need for "flexible" operation, as customers demand broader product ranges and just-in-time production. As in many industries, customers have become increasingly reluctant to hold inventory, often pushing inventory back into those suppliers unable to deliver "just-in-time". Not only does this incur additional expense for the supplier, but the proliferation of products often makes inventory-holding an infeasible method of providing quick-response. This is particularly true in the industry described in this paper. To meet these competitive imperatives, firms are anxious to build supporting capabilities at the plant level. To provide a path for improvement, they must determine what characteristics of a plant support flexibility in operations.
The ability to provide quick-response to customers demands capabilities in many areas, as well as than in the manufacturing shop itself. Hammond (1992) describes a number of these issues. There is, however, no substitute for a responsive plant, which can effortlessly switch between products to respond to growing volatility and product proliferation.
Despite the fact that plants in the industry studied employ comparable processes, I find that the differences in changeover times between plants is very large---the fastest plants differing from the slowest by a factor of between 15 and 200, depending on the type of change. I find no observable relationship between the frequency with which a plant performs a particular type of change and its ability to effect that change quickly. However, I find that 40 to 50% of the variance in changeover times across plants is explained by a combination of structural factors, such as scale and technology vintage; infrastructural factors, such as workforce composition and computer integration; and managerial factors, such as the emphasis placed on changing grades quickly.
Section 2 describes the industry in which the field research was carried out while §3 describes the process of changing grades in detail, summarizing the factors hypothesized to be related to the time taken to change grades. Section §4 describes the quantitative research methods and presents descriptive statistics on the data collected while §5 presents some empirical models which are estimated from those data. Section 6 discusses the results, examines some issues in more depth, and offers suggestions for further research.
I define flexibility as the ability to change with little penalty in time, effort, cost or performance. There are many aspects of manufacturing performance which fall under this broad umbrella. The framework used to distinguish the particular manifestation of flexibility explored here is described in Upton (1993). In this paper, the aspect of flexibility I focus on is the ability to change the product being manufactured. The time period over which the changes take place is operational---I am concerned with the ability of a plant to be flexible in making product changes on a day-to-day basis rather than over the long-term. The particular element of flexibility I explore within this context is the mobility of the plant among the products being manufactured, rather than the ability, say, to produce a broad range(4) of products. The motivating question throughout this paper is why some plants are able to change more quickly between products than others. To begin to answer this question, I selected the uncoated fine paper industry, because it has characteristics that greatly facilitate direct comparisons between plants, as described below.
Figure 1 Network Structure
-----
(fig)
-----
A paper mill usually comprises a pulp plant which provides pulp for a group of two to ten converting plants, which turn the pulp into paper. The manufacturing process, industry history and competitive context are each important in understanding the drivers of flexibility.
The papermaking process is a essentially water-removal procedure. Water is removed from a pulp slurry by use of gravity, squeezing and heating. The pulp slurry is laid onto a moving fabric from which water drains off. The pulp web is squeezed and heated by a series of rollers until it is strong enough to support its own weight. Further roller heating then takes place until the paper moisture content is just below that of the ambient atmosphere. The paper is then collected on a reel in the plant to be subsequently sliced transversely and longitudinally into roll-sizes convenient for customers and sheeting machines.
Figure 2 Typical Fine Paper Mill
-----
(fig)
-----
For the past century, converting plants have been increasing in size, output rate and investment requirements. The paper industry has typically competed on cost --- as increasing advantage has been taken of returns to scale in converting plants, the capital investment and size of plant necessary to remain cost competitive have increased. Companies accumulate funds through depreciation and retained earnings, and `leapfrog' the competition with a bigger, more expensive plant, usually on an existing site (the minimum efficient scale for pulp mills exceeds that of converting plants, so a number of plants tend to be co-located with a pulp mill --- incremental expansion is thus considerably cheaper than a greenfield site.) Expansions tend to occur together as firms each identify the same window of market or technological opportunity. This accentuates the natural cyclicality of the industry and is particularly damaging when industry-capacity cycles and business cycles coincide as they did in 1992.
Figure 3 Current Gross Output per day versus Original
Installation Date: Sample of Fine Paper Plants
-----
(fig)
-----
In general, and in downturns in particular, the existence of a few large `commodity" plants(5), and the growing demand for broader variety mean that here are hundreds of plants that find themselves unable to compete on cost, and are forced to stay busy and add value to the pulp by making "cats and dogs." The issue of flexibility improvement, and the ability to switch between grades quickly(6), has thus become an important part of the task of manufacturing managers---indeed, it has become a critical issue for the survival of many small and medium size plants, who can no longer live in a world of long, stable runs. Despite the importance of the issue, there is generally a very unclear understanding of the factors which improve changeover times(7). As the field research later shows, this haziness is understandable: grade change is not a mechanistic process, and there are a large number of factors which act together---each having an effect on the time taken to change between grades.
One of the key difficulties in comparing product changeover times across plants in general, is that the type of change may differ dramatically between plants. In particular, the ease with which a plant can change between products may be strongly dependent on the peculiarities of the process which it uses and the exact nature of the changes effected. These problems are rather conveniently avoided here, by selecting an industry with an inherently simple product, which is always manufactured by the same, fundamental process. This is not to say that the paper industry is representative of manufacturing in general, but rather that it provides a good starting point for examining the sources of those capabilities which support flexible production.
In paper-making, there is a one-to-one correspondence between the process which makes the product and the product itself. Therefore, the ability to change certain process parameters quickly is identical to the ability to change the product being manufactured. In addition, there are only a small number of dimensions along which product characteristics may change. This greatly facilitates the ability to make inter-plant comparisons, as there is a natural control on a number of critical, difficult-to-measure factors.(8)
Data from eleven companies and sixty-one converting plants were collected in this study. Structured and unstructured interviews were carried out at three levels for general background - Vice-President for Operations or COO for the company, plant managers and paper machine superintendents and operators. Detailed structured interviews were carried out with the operators as described later. I also observed operations in 15 plants in the U. S., Canada, Finland and Australia(9) and worked alongside operators on the equipment, to examine the way in which grade changes were made. All managers were invited to attend a one-day conference in which the results of the study were shared. Their comments are implicitly included in the discussion later.
The objectives of the field-research were to gain a detailed understanding of the factors affecting rapid process changes and thence to determine the relative importance of these factors.
A grade change in a paper plant involves the orchestration of the many drives and valves which determine the parameters of the process. The operators' objectives in changing grades are to perform the change as swiftly as possible between acceptable quality levels, while avoiding breaking the paper web. Paper breaks in a machine are very detrimental to plant utilization (as well as being fairly common(10)), and require operators to clamber into the machinery, tear out the broken web and re-thread the machine(11). The time taken to perform a grade change is well-defined. While the machine continues to run, paper made between grades is usually discarded and recycled through the plant as off-standard. Once acceptable quality levels on the new grade have been reached, the reelers again begin to collect paper (see Figure 4.)
Figure 4 The Grade Change Process
-----
(fig)
-----
While almost all plants are fitted with automation which controls process settings for static operation running one type of paper, there is a wide disparity in the degree of automation for the process of changing grades. There is considerable skill associated with the change process and it is as much an art as a science in many plants. Automatic grade changers operate in the upper part of the computer control hierarchy, and automate the transition between grades by gradually changing the process parameters according to a pre-programmed sequence. A number of operators viewed change automation with some scorn! One young operator described how his team had improved the grade change times on his machine:
"The computer is real slow---it get's kind of boring sitting watching [the machine run] so we figured out a better way to change manually. We've got it pretty much figured out now---we all practiced as a team, [...] worked out who should do what and when. We can always beat the computer, and do changeovers faster than they used to do them manually---it's just a matter of working out the right routine.
This kind of experimentation with the grade changing process involves some risk---output targets are often not reached on a shift in which there is a complex paper break, and experimentation is likely to precipitate one.
Plant level observation facilitated the generation of hypotheses concerning the factors likely to affect the time taken to change grades on a machine. These factors fall into four general categories:
- Plant Operations Policies, such as the frequency of grade change
- Plant Structure, such as the scale and vintage of the machinery.
- Plant Infrastructure, such as the experience of the workforce and the degree of computer integration
- Managerial Emphasis, such as the extent to which operators felt they were being encouraged to change grades quickly.
These four general categories are further broken down below.
The primary determinant of the time taken to affect a process change is the dimension along which the process is changed. There are four types of process change in a paper converting plant, listed here in decreasing order of "severity".
- Color Changes
Often require a complete plant wash-down, to remove any traces of the previous color. Few plants in the sample made a variety of colors, and these changes were therefore not explored statistically.
- Furnish Changes(12)
These require changes in the composition of the pulp and often the manual resetting of valves and drives, in order, for example to brighten the paper with Titanium Dioxide or vary the ratio of long-fiber to short-fiber pulp.
- Basis Weight Changes
Basis weight changes involve altering the rate of pulp deposition on the web to vary the area density of the paper. Changes usually require alterations in other plant parameters, such as web speed.
- Caliper Changes
These are minor adjustments to the thickness of the web at the dry end of the plant, and may often not be recorded by the mill.
The two dimensions of change considered in this paper are basis weight and furnish. These are the key dimensions which distinguish one type of paper from another. The changeover times on these dimensions are explored separately.
The time taken to effect a change is likely to increase with the magnitude of the change along the identified dimension. Paper plants recognize this fact and attempt to optimize their grade change sequence by running on a two-week production cycle, in which the machine makes a subset of the possible grades, for different lengths of time, depending on demand. For this reason, the "step-sizes" taken by plants as they move through the grade range is kept small, and plants do not differ greatly, though a few plants do make very large step changes when focused on a few customers and making customized papers. Anecdotal evidence from operators suggests that, unless the step is very large, or very small, it has little effect on the time taken to effect the change. The magnitude of a basis-weight change is numerically measurable. The magnitude of a furnish change is much more difficult to represent numerically, since it involves changing many variables, and adding a wide variety of chemicals and additives to the pulp.
Figure 5 Grade Change Sequencing
-----
(fig)
-----
Plants which make frequent changes might be expected to make faster grade changes. Frequent grade changes are likely to stimulate the plant to adopt operating procedures which improve changeover times. In addition, continual practice at making changes should make the plant faster at effecting them. On the other hand, when `change' is the status quo, it may be that the perceived pressure to get the process up and running again is attenuated. This latter factor seemed to dominate in our interviews:
"This plant makes a lot of customized paper, the kind our customers can only get in a couple of places---we're always changing grades. The main thing is to get [the paper] right when we get there [rather than getting to the new grade quickly]".
Back tender [a standard crew "position" at the end of the plant]
While scale is likely to affect the time taken for grade change, there are two possible mechanisms (and directions) for the relationship. First, increasing scale results in greater mechanical inertia of the components in the system. Tanks may take longer to clear of old grade furnish, and rolls make take longer to speed up and slow down, causing longer change times. However, the pressure that operators impose on themselves to change quickly may increase with the cost of losing production on a large, expensive plant, resulting in shorter change times.
Older, low-level control technology may also inhibit the ability of the plants' operators to change it's process parameters quickly and reliably. There was a wide variety of technology vintages in the plants observed, especially in the basic, mechanical structures. Firms periodically upgrade the motors, valves and low-level control mechanisms on plants to improve efficiency or increase the range of papers that can be made.
Unlike the low-level computer control technology which controls the speed of the drives and the flow rate of valves, computer integration aims to coordinate the various systems in and around the plant at a higher level, performing tasks such as monitoring quality and reporting operating statistics. High levels of computer integration may be associated with faster grade changes, since the process is ---almost by definition---under control. Even though plants upgrade their low-level drives, valves and controls frequently, and are very aware of when these systems are outdated, the level of computer integration is much more a decision based on managers' view of the appropriate application of the technology, rather than a straightforward "upgrade." There are a wide variety of options available concerning the level of integration they use to orchestrate the process; communicate with other parts of the mill; and monitor quality.
The experience of the operators in running the plant is also likely to be significant in the time taken to perform grade changes, though I saw many examples of young, inexperienced operators working quickly and effectively in making grade changes. Indeed, one of the best performing mills in the sample had one fundamental criterion in its hiring process for operators in its new mill: "you must never have worked in the paper industry before." The average experience in this plant was 18 months, compared to the average experience of 17 years.
Operators described a number of different ways in which they perceived managerial priorities. A number of plants emphasized the importance of producing every grade with equal efficiency, while operators at other plants said that the key emphasis was on producing a large range of paper grades or switching between them quickly. The extent to which operators perceived emphasis on each type of performance dimension is also likely to affect the rapidity of grade changes.
Figure 6 Factors influencing Process Changeover Times
-----
(fig)
-----
Grade changes are essentially an infrastructural activity, in that they involve a myriad of small operations, rather than just the flicking of a switch. The time taken to carry them out did not appear clearly to be a result of technology or the equipment being used (except in very highly automated plants). It seemed to be much more due to a combination of a very large number of small activities in which the operators were intimately involved. For this reason, a deterministic "engineering" model of the time taken to effect a change is not feasible, except in those plants in which changes are truly automatic. A wide variety of factors appear to influence the grade change time, although the way in which operating teams perceived the importance of changing grade quickly appeared, at least superficially, to be one of the more significant determinants.
The objective of the remainder of the paper is to determine the relative importance of each of the factors described above. Structured interviews were carried out with operators at each of the plants in the sample. I use these data to estimate a number of empirical models aimed at explaining the very large disparities in change times between plants. The following section describes the variables considered, and presents descriptive statistics on the data.
Operators were asked to estimate the minimum, maximum and mean time taken to change basis-weight and furnish. In some plants, these were cross-checked with plant records. However, plants rarely collect data on grade-change times (though they often collect data on overall non-productive time). I found operators' estimates to be consistent across people working on the same plant; with plant records that did exist; and with the timings I carried out in a small number of plants. The variables I aim to explain are the mean times taken to make changes in basis-weight and furnish as estimated by the operators: BTIME and FTIME. As shown in Table 1, there are very large differences across plants by these measures.
RANGE is the difference between the maximum and minimum basis weights(14) made by the plant in the previous year.
The variable MEANBWT is the mean basis weight at which the plant ran over the previous year.
FREQBWT and FREQFSH are respectively the number of basis-weight and furnish changes made per month .
The variables SPEED (Production Web Speed) and WIDTH (Trim Width) are representative of the scale of the plant. Web speed is the maximum speed at which a plant is run, while trim width is the width of the plant. The variable TONS represents the net output rate of the plant, and hence also reflects its scale. These data are net of off-standard output and broke(15). TONS is likely to be dependent on the way in which the plant is used, and the basis weights it makes, and so is a less reliable indicator of scale.
The variable VINTAGE represents the installation year of the plant or the year of the last major rebuild/update if one has been conducted. The scale of a plant (for example, its width) is related to date of its original installation, but scale effects are captured separately and explicitly, as described above.
Two aggregate variables are used to represent the degree of computer integration on each plant. The extent of computer integration is dichotomized in order to separate two forms of integration. The variable CIMGEN represents the degree of general computer integration not specifically installed for the purpose of improving changeovers. It is constructed as the sum of eight indicator variables in which a value of 1 represents the existence of a particular automation/integration system on the plant. These systems are:
- Mill-Wide Computer System
System to integrate the control computers of all converting plants in a mill, along with those controlling the pulp plant. They allow management to observe the performance of the mill as a whole from one, integrated system.
- Quality Control Computer
A computer system for analysis of trends in quality performance of the plant.
- Downtime data collection
An information system for the collection of data on plant utilization, and reasons for interruptions of production.
- Report system for downtime data
Systems for communicating utilization data automatically to the plant level.
- Cross-directional control of caliper (thickness)
Automatic system for controlling thickness of web in ten or more zones perpendicular to the direction of production.
- Cross-directional control of moisture content
Similar system, but focused on controlling moisture.
- Cross-directional control of basis weight
Similar system, but focussed on controlling area density.
- Digital Control on drives
The use of digital control systems (rather than the older analogue systems) on the motors which drive the plant rollers.
In cases where the existence of a system was unclear or was only rarely used, a score of 0.5 was assigned.
The variable CIMMOB (CIM for Mobility) represents the degree to which changes in grades could be made automatically through computer control of sub-systems. The degrees of automation for changes in basis weight, caliper (thickness) and furnish (chemical composition) were scored out of seven (seven being completely automatic, zero being manual). These three scores were then added and divided by seven, to give a maximum possible score of 3, corresponding to three fully automated methods of changing grade. Automation of color change was excluded since many plants in the sample did not make changes distinguished only by color.
The experience of a plant's operators is represented by the variable SERVICE. This is the mean number of years service (for the company) for the crew of the plant as estimated by the machine manager. While it is possible for operators to gain experience with other companies, the geographical separation of paper sites and low labor turnover rates in the industry mean that this is likely to be a good measure of the experience of the crew.
The following variables were used to represent the view of operators concerning the increase in the degree of managerial emphasis on the following factors:
- Producing a large range of grades (PUSHRNG).
- Producing all grades with equal efficiency (PUSHEQUAL).
- Quick grade changes (PUSHQUICK).
Operators rated the degree to which managerial emphasis had changed over the past five years on these factors. The respondents were also asked to rate the "static" emphasis, (i.e. rather than the change in emphasis, the degree to which the factors were seen as important). These static variables have been excluded from the models, since their variance was extremely low. For this reason, the temporal relative importance is used to represent the degree to which respondents felt management was emphasizing certain dimensions of performance. Ratings were made on a scale of 1 to 7 (7 = greatest degree of change in emphasis).
BREAKS represents the average number of breaks per week, and is therefore a measure of both the stability of the process, and the "quality" of the grade changes which are made. Poorly executed grade changes result in more frequent paper breaks.
Table 1 Variable Sample Characteristics
---------------------------------------------------------------------------------
Variable Description Units N Mean Std. Min Max
Dev.
---------------------------------------------------------------------------------
FMBW Basis Weight Change Time mins. 60 16.30 8.60 3 45
FMFS Furnish Change Time mins. 60 41.85 42.57 1 240
MEANBWT Average Basis Weight lbs 61 74.61 31.60 16.0 174.5
RANGE Machine Range lbs 61 66.80 46.22 0 209
FREQBWT Basis Weight Changes Per 60 30.10 19.81 0 90
Month
STEP RANGE/FREQBWT lbs 59 3.180 2.965 0.233 15.17
FREQFSH Furnish Changes Per Month 61 91.46 227 0 1000
VINTAGE Vintage year 57 1973 20.0 1919 1991
LOGAGE Log (1992-Vintage) 57 .978 .075 0.000 1.863
WIDTH Trim Width inches 60 160.4 66.0 72.0 330.0
SPEED Speed ft./min 60 1457 767 250 3200
TONS Net Output tons/ 61 230.81 248.7 18 1200
day
SERVICE Average Crew Service years 61 17.27 5.33 1 25
CIMGEN Computer Integration 0.0 to 61 5.06 1.92 1.5 8
(General) 8.0
CIMMOB Computer Integration 1.0 to 60 1.87 0.81 0.43 3
(Mobility) 3.0
BREAKS Break Frequency Breaks/ 61 17.3 14.6 1 60
Week
PUSHEQUAL Emphasis on producing all 1 to 7 60 5.25 1.40 2 7
grades with equal efficiency
PUSHRNG Emphasis on range of plant 1 to 7 60 4.90 1.45 2 7
PUSHQUICK Emphasis on quick grade 1 to 7 61 5.10 1.16 2 7
changes
---------------------------------------------------------------------------------
Three categories of empirical model are presented here. The first considers how mobility on the dimensions of basis weight and furnish are related to the operational policies in the plant. Second, I examine the relationship between changeover times and the structure and infrastructure of the plant. Finally, I include the effect of the managerial emphases perceived by the operators in the previous models.
Table 2 Regression Estimates: Change Times and Usage
---------------------------------------------------------------------------------
Basis Weight Change Time Furnish Change
Time
---------------------------------------------------------------------------------
Independent Variables(a)
Basis Weight Range n o o
Basis Weight changes per Month n o o
Furnish changes per month o o n
Mean BWT Step Size o n o
Number in Sample 60 59 60
R2 0.087 0.0002 0.015
---------------------------------------------------------------------------------
- (a)
- o = not included in model; n = included in model.
As described above, the initial hypothesis concerning mobility and usage was that changeover time should decrease with the frequency with which changes of each type were made, but increase with the distance of the changes. The "distance" of a furnish change is not numerical or unidimensional, and is not included in these models.
Regression summaries for a number of models relating basis-weight and furnish change times are shown in Table 2. The table shows that the frequency of change has no observed effect on either basis-weight or furnish change time. The average magnitude of the changes also has no observable effect on change times.
The relationship between changeover times and structural and infrastructural factors were also explored. A linear model for basis weight change time (Equation 1) was estimated from the data, along with a set of similar models for furnish change time (see Tables 2 and 4).
(1) 
.
Table 3 Regression Estimates:
Basis Weight Change Time (BTIME)
-----------------------------------------------------------------------------------------------------
Independent Coefficient Std. Coefficient Std.
Variable (Std. Error) Coeff. (Std. Error) Coeff.
-----------------------------------------------------------------------------------------------------
Constant 24.67 (4.867)(a) 11.556 (7.673)
Speed SPEED - 0.009 (0.003) -0.888 -0.007 (0.003) -0.726
Trim Width WIDTH 0.037 (0.025) 0.323 0.045 (0.023) 0.378
Log (1992- LOGAGE -5.313 (1.777) -0.391 -3.451 (1.830) -0.254
Vintage)
Computer CIMMOB 3.877 (1.853) 0.416 3.165 (1.711) 0.343
Integration
(Mobility)
Average Crew SERVICE -0.373 (0.199) -0.262 -0.234 (0.207) -0.152
Service
Break Frequency BREAKS 0.149 (0.072) 0.268 0.140 (0.064) 0.256
Emphasis on Equal PUSHEQUAL 2.466 (0.721) 0.463
Efficiency across
Grades
Emphasis on Quick PUSHQUICK -2.161 (0.781) -0.329
Changeovers
Emphasis on Large PUSHRNG 1.067 (0.764) 0.165
Process Range
Sample Size 54 54
R-squared (Adjusted R-squared) 0.325 (0.238) 0.503 (0.400)
Degrees of Freedom 47 43
F-Value 3.765 4.836
p (F > F*) 0.004 <0.001
-----------------------------------------------------------------------------------------------------
- (a)
- = significant at 1% level; = significant at 2% level; = significant at 5% level; significant at 10% level.
No statistical significance was found between the general degree of computer integration (CIMGEN) with either of the variables representing changeover times. It is excluded for parsimony. The computer integration variables included in models 1 through 2 are those representing general automation for mobility (CIMMOB).
The strongest determinant of basis-weight change time is the web speed of the plant. Faster machines make basis weight changes more quickly. One possible reason for this is that changes in basis weight propagate more quickly through a fast machine, and therefore take less time. A closer examination of the magnitude of the coefficient shows that this may well be the explanation. The coefficient implies that each 100 feet per minute in extra speed cuts 0.9 minutes from the mean change time (= 54 seconds). The length of a machine would be around 90 feet if this explanation were valid. This number is within the appropriate range, so this would appear to be a sound explanation of the observed result.
The width of the machine is detrimental to the basis weight change time, each 100 inches of trim width being associated with a 3-4 minute increase in changeover time. This might be expected from the anecdotal evidence described earlier: A wider machine means more inertia and hence slower actuation of the process parameter changes, although this result is statistically weak.(16)
Figure 7 Estimated Importance and Significance of
Factors affecting Basis Weight Change Times
-----
(fig)
-----
Older technology, surprisingly, results in faster basis-weight changes. An explanation for this arises from the opinions expressed by some of the operators. Old equipment in the paper industry is very old (often fifty years). The older equipment is far less valuable in accounting terms, and it may be that operators feel safer experimenting with equipment which is less strongly the focus of the organization's "cost" system.
The second strongest effect comes from the degree of computerization installed for making grade changes. While I found general computer integration to be statistically insignificant, it would appear that automatic grade changes are slower than their manual counterparts, in accordance with anecdotal field observation. Automation systems designers are not paid to make equipment that takes risks. Grade change automation systems are designed to work repeatably and reliably. With such an automation system in place, operators may be less likely to consider experimenting with the plant to push the speed of the changes---rather like a plane with an automatic pilot---one may rely on the system so much, that one never learns how to "loop-the-loop".
Crew service does not provide a strong enough statistical result to draw conclusions here. High break frequencies tend to be associated with an increase in the time taken to make basis weight changes, implying that unstable, ill-understood processes are also time consuming to change and that this effect overwhelms the alternative, that longer process changes result in fewer paper breaks because of the care and attention expended on them.
With the exception of LOGAGE, coefficients on all variables described above stay stable with the addition of the managerial variables, implying little interaction between the groups. The addition of these variables into the model increases the explanatory power significantly. The strongest effect in this group arises from the managerial emphasis on producing each grade with equal efficiency. This suggests that the emphasis on "getting it right when you get there" does indeed encourage operators to be slow and careful in changing grades. Emphasis on quick changeovers is strongly associated with rapid basis weight changes, as might be expected.
Figure 8 Importance and Significance of Factors
affecting Basis Weight Change Times (including
Figure 8 Managerial Variables)
-----
(fig)
-----
The direction of causation between managerial emphasis and change times must be considered carefully, since it may be that plants attract managerial attention and operators perceive encouragement in certain ways as a result of the plants' flexibility. However, there are two pieces of evidence which support the view that it is managerial emphasis that causes flexibility, rather than vice-versa. First, few of the plants had any idea that they were slower or faster than the mean in changing between grades. There are few measurement systems in place which inform people how fast or slow a plant is in making changes, so it is unlikely that plants would attract managerial emphasis because of their quick changeover times. Second, the effect I observe and explain through managerial emphasis exists only after controlling for each of the other variables. It is unlikely that plants would attract managerial attention because they have unusually low residuals of changeover time after controlling for the other factors---the effect is too obscure to be observed in a casual fashion.
Models 1 and 2 show estimates of similar models to explain variation in furnish change time. SPEED is statistically weak as an explanator in this model, as are all of the structural and technological variables. Average crew service is, however, significant in both models. The addition of the managerial variables adds strongly to the explanatory power of the model and gives statistically tighter estimates of the parameters. The parameter swings are wide enough to suggest significant interaction between the managerial group of variables and the structural and infrastructural group.
Table 4 Regression Estimates: Furnish Change Time (FTIME)
-------------------------------------------------------------------------------------------------------
Independent Coefficient Std. Coefficient Std.
Variable (Std. Error) Coeff. (Std. Error) Coeff.
-------------------------------------------------------------------------------------------------------
Constant 33.57 (29.50)(a) -79.032 (42.813)
Speed SPEED -0.025 (0.017) -0.441 -0.016 (0.014) -0.264
Trim Width WIDTH -0.059 (0.150) 0.089 0.031 (0.131) 0.044
Log (1992- LOGAGE -5.785 (10.770) -0.073 5.230 (10.211) 0.066
Vintage)
Computer CIMMOB 5.467 (11.232) 0.101 -2.066 (9.549) -0.038
Integration
(Mobility)
Average Crew SERVICE 2.478 (1.205) 0.299 4.674 (1.157) 0.516
Service
Break Frequency BREAKS -0.658 (0.425) -0.204 -0.684 (0.355) -0.213
Emphasis on Equal PUSHEQUAL 11.495 (4.020) 0.368
Efficiency across
Grades
Emphasis on Quick PUSHQUICK -7.734 (4.355) -0.201
Changeovers
Emphasis on Large PUSHRNG 9.487 (4.264) 0.249
Process Range
Sample Size 54 53
R-squared (Adjusted R-squared) 0.267 (0.173) 0.551 (0.457)
Degrees of Freedom 47 43
F-Value 2.849 5.870
p (F > F*) 0.019 <0.001
-------------------------------------------------------------------------------------------------------
- (a)
- = significant at 1% level; = significant at 2% level; = significant at 5% level; significant at 10% level.
Figure 9 Estimated Importance and Significance of
Factors affecting Furnish Change Times
-----
(fig)
-----
The level of computerization for grade changes is also significant but not a powerful explanator. A high proportion of the explanatory power of this model lies with the average crew service: each extra year of crew service is associated with four extra minutes taken to change grade, after controlling for other factors. This is an important and controversial result which will be examined more closely in the discussion, since it has significant implications for plant and network management. Break frequency is negatively related to change time in the case of furnish changes, implying that longer furnish changes are associated with fewer breaks (or vice-versa). The coherence of the web is much more strongly associated with change in furnish than with most basis-weight changes (except at the light end), and so fast furnish changes may be the cause of process instabilities. In addition, furnish changes take longer, on average, than basis weight changes and so the causal link between these two factors may not be obvious enough for the operators to see the link, and correct overly hasty change procedures.
Figure 10 Estimated Importance and Significance of
Factors affecting Furnish Change Times including
Figure 10 Managerial Variables
-----
(fig)
-----
Again, an emphasis on uniform performance across the grade range is associated with larger change times, and emphasis on quick changeovers is associated with smaller change times. An emphasis on producing a larger range of grades is detrimental to changeover performance. This last observation has some important ramifications for the selection of appropriate flexible operating capabilities as discussed below.
The models above show that the only structural variable important in determining change times is the speed of the plant, and this only for basis weight changes. In any case, the direction of the coefficient and the weakness of the effect of plant width contrast considerably with the general view of network managers in the industry, that large plants are less flexible and more difficult to change. Models that used tons output per day (not included here) as an indicator of scale show similar statistically strong evidence that scale alone was an unimportant determinant of changeover times. The lack of any negative statistical relationship between mobility and scale is important. This shows that there is no reason to assume that a plant is less able, purely by virtue of its scale, to produce paper in a "just-in-time" fashion. A plant's measurement system may recognize the cost of the downtime, yet ignore the advantages provided by flexibility. As a plant's rôle in the network gradually changes, its measurement systems and the custom and practice they foster may outlast their relevance. If large and mid-size machines continue to be considered less flexible for efficiency reasons, then many opportunities may be left unexploited. Even though the capital cost of large plants may lead managers to avoid the downtime associated with any change, it is also important to recognize that the value created by just-in-time production may be captured at a much larger scale without disproportionate penalty.
A surprising result from this study is the lack of observable effect of general computer integration and the negative effect of computerized grade-change automation on mobility. Since the flexibility offered by computer automation is one of its most often-quoted benefits this is an important result, though it is consistent with evidence from operators. It suggests that firms should look very closely at the benefits provided by computer automation of grade changes. It would appear that the superficial benefits of reliability and repeatability also rob the plant of its propensity to experiment and find new ways of changing grades more quickly(17).
One of the most important findings in this study concerns the strong negative relationship between furnish mobility and length of crew service. The fact that there is no significant relationship between basis weight change time and crew service helps prune the set of possible explanations. First is the possibility that plant assignments are made on the basis of furnish change time. There is no evidence to support this direction of causation. The second possibility is that it is the crew's long experience (or some unobserved correlate) which is causing the plants to be inflexible. The implications of this result are serious enough to warrant further investigation. To be clearer about the relationship identified, I followed the following procedure. First, Model 4 was estimated using all variables except SERVICE. A linear model of SERVICE was estimated from the same set of variables. Finally, the residuals of each model were plotted against one another to isolate the relationship. This procedure was also carried out using operator estimates of the minimum and maximum changeover times. This is a stringent test of the relationship, and one would expect to see any relationship weakened by the use only of the residuals of the remainder of the model.(18) However, Figure 11 clearly shows that the relationship persists(19). It may be that the operators' age, in concert with the physically arduous task of changing the furnish on a plant makes them slower because of their physical limitations. Attempts to explore this hypothesis using the average crew age were fruitless because of the very high correlation between experience and age in the sample.
FIGURE 11 Residuals of Furnish
Change times vs. Residuals of
Average Crew Service
-----
(fig)
-----
There is also a plausible behavioral explanation. People who learned papermaking at a time when quick changeovers were less important will place a smaller emphasis on its importance. This finding, regardless of explanation, to some extent vindicates the view of some managers in the industry that long service inhibits flexibility.
Another form of flexibility that can be exhibited by a plant is the ability to produce a broad range of products. The relationship between mobility and range is of interest because both are cited as examples of flexibility, and a strong relationship between the two capabilities would allow the identification of common, general characteristics which are shared by all "flexible" plants. Figure 12 shows that the two capabilities are not obviously related to one another. The figure ranks each plant in the sample according to the range of its process and its mobility within the range (both characteristics vary widely across plants). The graph shows each plant's rank on the two flexibility dimensions with plants grouped by firm. The results from model 4 also show that managerial emphasis on plant range is negatively associated with furnish mobility. This underscores a recurrent theme from the broader research project that generated these data: it is not enough to know that a plant needs to be "flexible"---one must also clearly identify precisely what kinds of flexible capabilities are required to support competitive needs, before embarking on a manufacturing flexibility improvement project.(20)
Figure 12 Scatterplot of Rank: Range and Mobility
-----
(fig)
-----
One of the strongest broad effects in both of the models above is that due to the emphasis which operators perceived management communicated to them. The mechanisms for this communication were sometimes overt (through "efficiency" measurement systems) and sometimes more subtle, through tacit approval of the custom and practice in the plant. One clear conclusion from the results, however, is that management makes a difference. This implies that the capabilities necessary for competitive performance can wither or never be built through not recognizing and underlining their importance. However, the capabilities that support flexibility can also be cultured ---by encouraging and emphasizing their importance to the operators in the plant.
In summary, we can draw the following conclusions from the data presented here. The size and computer technology in a plant are not important determinants of its mobility. Even though plant speed may affect mobility, it does so in a direction opposite to that generally suggested by industry opinion concerning scale and flexibility. This suggests that large and medium size plants are considered inflexible more because of measurement systems and industry custom and practice than because of any inherent lack of capability.
Much more important in determining the degree of this form of flexibility in a plant are the people in the plant, and the emphases communicated to them by their managers. This is both bad news and good news: Bad news, in that it would appear to be difficult simply to "buy" flexibility through computer-integrated technology and new equipment. Indeed, the results suggest that one way to foster flexible capabilities in a plant is to allow it to become old enough to escape the attention of the cost system that demands its relentless, "safe" use. It is good news, however, in that the capabilities that support this form of plant flexibility are strongly infrastructural in nature, and, once built, will be much more difficult for competitors to replicate.
The study carried out here explains, at best, 50% of the variance in flexibility across plants. There is still, therefore, a large unexplained variance which seems unlikely to be wholly a result of random error in our observations. The next stage in this research is to explain some of the outliers in our data to determine what other factors make a plant flexible, and to establish the degree to which the above models are misspecified. To do this, the data and models above have been translated into flexibility "benchmarks" for examination by the managers of each company in order to identify any missing factors.
The extent to which these results apply to other industries is of great interest. Data have now been gathered using the same techniques for steel mini-mills. This will hopefully permit the identification of more general truths concerning manufacturing flexibility that can then be used to inform a broader set of managers embarking on flexibility improvement projects.
The development of systematic methods for improving manufacturing flexibility is an important goal, and there is much work still to be done, as noted by Gerwin (1987). Only by developing a clear understanding of what is meant by the capability of flexibility---in a form suitable for manufacturing---and identifying the characteristics of operations that support that capability, can managers identify it, build it and exploit it to provide a lasting competitive advantage.
- References and Bibliography
- Anderson (1993) Shannon Anderson, "Measuring Manufacturing Flexibility: The Impact of Product Mix Complexity on Operating Performance and Manufacturing Overhead Cost," Doctoral Dissertation, Harvard University, Cambridge, MA, 1993.
- Arbel and Seidmann (1984) Ami Arbel and Abraham Seidmann, "Performance Evaluation of Flexible Manufacturing Systems," IEEE Transactions on Systems, Man and Cybernetics, vol. SMC-14, No.4, pp. 606-617, IEEE, July/August (1984).
- Beckman (1990) Sara L. Beckman, "Manufacturing Flexibility: The Next Source of Competitive Advantage," in Strategic Manufacturing edited by Patricia E. Moody, pp. 107-132, Dow Jones-Irwin, 1990.
- Browne et al. (1984) Jim Browne, Didier Dubois, Keith Rathmill, Suresh P. Sethi and Kathryn E. Stecke, "Classification of flexible manufacturing systems," The FMS Magazine, pp. 114-117, April 1984.
- Buzacott (1982) J A Buzacott, "The Fundamental Principles of Flexibility in Manufacturing," Proceedings. 1st International Conference on FMS, pp. 13-22, IFS (Publications) Ltd., Bedford, U.K., 1982.
- Carlsson (1989) B. Carlsson, "Flexibility and the Theory of the Firm," International Journal of Industrial Organization, vol. 7, pp. 179-203, 1989.
- Chatterjee et al. (1984) Arunabha Chatterjee, Morris A. Cohen, William L. Maxwell and Louis L. Miller, "Manufacturing Flexibility: Models and Measurements," ORSA/TIMS Special Interest Conference on FMS, Ann Arbor, Michigan, 1984.
- Cox (1989) Taylor Cox, "Toward the Measurement of Manufacturing Flexibility," Production and Inventory Management Journal, pp. 68-72, First Quarter, 1989.
- Dapiran and Manieri (1983) A. Dapiran and M. Manieri, "The Cost of Flexibility in the FMS," Proceedings of the Second International Conference on FMS, pp. 555-568, IFS (Publications) Ltd., 1983.
- DeMeyer et al. (1989) Arnoud DeMeyer, Jinichiro Nakane, Jeffrey G. Miller and Kasra Ferdows, "Flexibility: The Next Competitive Battle The Manufacturing Futures Survey," Strategic Management Journal, vol. 10, pp. 135-144, 1989.
- Falkner (1986) Charles H. Falkner, "Flexibility in Manufacturing Plants," Proceedings. 2nd ORSA/TIMS Conference on FMS: Operations Research Models and Applications, pp. 95-106, Elsevier Science Publishers B.V., Amsterdam, 1986.
- Garvin (1988) David A. Garvin, Managing Quality, pp. 49-68, Free Press, 1988.
- Gerwin (1987) D. Gerwin, "An Agenda for Research on the Flexibility of Manufacturing Processes," International Journal of Operations and Production Management, vol. 7(1), 1987.
- Ghemawat (1991) Pankaj Ghemawat, Commitment: The Dynamic of Strategy, pp. 109-134, Free Press, 1991.
- Goldhar and Jelinek (1983) Joel D. Goldhar and Mariann Jelinek, "Plan for Economies of Scope," Harvard Business Review, pp. 141-148, November-December 1983.
- Graves and Jordan (1991) Stephen C. Graves and William C. Jordan, "An Analytic Approach for Demonstrating the Benefits of Limited Flexibility," Sloan School of Management, MIT, Working Paper, vol. Working Paper #3297-91-MSA, May 1991.
- Graves and Jordan (1991) Stephen C. Graves and William C. Jordan, "Principles on the Benefits of Manufacturing Process Flexibility," Sloan School of Management, MIT, Working Paper, vol. Working Paper #3296-91-MSA, May 1991.
- Gupta and Goyal (1989) Yash P. Gupta and Sameer Goyal, "Flexibility of Manufacturing systems: Concepts and measurements," European Journal of Operational Research , pp. 119-135, July 1989.
- Gustavsson (1982) Sten-Olof Gustavsson, "Flexibility and Productivity in complex Production Processes," International Journal of Production Research, vol. 22 no. 5, pp. 801-808, 1982.
- Hammond (1992) Janice H. Hammond, "Coordination in Textile and Apparel Channels: A Case for Virtual Integration," Harvard Business School Working Paper 92-007, 1992.
- Hayes et al. (1988) Robert H. Hayes, Steven C. Wheelwright and Kim B. Clark, Dynamic Manufacturing, Free Press, 1988.
- Holusha (1989) John Holusha, "Beating Japan at Its Own Game," The New York Times, July 16, 1989.
- Hyun and Ahn (1992) Jae-Ho Hyun and Byong-Hun Ahn, "A Unifying Framework for Manufacturing Flexibility," Manufacturing Review, vol. 5 (4), pp. 251-260, December 1992.
- Ito (1987) Yoshimi Ito, "Evaluation of FMS: State of the Art Regarding How to Evaluate System Flexibility," Robotics & Computer-Integrated Manufacturing, vol. Vol. 3 No. 3, pp. 327-334, Pergamon Journals, Ltd, 1987.
- Jaikumar (1986) Ramchandran Jaikumar, "Postindustrial Manufacturing," Harvard Business Review, November/December 1986.
- Jaikumar and Upton (1993) Ramchandran Jaikumar and David M. Upton,. " The Coordination of Global Manufacturing", in Globalization, Technology and Competition: The Fusion of Computers and Telecommunications, Edited by Stephen P. Bradley, Jerry A. Hausman and Richard L. Nolan, Harvard Business School Press, Boston, MA., 1993 .
- Jones and Ostroy (1984) Robert A. Jones and Joseph M. Ostroy, "Flexibility and Uncertainty," Review of Economic Studies, pp. 13-32, 1984.
- Kelley (1990) Maryellen R. Kelley, "Plant, Process, and Product-Specific Determinants of Efficiency in the Use of New Technology: The Case of Programmable Automation," Working Paper #90-26, School of Urban and Public Affairs, Carnegie Mellon University, September 1990 .
- Kumar (1986) Vinod Kumar, "On measurement of Flexibility in Flexible Manufacturing Systems: An information-theoretic approach," Proceedings. 2nd ORSA/TIMS Conference on FMS: Operations Research Models and Applications, pp. 131-143, Elsevier Science Publishers B.V., Amsterdam, 1986.
- Kurke (1988) Lance B. Kurke, "Does Adaptation Preclude Adaptability? Strategy and Performance," Journal: Institutional Patterns and Organizations Culture and Environment, pp. 199-222, Ballinger Publishing Company, Cambridge, MA, 1988.
- Lasserre and Roubellat (1985) J.B. Lasserre and F. Roubellat, "Measuring Decision Flexibility in Production Planning ," IEEE Transactions on Automatic Control, vol. AC 30 (5), pp. 447-452, May 1985.
- Liberatore (1990) M.J. Liberatore, "Selection and Evaluation of Advanced Manufacturing Technologies," Management Science, vol. 324 , 1990.
- Mandelbaum (1978) Marvin Mandelbaum, "Flexibility in Decision making: An Exploration and Unification," Doctoral Dissertation, Deparment of Industrial Engineering, University of Toronto., 1978.
- Muramatsu et al. (1985) Rintaro Muramatsu, Kazuyoshi Ishii and Katsuhiko Takahashi, "Some ways to increase Flexibility in Manufacturing Systems," International Journal of Production Research, vol. 23 No. 4, pp. 691-703, 1985.
- Pye (1978) Roger Pye, "A Formal, Decision-Theoretic Approach to Flexibility and Robustness," J. Opl. Res. Soc., vol. 29 No. 3, pp. 215-227, Pergamon Press, Great Britain, 1978.
- Ranta and Tchijov (1990) Jukka Ranta and Iouri Tchijov, "Economics and Success Factors of Flexible Manufacturing Systems: The Conventional Explanation Revisited," The International Journal of Flexible Manufacturing Systems, 2, pp. 169-190, Kluwer Academic Publishers, Boston, 1990.
- Sethi and Sethi (1990) Andrea Krasa Sethi and Suresh Pal Sethi, "Flexibility in Manufacturing: A Survey," The International Journal of Flexible Manufacturing Systems, vol. 2, pp. 289-328, 1990.
- Slack (1983) Nigel Slack, "Flexibility as a Manufacturing Objective," International Journal of Production Management, vol. 3 (3), pp. 4-13, 1983.
- Stecke and Raman (1986) Kathryn E. Stecke and Narayan Raman, "Production Flexibilities and Their Impact on Manufacturing Strategy," University of Michigan Graduate School of Business Administration, vol. Working Paper No. 484, 1986.
- Suarez et al. (1991) F. F. Suarez, M. A. Cusumano and C. H. Fine, "Flexibility and Performance: A Literature Critique and Strategic Framework," Working Paper #50-91 International Center for Research on the Management of Technology, MIT, Cambridge, November, 1991.
- Suarez (1992) F. F. Suarez, Doctoral Dissertation, Massachusetts Institute of Technology, 1992.
- Talaysum et al. (1987) Adil T. Talaysum, M. Zia Hassan and Joel D. Goldhar, "Uncertainty Reduction Through Flexible Manufacturing," IEEE Transactions on Engineering Management, vol. EM-34 (2), pp. 85-91, May 1987.
- Tidd (1991) Joseph Tidd, Flexible Manufacturing Technologies and International Competitiveness, pp. 17-32, Pinter Publishers, London, 1991.
- Tombak and Meyer (1988) Mihkel Tombak and Arnoud de Meyer, "Flexibility and FMS: An Empirical Analysis," IEEE Transactions on Engineering Management, vol. 35 (2), pp. 101-107, May 1988.
- Upton and Barash (1984) D. M. Upton and M. M. Barash, "A Grammatical Approach to Routing Flexibility in Large Manufacturing Systems", Journal of Manufacturing Systems, vol. 7 (3) pp. 209-221 1988.
- Upton (1993) David M. Upton, "The Management of Manufacturing Flexibility," Working Paper 93-046, Harvard Business School, Boston, MA, January, 1993.
- Upton (1993b) David M. Upton, "Computer Integration and Catastrophic Process Failure in Flexible Production" Working Paper 93-073, Harvard Business School, Boston, MA, May, 1993.
- Waldner (1990) Jean-Baptiste Waldner, CIM: Les Nouvelles Perspectives de la Production, pp. 38-39, Bordas, Paris, 1990.
- Webster and Tyberghein (1980) Dennis B Webster and Michael B Tyberghein, "Measuring Flexibility of job shop layouts," International Journal of Production Research, vol. 18 No. 1, pp. 21-29, 1980.
- Williams et al. (1986) D. J. Williams, P. Rogers and D. M. Upton, "Programming and Recovery in Cells for Factory Automation," International Journal of Advanced Manufacturing Technology, vol. 1 (2), pp. 37-47, IFS (Publications) Ltd., 1986.
- Zelenovic (1982) Dragutin M. Zelenovic, "Flexibility--a condition for effective production systems," Int. J. Prod. Res., vol. 20 (3), pp. 319-337, Taylor & Francis Ltd., 1982.
- Zuboff (1988) Zuboff, Shoshana, In the age of the smart machine : the future of work and power, Basic Books, New York, 1988.
-------------------------------------------------------------------------------------------------------------------------------------------------------------
CIMGEN CIMMOB RANGE MEANBWT LOGAGE PUSHRNG PUSHEQUAL WIDTH SPEED PUSHQUICK BTIME FTIME FREQFSH FREQBWT BREAKS SERVICE
-------------------------------------------------------------------------------------------------------------------------------------------------------------
CIMGEN .613 -.430 -.325 -.326 -.073 -.126 .583 .734 -.253 .035 -.141 .274 -.177 .269 .154
CIMMOB -.458 -.500 -.219 -.033 -.056 .507 .712 .105 .037 -.132 .166 -.091 .355 .410
RANGE .903(a) -.078 .136 .202 -.531 -.423 .102 .097 .383 -.187 .308 -.117 .054
MEANBWT -.137 .001 .243 -.421 -.354 .088 .043 .306 -.181 .282 -.161 .047
LOGAGE .316 -.223 -.288 -.272 .007 -.274 -.019 -.154 -.224 -.117 -.335
PUSHRNG .021 -.201 -.104 .178 .029 .249 -.120 -.128 .011 -.019
PUSHEQUAL -.349 -.302 .363 .481 .335 .116 -.072 .028 -.021
WIDTH .802 -.120 -.113 -.254 -.075 -.095 .159 .271
SPEED -.050 -.204 -.367 -.013 -.085 .286 .182
PUSHQUICK -.095 -.027 -.198 .083 .049 .003
BTIME .282 .346 .040 .250 .047
FTIME -.112 -.162 -.267 .416
FREQFSH .356 .645 .004
FREQBWT .368 -.092
BREAKS .063
-------------------------------------------------------------------------------------------------------------------------------------------------------------
- (a)
- This correlation is functionally induced since each is deterministic function of the same set of sample variables.
53 observations were used in this computation.
8 cases were omitted due to missing values.
FIGURE 3 Residuals of Furnish Change times vs. Residuals of Average Crew Service
Footnotes
- (1)
- Beckman (1990), DeMeyer et al. (1989), Holusha (1989), Goldhar and Jelinek (1983), Zelenovic (1982).
- (2)
- See Upton (1993)
- (3)
- The ability to switch quickly to novel products is also a form of mobility, but is not addressed in this paper.
- (4)
- This is explored in other papers resulting from this project.
- (5)
- These plants tend to make 18-24lb. Xerox paper.
- (6)
- Plants must try to minimize this, even when running `slack', to balance pulp flow and continue to add value to it. The pulp may be sold externally, but drying and transportation costs make it much more advantageous to make paper with it on-site.
- (7)
- Changeovers absorb between 2% and 20% of productive time, depending on the plant.
- (8)
- See Anderson (1993) for an approach to poly-dimensional product attributes in the textile industry.
- (9)
- All of the numerical data here come from US and Canadian firms. This is being pooled with comparable work being carried out in Brazil and Finland for inter-country comparisons.
- (10)
- Paper breaks occurred, on average, 17 times each week per plant in the sample.
- (11)
- This takes between 30 minutes and four hours.
- (12)
- Some furnish changes are simpler than some basis weights changes, so there is considerable overlap in the "severity" of these two types of change.
- (13)
- Auxiliary variables are presented in the descriptive results, but are not included in the regression models presented later.
- (14)
- Measured in lbs/standard area
- (15)
- Waste paper recycled through the plant.
- (16)
- Pairwise correlations are shown in Appendix I. Because of the strong collinearity of the speed and the width of the plants, models were also explored using the projections of the sample on the eigenvectors of the joint distribution of SPEED, WIDTH and TONS. The variance explained by the primary eigenvector (in direction [+,+,+]) was small, which is consistent with the conclusion that scale is an unimportant determinant of flexibility. The remaining orthogonal eigenvectors explained more variance, but what their directions represent is much more obscure. The results have therefore been presented in their original form. The collinearity of general computer integration with scale was also explored as a possible reason for its statistical insignificance.
- (17)
- Further research has shown that computer-integrated change systems may actually be more likely to precipitate paper breaks (see Upton (1993b)).
- (18)
- Thanks are due to Kim Clark for this suggestion
- (19)
- there is still the possibility that the model has been misspecified and that there is some unobserved positive correlate of both variables.
- (20)
- These data also show clearly the multi-plant flexibility strategy of each firm. This was of great interest to the firms participating in the studymost were generally unaware of their position on this "capability map." Most agree that their position on the map was roughly correct, though not necessarily where they wanted it be! The formation of multi-plant flexibility strategies is currently being researched.