You are on page 1of 20

MANAGEMENT 8CIEKCB

Vol, la, No, 10, June, 198»


Printeiin V.S.A.

THE FACILITIES LAYOUT PROBLEM IN PERSPECTIVE*


THOMAS E. VOLLMANNt AND E L W O O D S. B U F F A J

The relative location of facilities in a functional layout has been deter-


mined under the criterion of materials handling cost minimization. Recent
advances in analytical methods have provided superior solutions within thie
criterion hut the validity of certain underlying assumptions must be ques-
tioned. Materials handling as a criterion is deficient for many facilities layout
problems; yet a modified definition of "materials" may yield considerable
insight to the improved location of people. Each layout problem is made unique
by its particular assumptions, constraints, limitations, and the intrinsic activ-
ity of the components; a guide to establishing this iiniqueness and to selecting
appropriate tool3 for analysis is presented.

Where should the drill presses be placed relative to the engine lathes? What
is the best location, for the tool crib? Where should the industrial engineering
departnaent be located in our plant? How close should the accounting department
be to the purchasing department? Where should remote input-output stations
be located for a centralized computer? Where should we put the new numerically
controlled jig boring machine? Where can we put it?
These are typical problems faced by managers every day. How are these prob-
lems solved? Is there a better methodology that will provide improved solutions?
What are the long run implications of various facilities layout solutions? Are all
layout problems similar? Can we a priori expect to use the same methodology
for solving every layout problem? What are the necessary conditions for deter-
mining applicability of a particular methodology to a particular problem? Have
these conditions been met for the problems at hand? More basically, has the
question of the applicability of generalized layout models to particular problems
even been assessed?
The issue
The use of formalized layout approaches requires the acceptance of gross as-
sumptions. Cost and flow data are assumed to exist for definitionally imknown
conditions. Materials handling costs are assumed to be linear, incremental, and
assignable to specific activities. Materials handling cost is considered to be ihe
only significant factor in most layout models. Flow data with stochastic proper-
ties are assumed to be deterministic. The interaction of other system problems
with the layout problem is ignored. Approaches are frequently adopted or advo-
cated without adequate attention to the compatibility of the problem situation
with the model assumptions.

* Received November 1965.


t Amos Tuck School of Business Administration, Dartmouth College, Funds for parts of
this research were derived from a grant by the Alfred P. Sloan Foundation administered
by the Tuck School,
t Graduate School of Business Administration, University of California, Lo8 Angelea.
B-450
THE PACILITIBS LAYOUT PHOBLEM IN PERSPECTIVE B-451

The purpose of this paper is to investigate critically the underlying assump-


tions of formalized approaches to facilities layout. The intent is to examine the
operational implications of assumptions and to examine the effects of utilizing
other plausible a-ssumptions as postulates. In the process of investigating the
layout problem, it is the criterion of materials flow itself which we see as the
focal point for necessary revision. In this pursuit, the following are representative
of some questions which require answers:
1. What is the operational nature of materials flow as a criterion, i.e., how is
the criterion to be measured?
2. What is to be considered as the equivalent of materials in various situations?
3. What data are required for use in the various approaches to layout prob-
lems?
4. What are the lim^itations, sources of bias, or errors in such data?
5. What assumptions are inherent in utilizing different criteria?
6. Under what conditions would these assumptions be reasonable?
7. In what empirical situations are various criteria most applicable? Least
applicable?
In pursuit of answers to these questions, other questions arise which generate
still more. It is not the intention to prove or disprove the validity of any existing
model. Rather the question is, what are the limitations, assumptions, and omitted
variables in any technique and what are their ramifications in various situations?
Current Status of the Layout Problem
Traditional approaches
The traditional approaches to facilities layout problems seem to follow a
common path. First, layout problems are categorized as manufacturing or non-
manufacturing, product or process, and initial or relayout. The non-manufactur-
ing layout is frequently dismissed as being analogous to manufacturing layout,
with recommendations to adapt the tools of the one to the other. Layout by
product is typically considered elementary insofar as the location of the various
components is concerned because of the severe restrictions arising from se-
quential constraints; the problem of balance assumes primary consideration in a
line layout with the assumption of standardized methods. Conceptually, relayout
is considered to be the same as initial layout with added constraints (e.g., existing
offices, buildings, service facilities, etc.) which, however severe, are of a imique
nature. This traditional approach leads to the selection of the initial, manu-
facturing, job shop layout as the most complex case, with other types being
somewhat ancillary to it.
The measure of effectiveness for layout problems traditionally has been con-
cerned with the flow of materials. Changes in manufacturing methods have
complicated the determination of this flow and new techniques have evolved
for better examination of complex flow situations. The criterion for quantitative
layout models is now frequently stated as the minimization of materials handling
cost, which is assumed to be an incremental linear function of the distances
between the components of the system under study.
B-452 TH0MA8 E. VOLLMANN AND BLWOOD S. BUFFA

Graphic and schematic analysis


The most familiar tools utilized for the solution of layout problems have been
graphic and schematic models, particularly two-and three-dimensional tem-
plates, assembly charts, operation process charts, and product flow process
charts.
Recent improvements in graphic and schematic layout techniques include
"link analysis," "travel charting" and "operations sequence analysis." These
techniques utilize data collected on the amount of materials flowing from each
department to every other department for some time period. The data are ac-
cumulated into a matrix form which is frequently referred to as a "from-to
chart," "cross chart," or "fiow matrix". The methodology for reducing materials
flow is to locate departments in such a way so as to minimize the volume of non-
adjacent departmental flow.
Link analysis, travel charting, and operations sequence analysis, albeit great
improvements over preceding methodology, become virtually unmanageable
when the nmnber of departments becomes at all large (say above 10) unless the
flow has a dominant pattern.
CRAFT iComputerized Relative AUocatim of Facilities TechniqueY
The CRAFT model represents a breakthrough in the abihty to systematically
evaluate layout problems having complex flow. The superiority of results ob-
tained by earlier techniques was in no small way dependent upon the art of the
layout analyst. The CRAFT model can generate and evaluate layouts having
as many as 40 departments in less than one minute of computer time. In addi-
tion, any department or departments may be fixed in particvdar locations.
The inputs to the CRAFT model are similar to those used in link analysis,
travel charting and operations sequence analysis. CRAFT takes interdepart-
mental flow data and weights them by materials handling cost data; the result
is a cost weighted flow matrix. The only additional input is an initial block lay-
out which may be the existing layout or any proposed layout. The computer
determines the centroid of each department and measures the distances between
centroids on rectangular coordinates. The result is a distance matrix which is
multiplied by the cost weighted flow matrix to determine the materials handling
cost for that particular block layout.
The governing heuristic algorithm then looks for the largest potential materials
handling cost reduction obtained by switching the location of any 2 or 3 depart-
ments. The change so indicated is made, the new block diagram is printed out,
and the procedure is repeated until a block diagram is produced which cannot
be improved.

»8ee Buffa, E. 8., G. C. Armour, and T. E. Vollmann, "Allocating Facilities


with CRAFT," Barvard Biuiness Review, March-April, 1964, pp. 136-168 for a more detailed
explanation of this tool. One of the authors is presently preparing another article which is
an examination and comparison of CRAFT with other techniques (see Hillier, Frederick
S,, "Quantitative TOOIB for Plant Layout Analysis," Journal o/ Industrial Engineering,
Vol. XIV, No. 1, Jan.-Feb., 1963 pp. 33-40.
THE FACILITIES LAYOUT PROBLEM IN PERSPECTIVE B-453

Systematic layout planning^


This technique only utilizes graphic and schematic analysis for materials flow
but does present a weU-organized enumeration of pertinent qualitative factors.
These factors are incorporated into a methodology for the determination of
which sub-groups or departments most require adjacent placement as well as
for the evaluation of several proposed layout solutions.
Layout Model Assumptions
Cost data
Heuristic layout model cost inputs are necessarily linear materials handling
costs which vary as a function of the distance between departmental centroids.
Pick-up and put-down time are considered non-incremental. Under what con-
ditions is materials handling cost incremental and linear as a function of depart-
mental centroid spacing?
In the machine shop example examined by Buffa, Armour, and VoUmann", two types
of cost conditions existed, the first being when machinists were utilized for materials
handling and the second when specialized materials handlers were used. In the machinist
case, the use of linear and incremeatal materials handling coats assumed first of all that
if a machinist was handling materials he was not running his machine (a valid assump-
tion in most instances). Second, an aasumption was made that if a machinist was not
engaged in this particular non-productive work he in fact would have been engaged
in productive work rather than some alternative non-productive activity. It also
was necessary to assume that incremental amounts of work were available for any
time saved, or conversely that any time wasted due to poor department spacing caused
extra costs to be incurred, e.g., extra people hired, overtime, etc.

The validity of some of these assumptions may be questioned. From a systems


viewpoint, any machinist may be considered as a production system component
with certain inherent queuing aspects. That is, a certain amount of slack is pur-
posely built into the system (necessary in any queuing or waiting line model)
as a cost tradeoff for shorter throughput time, fewer bottlenecks, etc. At a given
point in time, it may be possible to obtain some benefit at no apparent incre-
mental cost during slack time, although such work must itself not be severely
imposed with time restrictions or the total system effectiveness may suffer. Ma-
terials handling wotild seem to have such time restrictions, i.e., if workers are
to do materials handling only In off moments, better worker utilization well may
be obtained at the expense of incurring scheduling problems.

' Muther, Richard, Systematic Layout Planning, Industrial Education Institute, Boston,
1961. A somewhat abridged version of this technique may be found in "Simplified Systematic
Layout Planning," Factory, August, September, and October, 1962.
»In addition to the two cost possibilities, twoflowpossibilities were examined creating
four alternatives. Theflowpattern could either be directly to the department of the next
operation or via a central hold area (Department K). The four alternatives are therefore
labeled K Machinists, Non-K Machinists, K Materials Handlers, and Non-K Materials
Handlers.
B454 THOMAS B. VOLLMAim AND KLWOOD S. BtJFFA

In tbe specialized materials handler case, problems with the linearity assumption are
apparent. A company cannot go out and purchase X dollars worth of materials handling
capacity when it is needed. Materials handling costs would necessarily follow a step
function. The relatire magnitude of the steps obviously varies from situation to situa-
tion.

Conceptually, the step phenomenon makes many different layout possibilities


equivalent in materials handling cost. Complicating any particular situation
are the above mentioned queuing aspects of the materials handling problem and
the possibility of readily postponable fill-in work. Thus, for example, many
companies utilize materials handlers for janitorial activities and vice-versa. The
magnitude of postponable fill-in activities, of course, varies greatly from com-
pany to company due to such factors as union considerations and scheduling.
In specific ceises it would be necessary to obtain some feeling for the magnitude
of error when this st«p function is approximated linearly.*
Flow data
The typical source of flow data advocated by most layout texts is the opera-
tion sheet or operation process chart. These documents basically provide se-
quence data which can be accimiulated into various forms, including the cross
chart or CRAFT flow matrix. Detailed operation sheets for a proposed plant
indeed may not even exist I Moreover, since the basic rationale for the job shop
mode of operation is flexibility (which may be defined as the ability to readily
adapt to changes in environment), the usual assumption that future work will
have routing similar to present work well may be unjustified.
To some extent, this problem is aggravated by the data collection process it-
self, or more specifically, by the design of the data collection process. Data col-
lection seems to end up with one set of data, e.g., a cross chart. Such a cross
chart may state that for a given time period 57 loads flow from Department B
to Department C: not 56 and not 58 but 57. All matrix cells in the cross chart
bave this characteristic. To treat such highly stochastic or variable data as de-
terministic or fixed is clearly an oversimplification. The results from such anal-
yses tend to be of a "pinpoint" nature with no real assurance of validity if any
interdepartmental flow or flows differ by some amount.
With this problem in mind, final solutions for the above mentioned machine shop
example were examined after randomly altering the flow data by as much as plus or mi-
nus 50%. In the interest of brevity, the detailed analysis has been omitted here. The
results of theBe simulated flow adjustments indicate that this particular final solution
is relatively insensitive to random changes in flow. (The reasons for this particular in-
aenaitivity will be examined in a later section).

* Other coat data assumptions which may be questioned include cost assignment (is the
process of assigning costs to particular loads necessarily arbitrary?); cost significance (is
materials handling cost that varies as a function of the distance between syatem components
important relative to other costs?); and cost weighting (are proposed materials handling
costs to be dictated by present methods?). These assumptions are examined in more detail
in a monograph which is currently in preparation.
THE FACILITIES LAYOTJT PROBLEM IN PEESPBCTIVE B-455

A related question concerns the impact of non-random changes. Operationally


this question might take the form of determining what would happen if produc-
tion of Product X were to be changed by some factor or percent.
A large aircraft manufacturer is currently concerned with this problem when attempt-
ing to plan a layout under unknownflowconditions. It is expected that the problem will
be handled by asking what type offlowfluctUELtionacan be expected. Instead of collecting
data for one set of circumstances, data will be collected with the following questions in
mind:
1. What is the distribution of "educated guessea" for the number of aircraft per time
period?
2. What is the spectrum of conceivable design changes, make-buy decisions, fabrica-
tion changes, etc.?
Utilizing this type of data in a simulation experiment may provide results that would
be very good for a large number of states of nature. In addition, knowledge would be
gained of the critical range of various factors, thereby establishing a management by
exception criterion for subsequent adjustments. The advantage of such results over
"pinpoint" results is obvious.
A similar project was deemed feasible for a large manufacturer of rocket engines. The
basic idea was to correlateflowdata to some familiar index of activity such as man hours
worked in a department. Since management was familiar with departmental man hour
requirements for various courses of action, these data could have been used for artificial
generation of flow data to simulate the ramifications of various proposed contracts on
layout.

A factor relat«d to the general consideration of stochastic flow data is flow


dominance. To what extent does a given flow matrix possess dominance? Opera-
tionally for facilities layout, the question may be expressed as: How sensitive is
the final layout to changes in the flow matrix? Does a dominant pattern exist
which effectively allows other flows to be ignored?
Two extremes can be identified. At one extreme is complete dominance, e.g.,
a tube drawing facility, where all work irrespective of size or shape goes through
a prescribed set of operations. At the other extreme, one can conceptualize a flow
matrix with identical values in all matrix cells (except for zeros in the main di-
agonal). In the complete dominance case, flow is obvious and materials handling
can be minimized by graphic or common sense approaches to the layout prob-
lem. At the other end of the continuum no "one best way" exists. Changes in
layout can be made indiscriminately with little or no effect on the objective
fimction.
The obvious operational question is whether a particular layout problem has
a flow pattern which is dominant enough to negate the need for a sophisticated
analytical tool. How is dominance to be measured? The determination of rela-
tive dominance for flow matrices requires a measurement which is independent
of the raw data magnitude; the coefficient of variation in some ways satisfies
this independence and has been selected for this purpose.
To illustrate, four layout matrices were aaalyjsed for flow dominance yielding the
following coefficients of variation: 135%, 201%, 252%, and 519%. The four examples
were laid out on an intuitive basis after inspection of the flow matrices; the resulting
layouts were used as initial inputs to the CRAFT model. The model was able to reduce
B-456 THOMAS E. VOLtMANN AND ELWOOD 8. BUFFA

the objective function value for the 135% example by 12%, the 201% example by 2%,
the 252% example by 5% and the 519% example by 3%.»
If these results are accepted as indicative, it would seem fair to estimate that
any particular layout problem having flow matrix data with a coefficient of var-
iation in excess of 200 % can probably be solved by inspection of the flow matrix
or some other approximation technique and be within 5 % of the CRAFT func-
tion value."''
Departmentaticm
The relative location of facilities problem assumes that each facility unit or
department is specified. That is, in an operations system, the components are
grouped into various subsystems without the layout question entering into the
grouping process itself. Departmentation (and organization in general) is not
static. Improving technology and the learning phenomenon provide better
methodology for the accomplishment of tasks. Changes in work methodology
well may call for different departmentation bases. In the machine shop the com-
mon characteristic of nimierical control has caused regroupings of many tra-
ditionally diversified machine operations. Other equipment improvements have
created a working environment necessitating less direct control of workera. Im-
proved commuiucation devices including computerized information retrieval
may change office layout patterns.
Traditional approaches to the layout problem treat department changes under
the concept of flexibility, stating that the layout should be flexible enough to
accommodate evolutionary changes. It would seem that the importance of the
layout problem to departmentation and oi^ganization should be included in any
thoroughgoing layout analysis. In fact, the information generated from a layout
study may be indicative of needed organizational changes and thereby facilitate
oi^anizational strength.
Philosophical Layout Considerations
Plant layout as a systems problem
The plant layout problem overlaps with other traditional problem areas. The
critical nature of any one problem area is dependent in no small way upon the
degree of sophistication with which other problem areas are handled. Thus a
superior production control operation can create increased tolerance for poor
layout. Conversely, good layout can tolerate relatively inferior production con-
trol, quality control, maintenance, etc. Good relationships between the functional
areas of sales and production can also ameliorate certain problem areas, e.g., the
' Brevity here dictates the omission of a detailed investigation of these four examples.
For the interested reader, the 519% example is the Non-K Handlers machine shop example.
* Large numbers of departments, unusual differences in areas, and practice will naturally
influence results obtained by intuitive methods.
' Another problem with a flow data assumption arises from the way materials flow is
grouped into loads. This assumption is again treated in more detail in the author's above-
mentioned monograph.
THE FACrUTIEB LATOTJT PROBLEM IK PERSPECnVB B-457

impact of changes in demand may be dampened. The same generalization is


true for the finance area, e.g., discounts and credit terms can be adjusted to
benefit other functional area needs.
A similar systems problem is concerned with the overall utilization of the
production system. Overcapacity of ^stem components as a subproblem solu-
tion has already been considered. Most companies similarly have some degree
of overcapacity in the total system. In a full capacity situation, physical size
is a limiting factor of the plant operation, and the layout problem as a systems
subproblem may become relatively more pressing. Conversely, as the level of
activity diminishes in an operations system, more effective space is available.
The criterion of space utilization plays an entirely different role in the two ca-
pacity situations.
A related consideration is the significance of the time dimension. From a sys-
tems viewpoint, one important time consideration is the layout change impact
on system effectiveness. The seriousness of this problem is linked to the utiliza-
tion intensiveness of the system. Thus, for example, a large manufacturer of
television picture tubes must work within the constraint of interrupting pro-
duction for layout changes only on weekends and during an annual three-week
shutdown.
Lang-run verms short-run layout problems
Frequent reference is made in layout texts to the concepts of expandability
and flexibility. Expandability is a rather subjective factor which basically im-
plies that layouts should provide for future expansions with minimal cost. The
even more nebulous concept of flexibility implies that layouts readily should be
able to accommodate future changes in product mix, demand, and technology.
It ah^ady has been pointed out that the common use of past data for fore-
casting orders, machine time, etc. in a deterministic manner cannot satisfy the
objective of flexibility. Expandability usually is approached by modular con-
struction, knock-down walls, accessible utilities, and overcapacity of employee
services, to name a few physical methods of providing slack in the system. Al-
though expandability and flexibility differ in goals, it can be seen that they have
an important factor in common. They both imply that layout is not a single or
one-time decision; adjustments will be necessary. The inclusion of the time di-
mension as a factor might be better conceptualized by a simple dichotomy of
long-run short-run.
In a long-run problem facilities change from one configuration to another hy
means of a single major layout change performed at one point in time. On the
other hand, the short-run change is made whenever significant parameters of
individual system components are altered. The major or long-run layout change
will be hereafter referred to as the major case and each of the series of short-run
layout changes as a minor case layout change.
Most written works on facilities layout have focused on the major case, while
in practice the majority of layout decisions are of the minor case. The difference
between the two cases is analogous to the dichotomy between the production
B-458 TBOMAS E. VOLLMANN AND ELWOOD S. BDFFA

system considered as a single monolithic system instead of as a collection of


component subsystems. It has been recognized that the criteria for and con-
straints of the production system as a single unit are different from the criteria
and constraints oi each component subsystem. Similarly the criteria and con-
straints applied to the major layout case are different from those of the minor
layout case. There is no reason to assume that the result of a major case layout
is the same as the net result of a series of minor layout changes. Symbolically this
could be stated as

\M} ^ {mitrh- • -mn) (Both on criteria and on constraints)


Where A is an initial layout configxuation, M a major case transformation,
wJim2 • • nu, & series of minor case transformations, and B and B' resulting layout
configurations.
In the major case, the system boundary constraints are much less rigorous
than those for the minor case. There is no question that the two layout cases are
related; in most instances it will be difficult to precisely define a dividing line.
One large difFereace between major and minor case layouts is the determina-
tion of the number of departments to be considered as possible candidates for
replacement. The location to which a department can be moved is typically de-
cided on a space arailable basis in the minor case; in the idealized major case,
no location is sacrosanct. The usual asstunption of materials handling costs is
that no incremental cost will be incurred for the minor case layout change. The
major case typically assumes linear as well as incremental materials handling
costs.
Another implication of major versus minor case layouts may be thought of in
t€nns of the time period for which data are collected. Should a lai^e sample be
taken over a long period to effectively get an average fiow? How much would
costs be reduced if the layout were changed more frequently?
In an attempt to teat this implication of minor case layout changes the machine shop
example considered above was analyzed further; additional flow data were collected for
29 weekly periods. Using the final solution to the Non-K Materials Handlers situation*
as an input or startLug layout, each weekly flow situation waa evaluated; conversely,
the starting solution was evaluated on the basis of actual flow data.
The results indieated that the selected starting layout was relatively adequate for
all the weekly flows considered. The largest amount that the CRAFT model was able
to reduce the initial objective function for any one week was 8%; the average reduction
was less than 4%. Over 29 weeks, the value of the objective function varied from $193.37
to $451.97 (135% change); yet changes from CRAFT improvements averaged less than
4%. When the 29 weeklyflowswere accumulated into oneflowmatrix, the initial layout
waa improved by only .1%.

What can be generalized from this one example? To some extent the stability
of the Non-K Handlers situation is predictable because of the relatively high

»See Footnote 3.
THE FACILITIES LAYOOJ PBOBLEM IN PERSPECTIVE B-459

flow dominance (519%). In the context of random flow changes it has been
found that flow dominance allows the impact of these changes to be ignored. In
addition, it now would seem that once a superior layout has been determined
for a donvinant situation, little change is required as long as the basic dominance
prevails.
In terms of layout in general, the question arises as to whether this case is a
typical or atypical machine shop. It is these authors' opinion that every layout
problem is unique. The assertion that a layout problem is "typical" is about as
meaningful as stating that on the average a company always meets its delivery
dates. This statement is unfortunately true when 50% of the orders are early
and 50 % are late. The uniqueness of layout problems is generally due to the
assumptions, constraints, and conditions of any one situation.
A particular facet of the machine shop example may serve to highlight this unique-
ness. The flow data have been found to be highly dominant. The primary cause of this
dominance can be identified as stemming from three high flow departments: Deburring,
degreasing, and inspection. The key point here is, however, that the highflowarises from
routing conventions, i.e., jobs are routed to these areas after almost every manufacturing
operation.
These conventions can readily be changed by policy decisions, e.g., roving floor in-
spectors rather than a centralized inspection facility. A potential impact of a change in
these policy decisions would be to reduce drastically materials handling cost as presently
stated and to change the best relative location of departments. Clearly some assessment
of inherent data bias is called for in aay particular problem.

A modified materials handling cost criterion


At this point it would be useful to examine a situation where a variation of
the materials handling criterion in a non-manufacturing situation may be prof-
itably employed.
Before this example can be presented it is necessary to broaden the typical
definition of materials handling. The usual distinction would focus attention on
the parts or product in a manufacturing layout, on paper work in an office layout,
or on patients in a hospital. Such an assumption is excessively constrained since
in many instances (e.g. hospital patients) the operations function is carried out
to a large extent by bringing the services to the materials rather than the reverse.
Materials handling operationally defined for relative location of facilities prob-
lemis should then be concerned with any cost that varies incrementally as a func-
tion of the distances between components of that system. This definition would
immediately include trips to tool cribs, personnel offices, supervisors, or any
contact requisite to performing a job. If we view such contacts as being for the
purpose of transmitting and receiving information, the resultant layout may be
deemed layout by information requirements.
With a given flow matrix for information needs, an appropriate cost matrix
can be constructed utilizing hourly wages per employee and the traditional
industrial engineering standard of three miles per hour to compute the various
costs for moving some unit of distance. The two types of data can be coupled
with the existing layout and analyzed by a materials handling model, e.g.,
CRAFT.
B-460 THOMAS E. VOLLMANN AND BLWOOD S. BUFFA

It should be pointed out that when a layout analysis is made on the basis of
information needs as outlined above, the cost is determined by the walking time
of the individuals concerned. The cost savings so generated would be extremely
conjservative, as no account ia taken of the benefits accruing from more eflFective
commimication and organization. It would seem reasonable to inflate any sav-
ings by some factor such as two or three to more fairly present anticipated sav-
ings. The savings indeed may not take the form of fewer employees but rather
be obtained by getting more work through a given facility in a given time, or
the same work through in less time.
An example utilizing this expanded concept of materials handling coet was done in
an industrial planniDg ofllce of a large aircraft plant in Los Angeles. The approach uti-
lised in this layout project seemingly was to ignore paper flow. On closer examination,
however, emphasis was placed on the strategic elements. The method of analysis was
to focus attention on the movements of personnel in the department, whether in the
transportation of paper work or in other communications. The way these data were
collectecl was to ask each person in the area to record the face-to-face contacts he initi-
ated with people in the other groups. The data thus gathered became the flow data for
the CRAFT model when accumulated into matrix form. Cost weighting data were ob-
tained by utilizing the workers' wages.
The industrial planning department had the responsibility for company-wide layout
problems. It therefore followed that considerable thought had gone into the factors de-
termining its own internal arrangements. The final results indicated a relatively small
potential improvement (10%), which did not detract from the usefulness or applicability
of the approach. The improvement magnitude did signify that the initial layout was
good (a reasonable expectation for layout specialiBts). A more significant result was the
feeling of these specialists that the basic approach was sound and did a superior job of
handling complex relationships. As a tool, the basic approach was considered good
enough to use thereafter on a maintenance department relayout problem. The interesting
point in this subsequent application was that all flow of materials per se was ignored
and again the result was a relayout that seemed better to experts than their own pro-
posed relayout.

An additional aspect of the modified materials handling criterion warrants


attention. The only communication meditim examined was face-to-face contacts.
What about other forms of communication, e.g., telephone, closed circuit tele-
vision, written reports? The use of face-to-face commimication weights the lay-
out results in favor of the status quo since for a given information need, the use
of face-to-face communication grows proportionately larger as the distance from
the source grows smaller.
The favoring of the existing conditions can be justified in a pragmatic sense.
By so doing, the costs of relocation tend to be minimized and the existing infor-
mal organization tends to be perpetuated (which could, however, have negative
value).
The personal communication issue ideally must also take account of the social
needs of the individuals and the potential suboptimization of enterprise goals
by the thwarting of individual needs. In this context it should be pointed out
thiat the CRAFT model potentially could fulfill a useful function. In almost any
company, the issue of status dictates that when two individuals' work place lo-
cations are exchanged, at least one of the two feels he somehow has been de-
THE FACILITIES LATOTTT PROBLBM IN PBESPBCTITE B-461

moted. To minimize this effect, many companies have elaborate specifications of


office square footle, furniture, etc., for various job specifications. Nevertheless,
the basic fear of change itself is detrimental. This entire problem conceivably is
Eutneliorated by the objectivity of the CRAFT program. If an individual is told
he is to be moved and why (because on the basis of objective data at the new
location he is going to be able to do his job better), the problem of status loss
may be minimized. It also may be possible to have some d^ree of group par-
ticipation incorporated into the data collection process and thereby further lessen
individual fears.'
An Operational Guide to Analysis of Layout Problems
Purpose
Our research has been empirically oriented. Within this framework, it now
seems appropriate to amalgamate our findings into a means of analysis for lay-
out practitioners, The critical factor of layout problem uniqueness has been
discussed above; a guide has been devised in an attempt to clearly delineate the
unique aspects of a particular problem. The layout problem has always been
complex; the guide attempts to highlight this complexity in a way that will
better enable the practitioner to identify the strategic element or elements in
his particular problem and attack the problem accordingly. One or more of the
steps in the guide have been known or used before and each could be the sub-
ject of considerable discussion; what we believe to be unique is the approach in
toLo and the attempt to view the interaction of the issues rather than separately.
An expanded viewpoint of the problem does not necessarily dictate a more
painstaking analysis. We believe that the use of the operational guide will con-
siderably shorten many, if not all, layout analyses. The guide permits the analyst
to initially select the appropriate tool (which indeed may be no tool) for the
analysis of a particular problem. The approach is dictated by the problem, not
by the availability of a tool which may or may not be appropriate.
It is also beheved that the operational guide to analysis of layout problems
will be valuable to those managers who must pass judgment on layout analsnses.
The guide provides a systematic method for determining whether or not a given
problem has been properly investigated and the proper criterion or criteria used.
However, this type of ex post facto evaluation is obviously wasteful. We believe
the primary benefit can be obtained by using the gxiide before any analysis is
performed.
Step 1: Detennine the CompatibilUy of Materials Handling Layout Modds With
the Problem Under Sttidy
Examine the specific problem at hand and attempt to determine what factor
or factors can be conceivably modeled as materialsflow.Traditional plant layout
approaches have been overly constrained in their definition of the term "ma-

• Two additional considerations, the role of qualitative factors and the incremental im-
pact of unique problem constraints, are also examined in the monograph.
B-462 THOMAS E. VOfcliMANN AND ELWOOD 8. BtTFFA

terials." Existing models provide useful results by utilising any factor as a cri-
terion which vades as a fxmction of the distance between system components.
If in the problem under study any factor or factors can be identifled as being
influenced by the relative location of system components, the analysis can pro-
ceed to Step 2 (Determine the Basic Subunits for Analysis). If, on the other
hand, no factor exists in the problem that can conceivably be treated as "ma-
terial," the investigation should branch to Step 6 (Determine the Long-Run
Implications of the Problem).

Step 2: Determine the Basic Subunits for Analysis


Identify what is to be the basic imit or department in this particular problem.
The relative location of facilities problem typically assumes that each subsystem
is specified and that the grouping process itself is not a part of the layout prob-
lem. However, departmental organization in fact may be determined by layout
investigation. For any particular problem, answers to the following questions
may provide considerable insight:
Is the present grouping system (assuming one exists) to dictate future group-
ing? What changes in work methodology are expected to occxir in this problem
situation? For a given problem the salient question is whether or not a flxed
departmentation basis is operative. If the answer is yes, the analysis can proceed
to Step 3 (Determine the Compatibility of the Problem Cost Nature with the
Model Cost Nature). If the departmentation basis is hazy, two things may be
done. First, the existing departments may be subdivided into smaller units.
Second, the departmentation bases themselves can be evaluated by treating
them as variables to be investigated by simulation.

Step 3: Determine th.e Compatibility of the Problem Cost Nature with the Model
Cost Nature
Determine whether the "materials handling cost" is linear, incremental, and
assignable for the particular problem. Materials handling cost models necessarily
assume these cost characteristics. Is the "material handling cost" selected in
Step 1 of an incremental nature?'" If so, can the cost be readily assigned to par-
ticular activities or must the assignment process necessarily become arbitrary as
is true for overhead allocation? Is the cost selected in Step 1 linear? If not, is
linear approximation satisfactory? Is the cost significant, i.e., how does the cost
under study compare to other relative system costs?
" It is absolutely essential that the analyst be clear on the definition of the often mis-
used and complex term "incremental." The underlying reason for its complexity lies in
the nature of the word itself. An incremental cost can only be considered in regard to an
incremental cost analysis which is valid only at one point in time. In any incremental cost
analysis, some assumptions are made as to the parameters which are to remain constant
or be truncated. Incremental costs cut across the lines of fixed or variable costs and can
include or exclude either. The coats that are incremental to a particular facilities layout
decision are then only those costs that will not be the same for the particular alternatives
considered. This point becomes increasingly germane as tbe system constraints or number
of Altematires considered becomes narrowed.
THE FACILITIES LAYOUT PROBLEM IN PERSPECTIVE B-463

The pivotal question for any particular problem situation ia whether the
nature of the problem cost is consistent with the above concepts. If the answer
is a clear-cut yes, the interrogation process can move to Step 4 (Examine the
Impact of Flow Data Assumptions). If the answer to the cost nature question is
a ciear-cut no, the process can branch to Step 6 (Determine the Long-Run Im-
plications of the Problem). If the answer ia not clear-cut, cost nature may be
considered for possible simulation.
Step 4: Examine the Impact of Flmv Data Assumptions
Examine the specific problem and attempt to ascertain how sensitive the final
results are to flow data changes. In this pursuit, answers to the following ques-
tions may provide considerable insight.
To what extent will future flow data differ from the data used for model input?
Are existing data representative? How sensitive is a layout generated on the
basis of existing data to random changes in the data? What change in flow data
would occur if production of some product were to change by some factor? What
products would cause the most difference? Is there some layout which will pro-
vide superior results for several states of nature?
Determine the coeflBcient of variation for the flow data. At this point La the
analysis the decision has been made that "materials handling cost" is an appro-
priate measure of effectiveness for the particular problem under study. This
does not mean, however, that a computerized model is necessarily required to
determine the "best" layout. Before the decision is made to use a sophisticated
model, the flow data should be subjected to an investigation for dominance.
The use of layout by inspection of the flow matrix or some other shortcut
technique is largely determined by whether or not the criterion is relatively
straightforward with no real need for the simulation of assumption changes. If
the criterion is straightforward and theflowdata has a relatively large coefficient
of variation, it would seem appropriate to consider the particular problem as
capable of solution by some shortcut technique. The analysis could then branch
to Step 6 (Determine the Long-Run Implications of the Problem). If high dom-
inance is not the case, and/or simulation seems called for, the problem should
be tentatively considered as amenable to analysis by CRAFT. The flow data
should also be assessed in terms of how they are grouped relative to future ma-
terial handling, inventory, and lot splitting conventions.
Step 5: Recognize the Model Idiosyncrasies, and Look for Possible Improvements
Most heuristic models bave certain idiosyncrasies that affect the final solution
values obtained. Two such idiosyncrasies in the CRAFT model can be high-
lighted. First, the CRAFT heuristic always selects the largest potential cost
reduction as the basis for exchange. Initial iterations usually cause relatively
large reductions in the objective function, thereby allowing certain iterations to
he ignored where the change in the objective function is trivial. Second, the
CRAFT model necessarily approximates departmental areas by their centroids.
It is possible to achieve suboptimal improvements in the objective function by
the use of odd-shaped areas, e.g., all departments laid out as long thin strips.
B-464 THOMAS B. VOIXMANK AND BLWOOD S. BXJFFA

Search for potential improvement in the model. No model is perfect and im-
provements are always possible. If the ultimate test for model applicability is
to be empirical significance, it well may be empiricists who lead the way.
Step 6: Determine the Lang-Run Implications of the Problem
Determine if the problem under study is in fact the right problem for analysis
at this time.
Several steps in the guide have provided for branching to Step 6. The primary
reason for this branching is because the problem under study is not compatible
with model assumptions. The problem may be compatible, however, if it is con-
sidered in a larger context, dove-tailing with other layout decisions. Many lay-
out investigations only deal with minor case problems when in fact at some
point a major case layout may be ji:istified. Perhaps a more desirable way to
approach the issue is to determine the appropriate nsiajor case layout and its
evaluative criterion and examine the impact of handling the particular minor
case problem as a major case problem. If the general direction of the major case
layout is identified perhaps minor case problems could attempt to approach the
major ideal rather than being investigated under minor case criteria and con-
straints. If a major case (or for that matter a more inclusive minor case) approach
seems justified, the problem should be restated and the analysis should branch
back to Step 1 (Determine the Compatibility of Materials Handling Models
with the Problem Under Study). If, on the other hand, the present delineation
of the problem seems appropriate, the analysis can move to Step 7 (Examine
the Layout Problem as a Systems Problem).
Step 7: Examine the Layout Problem as a Systems Problem
Determine the relative importance of this particular layout problem with
respect to other system problemR Answers to the following questions will aid in
this determination:
Is excess capacity purposely built into the materials handling system? What
is the present and probable future plant utilization? Is excess space available?
Is more space needed? Is the layout question critical? What is the minimum cost
method of alleviating the problem? Will refinements be worth their incremental
cost or should a "rough and ready" approach be adopted?
If a significant change in plant activity is anticipated, perhaps the analysis
should branch back to Step 6 (Determine the Long-Run Implications of the
Problem). If a more zealous devotion to the layout problem would provide for
significant dack in other system problem areas, perhaps the problem should be
restated with the analysis branching back to Step 1. If a branch to Step 1 or
Step 6 cannot be justified, the analysis should proceed to Step 8 (Weigh the
Qualitative Factors).
Step 8: Weigh ihe QualiUriive Factors
Determine the role of qualitative factors in this particular problem. This
analytical guide hopefully should have determined the extent to which the prob-
lem may be modeled by a materials handling model.
THE FACILITIES LAYOUT PROBLEM IN PERSPECTIVE B-465

If a materials handling model fits the problem well, qualitative factors per-
haps may not be worthy of pursuit. As qualitative factors increase in significance,
their effect on the objective function may be determined by simulation in the
CRAFT model. As materials handling model approaches become less important,
qualitative factors perhaps may become all-important and be analyzed by a
technique such as Muther's Systematic Layout Planning," or indeed, perhaps
by no formalized approach.
Step 9: Select the Appropriate Tools for Analysis
Only at this point should analytical procedures be started. Let the analytical
procedures as well as detail be dictated by the unique considerations of the prob-
lem at hand.
Conclusions
The results of this research indicate that layout problems typically are some-
what atypical. A proclamation of the general applicability of materials flow (or
any other factor) is a disservice to practitioners. Each layout problem is made
unique by its particular assumptions, constraints, and limitations as well as by
the intrinsic activity of the components. Forcing problems into models is not a
technique which will pass the ultimate test of empirical significance. Each spe-
cific problem must be tested for compatibility with a model before it can be
modeled; unique problems well may be ill-treated by a generalized methodology.
Analytical solutions must be synthesized into a broader operations context.
Such a synthesis must take account of other unique system conditions and prob-
lems.
The treatment of each problem as unique allows a sharper focus and greater
subsequent alleviation of a specific problem. A recognition of necessary model
assumptions and their impact on the problem under study cannot help but aid
in selecting the right method of analysis as well as in determining the type of
problems to which specific tools are most applicable. An attempt to isolate the
key aspect of a particular problem unhindered by folklore approaches, e.g., "how
we have always done it," frequently aids immeasurably in the selection of the
right analytical tool.
A typical result of considering a layout problem as unique may be the dis-
covery that the criteria themselves are vague. That is, management knows it
wants a layout hut cannot specify a criterion of evaluation, or may have several
criteria which are somewhat conflicting. To the extent that problems with con-
flicting criteria can be examined by materials handling cost models, subjective
factors and changes in criteria and assumptions on the objective function can
be measured by simulation.
If, on the other hand, the problem seems to be entirely dictated by qualitative
factors, it should be stated and treated as such with a recognition of the limita-
tions in methodology. Uimecessary problem sophistication sho\ild be avoided.
If no clear measure of effectiveness for a particular problem is evident, a good
" The "Factor Analysis" portion of this work contains an improper coiiyersion of an
ordinal to an intenrsl scale which is not difficult to rectify.
AN OPERATIONAL GUIDE TO ANALYSIS
OF LAYOUT PROBLEMS-FLOW CHART
STEP 1 Determine the compatibility
of models with the problem

STEP 2 Determine the basic


subunits for analysis

Is the departmen.- Yes


tation basis fixed?

Can the units


be further
Yes ubdivided
No
Consider departmentation
as a potential simulation
variable

STEP 3 Determine cost


nature compatibility

Are the
Yes model and problem
cost natures
onsistent?

Can the cost na-


ture issue be sim-
ulated?
Yes
Consider cost nature as
a potential simulation
variable

STEP 4
1.
Examine the impact of
deterministic flow data

Assess the flow


data grouping

B-466
No

Yes
Consider the problem
amenable to heuristic
model solution

STEP 6 Recognize the model


idiosyncrasies and
look for improvement

STEP 6 Determine the long-run


implications of the problem

s the
problem the
right problem
or analysis

STEP 7 Examine the layout problem


as a systems problem

Is a
wgnificant change
in plant activity
anticipated?

ilia
superior solution
to the layout problem
provide for slack in
other problem
reas?

Restate the problem


Go to step 1

STEP 8 Weigh the


qualitative factors

STEP 9 Select the appropriate


tools for analysis
B-467
B-468 THOMAS E. VOLLMANN AND ELWOOD 8.

opportimity for the workers to participate in the decision may be provided. The
qualitative benefits thereby obtained may prove greater than any benefits ob-
tained by rigorous solutions.
In summary, the layout problem should be considered iri the light of problem
uniqueness, the concomitant uniqueness of specific problem criteria, and the
need to reflect this uniqueness in problem approaches. The facilities layout
problem is inherently multi-valued and is not properly handled by a single cri-
terion model. Problems cannot be forced into models; models must be adapted
to problems.
References
1. ABMOUB, G . C . ANP BurFA, E. S., "A Heuristic Algorithm and Simulation Approach to
the Relative Location of Facilities," Management Science, Vol. 9, No. 2, January, 1963.
2. BcPFA, ELWOOB 8., "Sequence Analysis for Functional Layouts," Joumal of Industrial
Engineering, Vol. VI, No. 2, March-April, 1955, pp. 12-13, 26.
3. BuFFA, ELWOOD S., AsMOtTB, GORDON C . AND VOLIMANN, THOMAS E . , "Allocating Fa-
cilities with CRAFT," Harvard Butiness Review, March-April, 1964, pp. 136-158.
4. HiLUER, FBEDKKicaK S., "Quantitative Tools for Plant Layout Analysis," Joumal of
Industrial Engineering, Vol. XIV, No. 1, January-February, 1963, pp. 33-40.
5. MTTTHEB, RICHARD, Systematic Layout Planning, Industrial Education Institute, Boston,
1961.
6. VoLLMANN, THOMAS B . , "An Investigation of the Bases for the Relative Location of
FacUities." Unpubliahed doctoral dissertation. University of California, Los Angeles,
1964.

You might also like