INTERFACES

Vol. 4, No. 3
May 1974

ORGANIZATION DESIGN:
AN INFORMATION PROCESSING VIEW

JAY R. GALBRAITH
European Institute for Advanced Studies

JAY R. GALBRAITH is currently serving as Professor ot Organiza-

tional Behavior at the European Institute for .Advanced Studies in Man-
agement. Foniierly. he was affiliated with the Sloan .School of Manage-
ment, M.I.T. Professor Galbrailh i.s interested in the design of organiza-
tion structures and his most recent work is Designing Complex Organiza-
lions (Addisoii-VVesIey, 1973).

The Information Processing Mode/

A basic proposition is that the greater the uncertainty of the task, the
greater the amount of information that has to be processed between decision
makers during the execution of the task. If the task is well understood prior to
performing it, much of the activity can be preplanned. If it is not under-
stood, then during the actual task execution more knowledge is acquired
which leads to changes in resource allocations, schedules, and priorities. All
these changes require information processing during task performance. There-
fore the greater the task uncertainty, the greater the amount of information
that must be processed among decision makers during task execution in
order to achieve a given level of performance. The basic effect of uncertainty
is to limit the ability of the organization to preplan or to make decisions about
activities in advance of their execution. Therefore it is hypothesized that the
observed variations in organizational forms are variations in the strategies
of organizations to 1) increase their ability to preplan, 2) increase their
flexibility to adapt to their inability to preplan, or, 3) to decrease the
level of performance required for continued viability. Which strategy is
chosen depends on the relative costs of the strategies. The function of the
framework is to identify these strategies and their costs.

The Mechanistic Modei

This framework is best developed by keeping in mind a hypothetical or-
ganization. Assume it is large and employs a number of specialist groups and
resources in providing the output. After the task has been divided, into
specialist substasks, the problem is to integrate the subtasks around the com-
pletion of the global task. This is the problem of organization design. The be-
haviors that occur in one subtask cannot be judged as good or bad per se. The
behaviors are more effective or ineffective depending upon the behaviors of
the other subtask performers. There is a design problem because the executors
of the behaviors cannot communicate with all the roles with whom they are
interdependent. Therefore the design problem is to create mechanisms that
permit coordinated action across large numbers of interdependent roles. Each
of these mechanisms, however, has a limited range over which it is effective
at handling the information requirements necessary to coordinate the inter-
dependent roles. As the amount of uncertainty increases, and therefore in-
Copyright ® 1974, The Institute of Management Sdencca

28
formation processing increases, the organization must adopt integrating
mechanisms which increase its information processing capabilities.
1. Coordination by Rules or Programs
For routine predictable tasks March and Simon have identified the use of
rules or programs to coordinate behavior between inierdependent subtasks
[March and Simon. 1958, Chap. 6]. To the extent that job related situations
can be predicted in advance, and behaviors specified for these situations, pro-
grams allow an interdependent set of activities to be performed without ihe
need for inter unit communication. Each role occupant simply executes the
behavior which is appropriate for the task related situation with which he
is faced.
2. Hierarchy
As the organization faces greater uncertainty its participants face situa-
tions for which they have no rules. At this point the hierarchy is employed
on an exception basis. The recurring job situations are programmed with
rules while infrequent situations are referred to that level in the hierarchy
where a global perspective exists for all affected subunits. However, the hier-
archy also has a limited range. As uncertainty increases the number of ex-
ceptions increases until the hierarchy becomes overloaded.
3. Coordination by Targets or Goals
As the uncertainty of the organization's task increases, coordination in-
creasingly takes place by specifying outputs, goals or targets [March and
Simon, 1958, p. 145]. Instead of specifying specific behaviors to be enacted, the
organization undertakes processes to set goals to be achieved and the em-
ployees select the behaviors which lead to goal accomplishment. Planning
reduces the amount of information processing in the hierarchy by increasing
the amount of discretion exercised at lower levels. Like the use of rules, plan-
ning acliieves integrated action and also eliminates the need for continuous
communication among interdependent subunits as long as task performance
stays within the planned task specifications, budget limits and within targeted
completion dates. If it does not, the hierarchy is again employed on an ex-
ception basis.
The ability of an organization to coordinate interdependent tasks depends
on its ability to compute meaningful subgoals to guide subunit action. When
uncertainty increases because of introducing new products, entering new
markets, or employing new technologies these subgoals are incorrect. The
result is more exceptions, more information processing, and an overloaded
hierarchy.

Design Strategies
The ability of an organization to successfully utilize coordination by
goal setting, hierarchy, and rules depends on the combination of the frequency
of exceptions and the capacity of the hierarchy to handle them. As the task
uncertainty increases the organization must again take organization design
action. It can proceed in either of two general ways. First, it can act in two
ways to reduce the amount of infomiation that is processed. And second, the
organization can act in two ways to increase its capacity to handle more in-
formation. The two methods for reducing the need for information and the
two methods for increasing processing capacity are shown schematically in
Figure 1. The effect of all these actions is to reduce the number of exceptional

INTERFACES May 197-1 29

cases referred upward into the organization through hierarchical channels.
The assumption is that the critical limiting factor of an organizational form is
its ability to handle the non-routine, consequential events that cannot be
anticipated and planned for in advanced. The non-programmed events place
the greatest communication load on the organization.
1. Rules and Programs
2. Hierarchical Referral
3. Goal Setting

4, Creation of 5. Creation of (l. Invesimeni in 7. ('realii)n o f

Slai-k Resources Sclt-Ci>nlai Veriical Lateral
Tasks Information Relations

Reduce ihe Need for Increase the Capacity

to Process Information

I. Organization Design Strategies

1. Creation of Slack Resources

As the number of exceptions begin to overload the hierarchy, one re-
sponse is to increase the planning targets so that fewer exceptions occur. For
example, completion dates can be extended until the number of exceptions
that occur are within the existing information processing capacity of the
organization. This has been the practice in solving job shop scheduling prob-
lems [Pounds, 1963]. Job shops quote delivery times that are long enough to
keep the sclieduling problem within the computational and information pro-
cessing limits of the organization. Since every job shop has the same problem
standard lead times evolve in the industry. Similarly budget targets could be
raised, buffer inventories employed, etc. I'he greater ihe uncertainty, the
greater the magnitude of the inventory, lead time or budget needed to reduce
an overload.
All of these examples have a similar effect. They represent the use of
slack resources to reduce the amount of interdependence between subunits
[March and Simon, 1958, Cyert and March, 1963]. This keeps the required
amount of information within the capacity of the organization to process it.
Information processing is reduced because an exception is less likely to occur
and reduced interdependence means Lhai fewer factors need lo be considered
simultaneously when an exception does occur.
The strategy of using slack resources has its costs. Relaxing budget targets
has the obvious cost of requiring more budget. Increasing ihe time to comple-
tion date has the effect of delaying the customer. Inventories require the
investment of capital funds which could be used elsewhere. Reduction of de-
sign optimization reduces the performance of the article being designed.
Whether slack resources are used to reduce information or not depends on the
relative cost of the other alternatives.
The design choices are: 1) among which factors to change (lead time,
overtime, machine utilization, etc.) to create the slack, and 2) by what amount
30 INTERFACES May 1974
should the factor be changed. Many operations research models are useful in
choosing factors and amounts. The time-cost trade off problem in project
networks is a good example.
2. Creation of Self-CofUained Tasks
The second method of reducing the amount of information processed
is to change the subiask grouping.s from resource (input) based to output based
categories and gi\e each group the resources it needs to supply the output.
For example, the functional organization could be changed to product groups.
Each group would have its own product engineers, process engineers, fabri-
cating and assembly operations, and marketing activities. In other situations,
groups can be created around product lines, geographical areas, projects,
client groups, markets, etc., each of which would contain the input resources
necessary for creation of the output.
The strategy of self containment shifts the basis of the authority structure
from one ba.sed on input, resource, skill, or occupational categories to one
based on output or geographical categories. The shift reduces the amount of
information processing through several mechanisms. First, it reduces the
amount of output diversity faced by a single collection of resources. For ex-
ample, a professional organization with multiple skill specialties providing
service to three different client groups must schedule the use of these special-
ties across three demands for their services and determine priorities when con-
flicts occur. But, if the organization changed to three groups, one for each
client category, each with its own full compliment o£ specialties, the schedule
conflicts across client groups disappears and there is no need to process in-
formation to determine priorities.
Tbe second source of information reduction occurs through a reduced
division of labor. The functional or resource specialized structure pools the
demand for skills across all output categories. In the example above each
client generates approximately one-third of tbe demand for each skill. Since
tbe division of labor is limited by the extent of the market, the division of
labor must decrease as the demand decreases. In the professional organization,
each client group may have generated a need for one-third of a computer
programmer. The functional organization would have hired one programmer
and shared him across the groups. In the self-contained structure there is in-
sufficient demand in each group for a programmer so the professionals must
do their own programming. Specialization is reduced but there is no problem
of scheduling tbe programmer's time across the three possible uses for it.
The cost of the self-containment strategy is tbe loss of resource specializa-
tion. In the example, the organization foregoes the benefit of a specialist in
computer programming. If there is physical equipment, there is a loss of
economies of scale. The professional organization would require three
machines in the self-contained form but only a large time-shared machine in
the functional form. But those resources which have large economies of scale
or for which specialization is necessary may remain centralized. Thus, it is
the degree of self containment that is the variable. The greater the degree of
uncertainty, other things equal, the greater the degree of self-containment.
The design choices are the basis for the self-contained structure and the
number of resources to be contained in the groups. No groups are completely
self-contained or they would not be part of the same organization. But one
product divisionalized firm may have eight of fifteen functions in the division
while another may have twelve of fifteen in the divisions. Usually accounting,
INTERFACES May 1974 31
finance, and legal services are centralized and shared. Those functions which
have economies of scale, require specialization or are necessary for control
remain centralized and not part of the self-contained group.
The first two strategies reduced the amount of information by lower per-
formance standards and creating small autonomous groups to provide the
output. Information is reduced because an exception is less likely to occur and
fewer factors need to be considered when an exception does occur. The next
two strategies accept the performance standards and division of labor as given
and adapt the organization so as to process the new information which is
created during task performance.
3. Investment in Vertical Information Systems
The organization can invest in mechanisms which allow it to process in-
formation acquired during task performance without overloading the hierarchi-
cal communication channels. The investment occurs according to the following
logic. After the organization has created its plan or set of targets for in-
ventories, labor utilization, budgets, and schedules, unanticipated events
occur which generate exceptions requiring adjustments to the original plan.
At some point when the number of exceptions becomes substantial, it is
preferable to generate a new plan rather than make incremental changes
with each exception. The issue is then how frequently should plans be
revised — yearly, quarterly, or monthly? The greater the frequency of re-
planning the greater the resources, such as clerks, computer time, input-output
devices, etc., required to process information about relevant factors.
The cost of information processing resources can be minimized if the
language is formalized. Formalization of a decision-making language simply
means that more information is transmitted with the same number of symbols.
It is assumed that information processing resources are consumed in pro-
portion to the number of symbols transmitted. The accounting system is an
example of a formalized language.
Providing more information, more often, may simply overload the decision
maker. Investment may be required to increase the capacity of the decision
maker by employing computers, various man-machine combinations, assistants-
to, etc. The cost of this strategy is the cost of the information processing
resources consumed in transmitting and processing the data.
The design variables of this strategy are the decision frequency, the degree
of formalization of language, and the type of decision mechanism which will
make the choice. This strategy is usually operationalized by creating redundant
information channels which transmit data from the point of origination up-
ward in the hierarchy where the point of decision rests. If data is formalized
and quantifiable, ihis strategy is effective. If the relevant data are qualitative
and ambiguous, then it may prove easier to bring the decisions down to where
the information exists.
4. Creation of Lateral Relationships
The last strategy is to employ selectively joint decision processes which
cut across lines of authority. This strategy moves the level of decision making
down in the organization to where the information exists but does so without
reorganizing around self-contained groups. There are several types of lateral
decision processes. Some processes are usually referred to as the informal
organization. However, these informal processes do not always arise spontane-
ously out of the needs of the task. This is particularly true in multi-national

SZ INTERFACES May 1974

organizations in which participants are separated by physical barriers, langu-
age differences, and culttiral differences. Under these circumstances lateral
processes need to be designed. The lateral processes evolve as follows with
increases in uncertainty.
4.1. Direct Contact between managers who share a problem. If a prob-
lem arises on the shop floor, the foreman can simply call the design engineer,
and they can jointly agree upon a solution. From an information processing
view, the joint decision prevents an upward referral and unloads the hierarchy.
4.2. Liaison Roles — when the volume of contacts between any two de-
partments grows, it becomes economical to set up a specialized role to handle
this communication. Liaison men are typical examples of specialized roles de-
signed to facilitate communication between two interdependent departments
and to bypass the long lines of communication involved in upward referral.
Liaison roles arise at lower and middle levels of management.
4.3. Task Forces. Direct contact and liaison roles, like the integration
mechanisms before them, have a limited range of usefulness. They work when
two managers or functions are involved. When problems arise involving seven
or eight departments, the decision making capacity of direct contacts ts ex-
ceeded. Then these problems must be referred upward. For uncertain, inter-
dependent tasks such situatioJis arise frequently. Task forces are a form of
horizontal contact which is designed for problems of multiple departments.
The task force is made up of representatives from each of the affected
departments. Some are full-time members, others may be part-time. The task
force is a temporary group. It exists only as long as the problem remains.
When a solution is reached, each participant returns to his normal tasks.
To the extent that they are successful, task forces remove problems from
higher levels of the hierarchy. The decisions are made at lower levels in the
organization. In order to guarantee integration, a group problem solving
approach is taken. Each affected subunit contributes a member and therefore
provides the information necessary to judge the impact on all units.
4.4. Teams. The next extension is to incorporate the group decision
process into the permanent decision processes. That is, as certain decisions
consistently arise, the task forces become permanent. These groups are labeled
teams. There are many design issues concerned in team decision making such
as at what level do they operate, who participates, etc. [Galbraith, 1973,
Chapters 6 and 7]. One design decision is particularly critical. This is the
choice of leadership. Sometimes a problem exists largely in one department
so that the department manager is the leader. Sometimes the leadership passes
from one manager to another. As a new product moves to the market place,
the leader of the new product team is first the technical manager followed
by the production and then the marketing manager. The result is that if
the team cannot reach a consensus decision and the leader decides, the goals
of the leader are consistent with the goals of the organization for the decision
in question. But quite often obvious leaders cannot be found. Another
mechanism must be introduced.
4.5. Integrating Roles. The leadership issue is solved by creating a new
role - an integrating role [Lawrence and Lorsch, 1967, Chapter 3]. These
roles carry the labels of product managers, program managers, project man-
agers, unit managers (hospitals), materials managers, etc. After the role
is created, the design problem is to create enough power in the role
to influence the decision process. These roles have power even when no one

INTERFACES May 1974 33

reports directly to them. They have some power because they report to the
general manager. But if they are selected so as to be unbiased with respect to
the groups they integrate and to have technical competence, they have expert
power. They collect information and equalize power differences due to pre-
ferential access to knowledge and information. 7 he power equalization
increases trust and the quality of the joint decision process. But power
equalization occurs only if the integrating role is staffed with someone who
can exercise expert power in the form of persuasion and informal influences
rather than exert the power of rank or authority.
4.6. Managerial Linking Roles. As tasks become more uncertain, it is
more difficult to exercise expert power. The role must get more power of the
formal authority type in order to be effective at coordinating the joint de-
cisions which occur at lower levels of the organization. This position power
changes the nature of the role which for lack of a better name is labeled a
managerial linking role. It is not like the integrating role because it possesses
formal position power but is different from line managerial roles in that
participants do not report to the linking manager. The power is added by the
following successive changes:
a) The integrator receives approval power of btidgets formulated in the
departments to be integrated.
b) The planning and btidgeting process starts with the integrator mak-
ing his initiation in budgeting legitimate.
c) Linking manager receives the budget for the area of responsibility
and buys resources from the specialist groups.

Fic.UKi: 2. A Pure Mainx Or^iani/uliim

Gcnttal Manager

Pruducl
bng neering Manutac luring Mu kel 11);
Man age me 111

Pniduci Prudui:!
HIetirital Mechanical abncaliun ANscnibly Priiiiuitii>n
Une #1 Line «

\
•V, "-
\
\

Subproduci Subpruduct SubproduL'l Subprudm.) SubproJuci Subpruduct

#2 #2 f2 =2

= Tci:liniL-al aiilliiirity i)vcr ihe ptiidutt

= Formal suihurily over the product (in product organi/aiion. these rcktiunships may be reversed)

34 INTERFACES May 1974

 These mechanisms permit the manager to exercise influence even though
no one works directly for him. The role is concerned with integration but
exercises power through the formal power of the position. If this power is
insufficient to integrate the subtasks and creation of self-contained groups
is not feasible, there is one last step.
4.7. Matrix Organization. The last step is to create the dual authority
relationship and the tnatrix organization [Galbraith, 1971]. At some point in
tlie organization some roles have two superiors. The design issue is to select
(he locus of these roles. The result is a balance of power between the man-
agerial linking roles and the normal line organization roles. Figure 2 depicts
the pure matrix design.
The work of Lawrence and Lorsch is highly consistent with the a!ssertions
concerning lateral relations [Lawrence and Lorsch, 1967, Lorsch and Lawrence,
1968]. They compared the types of lateral relations undertaken by the most
successful firm in three different industries. Their data are summarized in
Table I. The plastics firm has the greatest rate of new product introduction
(uncertainty) and the greatest utilization of lateral processes. The container
firm was also very successful but utilized only standard practices because its
infonnation processing task is much less formidable. THius, the greater the
uncertainty the lower the level of decision making and the integration is main-
tained by lateral relations.

TABLE 1

Plastics Food Container

% new products in last 35% 20%

ten years

Integrating Devices Rules Rules Rules

Hierarchy Hierarchy Hierarchy
Planning Planning Planning
Direct Direct Direct
Contact Contact Contact
Teams at 3 Task forces
levels
Integrating Integrators
Etept.
% Integrators/Managers 22% 17% | 0%

[Adopted from Lawrence and Lorsch, 1967, pp. 86-138 and Lorsch and Lawrence,
1968].

Table 1 points out the cost of using lateral relations. The plastics firm has
22% of its managers in integration roles. Thus, the greater the use of lateral
relations the greater the managerial intensity. This cost must be balanced
against the cost of slack resources, self-contaitied groups and information
systems.

INTERFACES May 1974 35

Choice of Strategy
Each of the four strategies has been briefly presented. The organization
can follow one or some combination of several if it chooses. It will choose
that strategy which has the least cost in its environmental context. [For an
example, see Galbraith, 1970], However, what may be lost in all of the ex-
planations is that the four strategies are hypothesized to be an exhaustive
set of alternatives. That is, if the organization is faced with greater uncertainty
due to technological change, higher performance standards due to increased
competition, or diversifies its product line to reduce dependence, the amounl
of information processing is increased. The organization must adopt at least
one of the four strategies when faced with greater uncertainty. If it does not
consciously choose one of the four, then the first, reduced performance stand-
ards, will happen automatically. The task information requirements and
the capacity of the organization to process information are always matched.
If the organization does not consciously match them, reduced performance
through budget overruns, schedule overruns will occur in order to bring about
equality. Thus the organization should be planned and designed simultaneous-
ly with the planning of the strategy and resource allocations. But if the strategy
involves introducing new products, entering new markets, etc., then some
provision for increased information must be made. Not to decide is to decide,
and it is to decide upon slack resources as the strategy to remove hierarchical
overload.
There is probably a fifth strategy which is not articulated here. Instead of
changing the organization in response to task uncertainty, the organization can
operate on its environment to reduce uncertainty. The organization through
strategic decisions, long term contracts, coalitions, etc., can control its environ-
ment. But these maneuvers have costs also. They should be compared with
costs of the four design strategies presented above.

Summary
The purpose of this paper has been to explain why task xmcertainty is
related to organizational form. In so doing the cognitive limits theory of
Herbert Simon was the guiding influence. As the consequences of cognitive
limits were traced through the framework various organization design stra-
tegies were articulated. The framework provides a basis for integrating organi-^
zational interventions, such as information systems and group problem solv-
ing, which have been treated separately before.

BIBLIOGRAPHY

Cyerl, RicharH, and March, James, The Behavioral Theory of the Firm, Prentice-Hall,
Englewood Cliffs, N. J., 1963.
Galbraith, Jay, "Environmental and Technological Determincnts of Organization Design:
A Case Study" in Lawrence and Lorsch (ed.) Studies in Organization Design, Richard
D. Irwin Inc.. Homewood, 111., 1970.
Galbraith, Jay, "Designing Matrix Organizations", Business Horizons, (Feb. 1971), pp. 29-40.
Galbraith, Jay, Organization Design, .A.ddison-Wesley Pub. Co., Reading, Mass., 1973.
Lawrence, Paul, and Lorsch, Jay, Organization and Environment, Division of Research,
Harvard Business School, Boston, Mass., 1967.
Lorscb, Jay, and Lawrence, Paul, "Environmental Factors and Organization Integration",
Paper read at the .Annual Meeting of the American Sociological Association, August 27,
1968, Boston, Mass.
March, James, and Simon, Herbert, Organizations, John Wiley & Sons, New York. N. Y., 1958.
Pounds, William, "The Scheduling Environment" in Muth and Thompson (eds.) Industrial
Scheduling, Prentice-Hall Inc., Englewood CUtis, N. J., 1963.
Simon, Herbert, Models of Man, John Wiley & Sons, New York, N. Y., 1957.

36 INTERFACES May 1974