Modeling and evaluation of alternatives in information systems

Information Systems Vol. 16, No. 5, pp. 537-558, Printed in Great Britain. All rights reserved

MODELING

1991 Copyright

0

0306-4379/91 $3.00 + 0.00 1991 Pergamon Press plc

AND EVALUATION OF ALTERNATIVES INFORMATION SYSTEMS

IN

OSCARBARROS Universidad de Chile, Facultat de Ciencias Fisicas y Mbtematicas, Departamento de Ingenieria Industrial, Casilla 2777, Santiago, Chile (Received 7 August 1989; in revised form I7 October 1990; received for publication I3 June 1991)

generalized graphical scheme for organizational modeling is proposed. Such a scheme is based on systems theory and general patterns of organizational processes regulation derived from empirical observation and experience. This type of modeling is shown to be useful in approaching the design of the components external to the computer of an information system, or doing external design. Alternative external designs or structures are related to organizational coordination structures and other concepts proposed in organization design theory. Since alternative designs generate alternative results, a quantitative approach is developed to evaluate such alternatives and select the most cost-effective one. Abstract-A

Key won&: Management information

systems, simulation, organization design, modeling, cost-effective-

ness analysis

1. INTRODUCTION

Many approaches for modeling information systems (IS) have been proposed. A few of the most popular ones are structured analysis for the graphic modeling of data flows [l]; entity relationship approach, semantic modeling [2,3] and information engineering [4] for data modeling; and structured design for software modeling [5]. None of these methods explicitly consider users’ tasks performed outside the computer as objects of analysis and design, in the same way as computer programs and/or data-file structures are. The main consequence of this omission is that alternative IS structures associated to alternative ways of performing users’ tasks-clearly related to decision making and organization structure-are not explicitly brought out as design options. Thus design domain is arbitrarily reduced, obviously resulting in suboptimization. There are a few exceptions among IS design methods to this lack of formal consideration of IS alternative structures. Most notable among these is BIAIT [6] defined by Davis [7] as a normative approach, which provides an empirical characterization of IS structure by establishing general patterns of organizational purpose (people and computer implemented), functions and data. This method posits that IS structure is dependent on the values of seven variables that completely characterize an organization. However, details about this methodology, which is now proprietary, are not known. Our work aims at extending the formal consideration of structure in IS design, by explicitly bringing out the problem of alternative structures and their evaluation. We do this by: (i) Developing a general organizational graphic model, based on systems theory and general patterns of organizational process regulation derived from empirical observation and experience. These processes are related to product-resource lifecycle stages defined in other methodologies [8]. (ii) Using the general organizational model to derive alternative information system structures. Here we define the concept of structure to include not only what will be computerized but also the prescription of the decision making behavior of the users of information by means of policies, rules, procedures and the like. Since there always are alternative behaviors and, consequently, alternative sets of information to support such behaviors we are really talking about design, We chose to call this external design, as opposed to internal or computer subsystem design. This problem of jointly studying alternative behaviors and sets of information or external designs (structures) can be related to and supported by organization design theory [g-11]. In fact we will show that these designs correspond to organizational coordination structures, as defined by Malone [l 11. 537

538

OSCARBARROS

(iii) Developing a quantitative modeling approach to be able to evaluate alternatives derived in (ii) and select the best one. The connection with organization theory makes it possible to base modeling on the measuring of the organizational effectiveness of each alternative. Concepts and tools for modeling alternatives are related to Marschack’s value of information ideas [ 121, modeling of computer processes [ 131, operations research and management science (~R/~S) modeling and the software packages available to implement models in practice [14]. The main thrust of this paper is integration and operationalization of ideas. Thus many diverse concepts coming from fields such as systems theory, organization theory, information value theory, OR/MS modeling and modeling packages are put to work together in an operational methodology for the dete~ination and evaluation of alternative requirements and structures. The methodolo~ is purposely practical, for we have based its development on many real cases from which ideas have been derived and to which the methodology has been successfully applied. 2. AN ORGANIZATIONAL

MODEL

We start by recognizing that an IS is embedded within the management system of an organization. Now, the purpose of the management system, and hence of the IS, is to regulate the behavior of the organization under the perturbations generated by a changing and non-controllable environment. Thus a relevant systemic theory to better understand IS is regulation theory [1.5]. According to regulation theory, we begin by identifying the entities that are regulated in any organization. These are organization resources, i.e. materials, including products in which they are used, money, capita1 goods (assets) and human resources. Regulation is then exercised over the flow of these entities determining the system’s behavior. A fifth entity appears in some cases, that is data. For example in a conference organization, one of the regulated entities is “paper”, which contain data in itself (the technical content). But we need further information about a paper for regulation purposes: received date, referees assigned to it, results of the refereeing process, date when result was informed to the author; if accepted, session it was assigned to, etc. So, sometimes, we may find regulated data and data about data for regulation purposes.? This brings out the dual nature of other resources which can also be regulated and at the same time regulator. This may be the case of human resources when executing tasks are regulated and when issuing orders are regulators. We will treat regulated entities as the primitives of our organization model and assume they are sufficient, in the sense that no other entities are needed to describe a real situation. Now, in the act of regulation, entities (resources) flow from the environment into the organization, are operated on within the organization and exit, not necessarily with the same identity, to the environment. The operations performed over the entities are the means for implementing the regulation. We will call these operations processes. Examples of these processes are shown in Table 1. Processes can be classified into the generalized types given in Table 2, where examples of instances of these generalized processes for each regulated entity (resource) are also shown. The generalized types and instances are based on the experience derived from many real cases known to the author. Although independently developed, our concept of processes is similar to the concept of a process proposed in the Jackson System Development (JSD) methodology 1161.JSD defines a process as a sequence of actions performed over a real life object (our resources). For instance, in Cameron [16], the process “book” is defined as composed of the actions: acquire, classify, loan part and end part. It is obvious that these actions are respectively equivalent to our processes: acquire, store, apply and transfer. The difference is that our processes are generalized and valid for any IS and a JSD process is particular of an application. Furthermore, the use given to the idea of a process TThis subject can be further extended by thinking of databases as resources and defining their management as a regulation problem (information resource management). This problem is a meta problem to us and we will not pursue this idea any further here.

Modeling

and evaluation

of alternatives

Table 1. ExamDIes of oraanizational Regulated entity (resource) Materials Money Capital assets Human resources

539

orocesses

FVoCesSeS

Inspect, transport, transform, machine, assemble, dispose, dispatch, etc. Borrow, collect, invest, apply, pay, distribute, etc. Install, use, dispose, etc. Hire, train, develop, promote, assign, motivate, tire, etc. Table 2. Instances of generalized processes for resources Generalized processes

Regulated entity

Acquire

Store

Improve

Maintain

Transfer

Materials

Transport, inspect

Move, inventory, distribute

APPLY Use, handle, machine transform, assemble

Treat, correct

Fix, clean

Dispatch, dispose

Money

Borrow, collect, inspect

Move, distribute

Use

Invest

Protect

Pay

Capital assets

Transport, test

Move, inventory, distribute, install

Assign, use, operate

Modify, rebuild

Fix, repair, prevent, overhaul

Dispose

Human resource

Recruit, select, hire

Pool

Assign, instruct eive task

Train, develop, oromote

Motivate, reward, check-un

Fire, retire

diverges in JSD and our model. JSD goes on to convert the set of actions that make up a process into a computing process that mimics the real-life object [16]. We will use the processes to derive the regulation (and information) needs an IS should satisfy, as we show next. The idea of processes is also related to what other authors call object systems [7, 171 which can be defined in brief as to what gives rise to the need of an IS. We think that processes give a generalized operational description of such an object system. Based on what we have seen so far, we can state that the regulation problem in an organization is to decide what processes to perform over resources at any given time, based on the knowledge of the state of such processes previous to the decision and, possibly, a prediction about the future state of the environment. Regulation as defined here is related to Malone’s [ 1l] idea of coordination which he considers to be associated with the assignment of tasks to processors (our processes). We now identify the decision and other types of functions necessary to perform regulation in the sense we have defined. These functions make up what is the regulator and controller in a general regulation scheme [ 181.In identifying regulation functions, we associate to each generalized process decision and data manipulation activities that may be necessary for its occurrence. The derived functions are listed in Table 3 and are shown in Fig. 1, together with their relationships to processes by means of information flows. In Fig. 1, the symbology in Table 4 has been used. The flows are given just as examples since, as we will discuss later, they are at the heart of the IS design problem. The other attempt we know of to identify generalized functions is BIAIT’s Information Handling Disciplines (IHD). Despite their name, these IHD are decision and data processing functions that may occur in IS. There are around 50 and were derived from an empirical study made at IBM, details of which are not known. According to BIAIT, the occurrence of these IHD in a particular case depends on the characteristics of the business where the IS exists [6]. Other authors have indepently identified stages in the product-resources lifecycle [8] which are similar to some of the processes and generalized functions we have defined. This is not surprising since our regulation model can also be thought to exist to manage resources (including products) through their lifecycles (from input to the organization to exit). These product-resources lifecycle Table 3. Generalized functions needed for generalized processes Generalized processes for a resource

Generalized functions

Acquire

Establish need and specify resource, select supplier and order, decide on quality

Store

Decide application (of stored resources)

APPLY

Decide application (present and possibly future; may include planning and programming), account for application (determine state of application of resources)

Improve

Decide to improve

Maintain

Decide to maintain

Transfer

Decide on transfer

OSCARBARROS

540

lnf~rmat~~ about ~n~ronrn~nt state, Request from Envimn~m State of reswrce

ksource

formation bout nvimnment

C)rder

I

Late

I

I/

instruction on appticalion

Dataabout ~licat~n

Sp&iiations

1

Instruction to maintain n

/

10 improve

Fig. 1. Generalized processes and functions and flows of information.

stages have been used to establish information requirements in a similar fashion tu what we will propose in the next section. Our set of generalized functions pretends to be minimally sufficient, in the sense that this is the least number of functions that allows rep~sen~tion of a problem of regulation of resources in an organization. Of course, we cannot prove this, but we have performed many tests of this model on actual cases that partially validate our presumption. However, we do not claim general applicability of the model. We think that, because of its emphasis on the regulation of resources, our model is suitable for o~mtion~ sit~tions where the IS problem can be clearly stated in terms of resource management. An example of an application of the model, which should be largely self-explanatory, is given in Figs 2 and 3. In this case of credit management two resources have been identified: clients (human] and money ~~nan~ial~. Each resource has its own regulation set of activities, but they relate through a common function: “store and generate client and collection information.” This is a design decision which has led to the centralization and integration of such information. The client regulation has as its main purposes to select good customers that apply for repetitive credit, to make a decision on each part&&r order for products and associated credit request and to eliminate

Modeling and evaluation of alternatives

541

Table 4. Symbology for regulation diagrams

Symbol

Meaning

Process

w

: transformation performed over a resource

Flow of a resource

c-i

Generalized regulation function which makes decision and /or stores I manipulates information about processes

c

cl

Information flow

Object from the environment

clients that are not creditworthy. Credit originates payments in installments that the money regulation part has an objective to collect. Functions in both regulation diagrams have been derived from the general model in Fig. 1, by establishing how each generalized function in such a figure gives rise to a particular activity needed to manage a resource. For example, for the resource “client”, “decide on quality” instantiates as some activity that evaluates a client in terms of creditworthiness and decides that he can be a member of a “client population” that repetitively orders products and applies for deferred payment (in installments); “decide application” (of the resource) has to do with whether we process a client accepting his order and his application for credit giving instructions for dispatching and invoicing products to him; and “account for application” deals with registering orders processed, credits accepted, invoices issued, installments due and so on, and also, all the supporting information that helps making a decision about a credit request. IS 16/5-F

OSCARBARROS

542

order 4

lnveicinginformation

Client information ‘7

r Issue credii policies

ESTABLISH NEED AND SPECIFY

-

promotion

I

t

L’

ACCOUNT POR

DECIDE APPLICATlON

SELECT SUPPLIER AND ORDER

Credit information

I

Ii’

_- . _ -1 _- .

Decisiin on dient

Crbdit policies

Decide on dient

O&t

T Client evaluation

De&ion on request

Roquost

New

DffSS

li-

I

t Credit information

information

credit

DECIDE ON QUALITY

ON TRANSFER

C&ii

for

Sales

t

store and genwat~ dient and mlbaion informatbn

D&de on request

Decision on client

(ENVIRONMENT)

t

ciients

1 I

II : Detail omission Fig. 2. Client regulation for credit management.

This type of analysis can be normative in the sense that we are designing a system from scratch or descriptive in the sense that we match existing activities and information flows to the ones prescribed by the general model. 3. ALTERNATIVE

STRUCTURE

DETERMINATION:

EXTERNAL

DESIGN

In this section we show how the generalized organizational model can be used to derive alternative IS structures. We include within the IS structure, the specification of how regulation functions will be performed-possibly by establishing policies, rules and procedures-and the information processing and flows needed to support such functions. It is obvious that alternative ways of implementing the regulation functions will create the need for different sets of information processing and flows. So, we may have alternative designs of an IS for a given situation. The problem of establishing such designs is called external design to distinguish it from the problem of a computer subsystem or internat design. The methodology we propose for generating and evaluating alternative structures for IS is summarized in Table 5, where we also show some theories that support and complement our

Modeling

and evaluation

of alternatives

543

Product

Invoices

information

I Collection Issue

Payment

policies

collection

AND

* client

payments

policies ESTABLISH

Store

due slips

Collect

L

SELECT

NEED

SUPPLIER

AND

SPECIFY

Past due slips

~

c and generate

and collection information

ACCOUNT

FOR

APPLICATION

ORDER

Other

t

collection actions

+ Invoices Control

Collected

payments

Payments

due information

payment Payments due slips

DECIDEON

.

Past

_

QUALITY

4

due slips Payment

Accepted

information

rejected

I

payment decision

it

f

/

Client

\

/ Suppliers

Money

Payments

MARKET

SUPPLIER

(ENVIRONMENT)

(ENVIRONMENT) \ b I

/

I Rejected

payments

Fig. 3. Money

rh+.i,nmiPPinn

II

regulation

for credit

management.

analysis. This methodology will be developed and examplified in this section and the following one. Our methodology is supported by a CASE-like product which allows for computer-based diagraming and also provides a dictionary for documentation of flows, decision and data manipulation, which we will not illustrate here. 3.1. IdentifVing resources and modeling current situation In approaching the design of a particular IS we start by identifying which the relevant resources are and modeling the current situation. The generalized model tells us what processes and regulation functions may exist to manage a resource. By matching current activities within the Table 5. Methodology

for generation Modeling

Step 1. Identify resources be managed

to

and evaluation

approach

Look for material/products, capital goods or information

human, money resources

of alternatives Supporting Regulation

theory

theory

theory

[151

2. Model current situation for each resource

For each resource use the generalized model in Fig. I and match activities in practice to those of the model; identify information flows among current activities; document current rules for decision functions

Regulation

3. Generate alternative structures for the IS

For each resource, use the generalized model to identify new or improved functions and information flows that could modify current situation; in particular identify new or improved rules for decision functions and new or improved information manipulation

Regulation theory, coordination theory [l I], organization design theory [lo], contingency theory

Quantitatively and perform-a

Petri nets [13], systems dynamics [19], DSS [14]

4. Evaluate

alternatives

model each alternative cost-effectiveness analysis

[9]

+j+

NEEO

SUPPLIER

Dispakbd material

Materialrequirement b ACCOUNT FOR APPLICATlOt4

SELECT SUPPLIER AND CXIDER

Mdrial

Inventory

W,,+ru,G

Compute Reorder Pointand Lot Size Order

*

Fig. 4. Current situation model for material management problem.

: Detailomission

AND Sl’l!CCWi’

ESTABLISH

Forecastmaterial comsurrglion cornpuleMaterial requirement

Materialinuentorv

I

Modeling and evaluation of alternatives

545

Bank account balance

Request Loan

SELECT SUPPLIER AND ORDER

Instructions for Payments and Short term investment

Collections and Loan received Loan requisition

Collection

i FINANCIAL MARKET AND DEBTORS

.II Money

Fig. 5. Current situation model for cash-flow management problem.

scope of the situation with generalized processes and functions, it is very easy to generate a model for the current situation. For example, assume we have a material management problem and a cash-flow management problem. Resources to be regulated are clearly material and money, respectively. Current situations models are shown in Figs 4 and 5 for both problems. The examples have been kept simple to facilitate understanding. In particular both examples cover just a part of only one resource regulation problem. There are no limitations in the model we discussed in Section 2 to fully cover one resource regulation and even the regulation of several resources interacting in a complex way. Also, these examples emphasize resource requisition, but there is not any inherent limitation to expand the problem to consider resource utilization and services or goods production. Both current situations are characterized by very simple decision function rules and limited information requirements. Thus, in the material management example, replenishment is decided by comparing “material inventory” to a “recorder point” and if the former is less than or equal to the latter, issuing an order for a “lot size”. Notice that the only basis for reorder criteria is “material requirement” based on historical consumption (dispatched material). In the cash-flow management problem, “loan requisition” is decided on the basis of getting negative “bank account balances” to zero and “short-term investment”, by allocating positive “bank account balances”. There are many other details which we do not consider for the time being, e.g. how is the material forecast done? We will come back to this when we quantitatively model these situations in order to be able to evaluate them. In particular we will consider the dynamics of the problem, e.g. the time frame in which decisions are made.

546

OSCARBARREN

In these simple examples just one level of diagramming is enough to fully represent the situation. In more complex situations, several levels of diagramming can be used by hierarc~cally decomposing functions in the style of structured analysis El]. 3.2. Generating alternative structures Now, given models for the current situation, we can generate alternative structures by discovering regulation functions or relationships absent or imperfectly implemented with respect to the generalized IS model in Fig. 1. Consider the material management example. Notice that the material forecast is based on just historical consumption information (dispatched material). Now, the general model in Fig. 1 indicates that there may be an obvious relations~p between the decision on what to apply the resoume to (decide application) and material (resource) required (establish need and specify). Hence an alternative based on this observation, shown in Fig. 6, will explicitly introduce within the system the function “determine manufacturing plan” (decide application) and link it to “compute material requirement” (establish need and specify) by means of the “manufacturing plan”. Then the material required will be computed strictly for the needs of the “manufacturing plan” and ordering done accordingly, instead of basing the decision on just historical material consumption.t In the cash management problem we observe that the only basis for loan and short-term investment decisions is current “bank account balance”. Now the general IS model indicates that the ordering (select supplier and order) of a resource (loan) may be based on its projected need (establish need and specify). This suggests defining an alternative, shown in Fig. 7, where the currently absent function “compute short-term cash flow forecast” (establish need and specify) is created to support the loan and investment decisions. In turn this implies expanding the “account for application” function to provide the necessary information to be able to forecast the cash flow. Searching for alternative st~ctures can be guided by organization design theory [9, lo]. As detailed elsewhere [15,21], different implementations of functions and relationships with corresponding sets of information requirements imply different degrees of coordination and organizational structures. These in turn will produce different consequences in terms of results achieved (measured in terms of quality of regulation in our terminology). But coordination can be approached with two types of mechanisms. On the one hand, we can reduce interaction-hence the need for coordination and flow of information-among units by using slack resources (e.g. buffer inventories and backlogs) and unit elf-containment. On the other hand, we can increase the capacity to coordinate by providing more info~ation and coordinating agents, like assistants, staff and lateral roles, e.g. liaisons [IO]. As is usual in IS, in both of the examples we have presented, alternatives go in the direction of increasing the capacity to coordinate and reducing use of slack resources. Thus, in the material management case, the computing of “material requirement” based on an explicit “manufacturing plan”, as shown in Fig. 6, will improve coordination between manufacturing and requisition of material. This will mean a reduction of the slack resources material inventory and/or delays in product orders satisfaction for, in the current situation, excess inventor has to be maintained to meet unforeseen demand or if not so, lack of material inventory may prevent satisfying products orders. The better adjustment between material needs and requisition produced by the alternative will mostly eliminate these problems, In the cash flow management alternative in Fig. 7, the explicit forecasting of cash flow allows for better planning and negotiation of the loans required (again better coordination between needs and requisition). This in turn affects the slack resource interest payments (income), since better loan (investment) conditions can be obtained if negotiated in advance. Of course these coordination arguments which we give a posteriori can be used a priori in their generation.% Also it is obvious that the identification of slack resources affected by the alternatives is a key factor in the evaluation we will attempt in the next section.

tNotice that current situation and alternative correspond to simplifications of policies very much used in practice: the first to a ROP (reorder point) policy and the second to an MRP (material requirements planning) policy. $For further examples of uses of coordination ideas in generating alternative structures see Barros (15,211.

ComputeMaterial

SUPPLIER

I

Order

ESTABLISH NEED ANDSPECCW

I

I

I-

I

4 +-7

I

I

ManufachwingPlan

Fig. 6. Model for an alternative

1

Parameters/

problem.

ManufaciuringPlan

in the material management

APPLlCATlON

+

Received and Ditched malerial

Inventory

complte

E’

f

4

n.

B

00

OSCARBARR-

548

Alternative IS designs are also related to the idea of coordination structures as defined by Malone he makes specific micro-level assumptions about how decisions are handled and the amount of information (messages) required. Different sets of assumptions lead to different generalized “pure” structure types: product hierarchy, descentralized market, functional hierarchy and centralized market. Thus, for example, in a product hierarchy, decisions about tasks (processes) are decentralized in a product manager and the processors he directs. It is clear that these types can be considered as general patterns of coordination one can use for an organization as a whole and that each type will generate different IS designs. For instance, consider the problem of material management (procurement) for the manufacture of all the organization’s products. The product hierarchy design will define an independent material management IS for each product line. Alternatively we have the possibility of just one material management IS (functional hierarchy) which unifies and satisfies the material requirement for all lines of products. Malone [ll] analyzes the different types of structures in terms of production, coordination and vulnerability costs. Rough approximations to these cost allow us to approximately prescribe the desirability of each structure for a given situation. Hence Malone’s ideas appear to be a good way to screen possible general alternative macro structures for the organizational information system. Then our generalized organizational model and the evaluation procedure we developed in Section 4 will allow us to detail the alternative IS structures and select the most cost-effective one. [l 11. Though Malone refers to the macro structure of an organization,

4. 4.1.

EVALUATING

ALTERNATIVES

Basis for modeling

Modeling the IS essentially means modeling of the processes. There are several ways in which we can model a process, e.g. using Petri nets [13]. We choose the approach that follows due to the fact that it can be supported by widely-available software packages, as we will show later, and that it can be related to most of the modeling work that has been done in OR/MS. A process is characterized at a given point in time by state variables. A given state is defined by a specific assignment of values to the state variables. Changes of or transitions between states are specified through an action function? which, for a given beginning state, establishes an immediate succeeding state for the process. Of course, by repetitive application of the action function, we can generate a simulated behavior, or a sequence of states, starting from a given initial state. The basic model relates directly to the graphical regulation model of Fig. 1. Thus, certain regulation fuctions such as “account for application” exist for computing the state of processes; others, like “decide application” are actually for generating action fucntions, i.e. taking action based on current state to carry a process to a certain new state. We can specify an action function by defining: Yt = state of a process at time t (given values of a vector of state variables or variable set at time t), X, = values of the vector of action or decision variables that affect the state of a process (decision variables are the ones that can be set freely by regulation functions). Then: y,+, = F(Y,> -0,

(1)

where F is a vector function that assigns a successor state to a beginning state for given values of the decision variables.

THere, we are talking about mathematical functions, which are not to be confused with the regulation functions defined in Section 2.

Modeling and evaluation of alternatives

549

NOW the determination of X, can be done in many ways, depending on the characteristics of the problem and degree of realism we wish to include in the model; for example: (.I deterministic

error correction: X=W’,,

Y,-,,

Y,-z, . . . . Y-d,

(2)

i.e. values of decision variables only depend on current or past state, which corresponds to the usual feedback control or regulation by the error scheme [19].

x,=G(Y,,Y

r-,, Y,-2, . . . . Yo;

~;+,,...,&IT),

(3)

where fZ is the projected future state, T is the horizon and G is a vector of a closed-expression mathematical functions. {LEEdeterministic untic~ation-~igorithmic: Here G is not a vector of close-expression mathematical functions, but the result of an algorithm.? e.g. this will be the case when functions in G form a system of simultaneous equations for the determination of X, or when X, is the solution of a mathematical programming optimi~tion problem. In these cases one finds that in projecting the future state one must also establish future decisions. Hence, current decision is based on projected future decisions. This is the case, for example, in production smoothing where, under seasonal demand, today’s production must consider future production and its relationship to future peak demand. This leads to the concept of simulation over a rolling horizon, i.e. at time t one solves equation (3) considering projected decisions up to the horizon T, then after implementing a decision for the next period one advances one time unit, collects new data about processes, update state, again solves (3) for t + 1 and so on. Considering the effect of future decisions on today’s decision also means forecasting future values for parameters which have not been explicited in equations (l-3). For example, in production smoothing this means forecasting product sales for each period of the horizon. Note that this type of specification for G allows us to use from very simple heuristics to complex OR/MS models. (2~1Stochastic: Since forecasting is subject to error, one can have a by representing future values of parameters as stochastic instead of case decision variables must be determined under un~ert~nty occurrence of future states and system’s behavior also established

better representation of reality deterministic variables. In this based on the probability of in a probabilistic way.

Up to now we have implicitly assumed time as discrete. If we wanted no loss of state info~ation, the time increment should be variable and determined by the time interval between events that change process state. However, we will choose to consider time increments as fixed but as small as necessary to avoid loss of state info~ation. The reason for this is that the software available we will use for implementing IS models only accepts discrete time increments. 4.2 Tools for mo~eiing The purpose in modeling an IS is to be able to accurately represent a given alternative-as discussed in Section %-and calculate its organizational effectiveness by simulating its behavior. This is geared to make a lost-effectiveness analysis of the alternative to determine if it is preferable to the current situation or to other possibly available alternatives. Hence, we need tools for implementing models of the type discussed in Section 4.1, that allows as quick and effortless an evaluation as possible. Fortunately, these tools exist, though they have been developed for other purposes. They are the modeling software packages available for building DSS [ 141.These packages use as a calculation paradigm a matrix, where columns usually represent time (in discrete increments) and rows or lines are items-such as states, decision variables or parameters--to be projected in time. They contain powerful expressions and modules that allow specification of how an item is to vary in time and to “solve” the specifications. For example, IFPS-one of the leading TFormally, it can also be the result of a heuristic.

OSCAR BARR&Y

550

packages of this type-has, among other facilities, built-in routines to project parameters under various assumptions, e.g. linear and non-linear regression; provides a simple logic for specifying conditional values for decision variables; allows to specification and solves linear and non-linear systems of equations for Y, and X,; allows to specification and solves linear and integer mathematical programs for establishing Y, and X,; permits the specification of parameters as stochastic variables and determines probability distributions of Y, by means of Monte-Carlo simulation; all this in a very friendly environment where data and model management are very much facilitated. For example, the problems of matrix generation and report production when solving a mathematical program are easily handled with the built-in facilities of the package [22-241. Hence these packages allow us to build a model of it in very little time, usually a few days, and calculate the results it would produce if applied in practice, hence its effectiveness. We have extensively tested this scheme with alternatives for a variety of IS using IFPS. In all events, the approach has worked quite well making possible the quantitative evaluation of alternatives. In the next section, we give examples of application of the approach. 4.3. Modeling examples We model the material management and cash flow management problems presented in Section 3 and shown in Figs 4-l. In the current situation for the material management case in Fig. 4, the material forecast is done by using exponential smoothing on the historical consumption [25]. Lot size and reorder point are calculated according to the usual formulas [25] and appropriate values for parameters in these Invoices, Purchases and Other

expenses per p%rtod, Bark accwnt balance Parameters

1

3ther

expenses

V Decide Payments

Compute short

term Cash flow

Request Loan Cash f& forecast SELECT SUPPLIER AND ORDER

account balance Sorl klvoices RJlchases and Other expenses

and Short term investmant DECIDE APPLICA~

+

ACCOUNT FOR ~~I~A~ON 4

Instruction for Payments and Short term investment

Loan requisition

Collections and Loan received

f / FINANCIAL MARKET AND DEBTORS (

account and short term investment

J

~$ammmm

Fig. ‘7. Model for an alternative in the cash-flow management problem.

Modeling and evaluation of alternatives

551

formulas are given. Simulation of this situation will be performed to try to satisfy the material needs for a given “manufacturing plan” over an horizon of 12 months. However, in this case, no knowledge about this plan is used in making decisions. A model that simulates the behavior under these conditions is given in Fig. 8, together with the results obtained for given data. The model clearly separates the implementation of an action function of type (1) in the first block, the projection of future states in the second and the implementation of a decision function of type (3) in the third block. In the alternative given in Fig. 5, the basic idea is to make explicit the use of the “manufacturing plan” to anticipate material requirements (in the function “compute material requirements”) over the planning horizon. Based on these anticipated requirements the decision about an order will be based on the Silver-Meal 1261heuristic for the so-called dynamic lot size. A model that simulates this alternative together with the results for a 12 month horizon is given in Fig. 9. Of course in COLUMNS 1 ..12 \\ PROCESS \\

STATE DETERMINATION BASED ON PREVIOUS ACTION. \\ MANUFACTURING PLAN=900,1150,1050,950,1050,1200,1300,1500,1050,900,1000,1000 DISPATCHED MATERIAL= MANUFACTURING PIAN MATERIAL UNIT COMSUMPTION RECEIVED MATERfAL = PREVIOUS ORDER MATERIAL INVENTORY = INITIAL INVENTORY - DISPATCHED MATERIAL,’ PREVIOUS + RECEIVED MATERIAL - DISPATCHED MATERIAL \\ PROJECT FUTURE STATE. \\ \\ MATERIAL FORECAST = DISPATCHES MATERIAL, ’ PREVIOUS (1 - ALPHA) + DISPATCHED MATERIAL * ALPHA MATERIAL REQUIREMENTS = MATERIAL FORECAST INVENTORY FORECAST = MATERIAL FORECAST LEADTIME + SS \\ DECISION BY DETERMINISTIC ANTICIPATION SIMPLE. \\ \\ ORDER = IF INVENTORY FORECAST > = MATERIAL INVENTORY THEN 0 ELSE 0 0 = ROUND (XPOWERY (2 A MATERIAL FORECAST/ IC, 0.5)) \\ SUMMARY OF EFFECTIVENESS. \\ \\ AVERAGE INVENTORY = MEAN (MATERIAL INVENTORY [I], MATERIAL INVENTORY) NUMBER OF ORDERS = IF ORDER > 0 THEN PREVIOUS + 1 ELSE PREVIOUS TOTAL BACKORDERS = IF MATERtAL INVENTORY < 0 THEN PREVIOUS + 1 ELSE PREVIOUS TOTAL COST = NUMBER OF ORDERS * A + AVERAGE INVENTORY * IC \\ \\ PARAMETERS. \\ A = 500 INITIAL INVENTORY = 0 ss=o 1 3 LEAD TIME = 1 MANUFACTURING PLAN 900 1150 2 I050 9640 ALPHA = 0.7 DISPATCHED MATERIAL 90 ii5 105 95 MATERIAL UNIT CONSUMPTION = 0.1 RECEIVED MATERIAL 0 212 232 0 l

l

l

l

IC = 2

l

MATERIAL INVENTORY MATERIAL FORECAST MATERIAL ~EOUIREME~TS INVENTORY FORECAST ORDER 0 AVERAGE INVENTORY NUMBER OF ORDERS TOTAL BACKORDERS TOTAL COST

-90 :: 90 212 212 -90

t

32:

7 108 108 108 232 232 .42 2 ,I:

134 106 106 106 0 230 17 2

I

1074

MANUFACTWING PLAN DISPATCHEO MATERIAL RECEIVED MATERNAL MATERIAL INVENTORY MATERIAL FORECAST MATERIAL REQUIREMENTS INVENTORY FORECAST ORDER CJ AVERAGE INVENTORY NUMBERS OF ORDERS

7 1300 130 240 146 I25 125 125 0 250 61 4

6 1500 150 0 -4 143 143 143 267 267 53

9 1050 105 267 I58 ii6 ii6 116 0 241 65

TOTAL COST BACKORDERS

21221

2606.:

2629z

Fig. 8. Model for current situation-material

management problem.

;z 98 292; 222 23 3

I

1545 10 900 90 0 66 98 98 2:: 221 65 6 ,I,‘,

5 1050 105 222 156 103 103 103 0 227 49 3

6 I200 120 0 36 Ii5 115 Ii5 240 240 47 4

1598

209:

t

11

12

1000

1000

100 221 189 99 ::: 0 223 76

3153:

100 0 89 IO0 100 too 223 223 77 7 36525

OSCARBARROS

552

this case the affected slack resource is inventory and it is seen that it is greatly reduced with respect to the current situation, with less backorders. Also overall cost is reduced with respect to such a situation. We consider now the current situation for the cash management problem given in Fig. 6. This is an example of the classical regulation by error correction scheme. Thus a loan is requested to satisfy the last period’s deficit and “short term investment” is made to use the last period’s surplus. Both loans and investments are due next period. A model that simulates this current situation for given series of collections and payments, together with the results for 12 periods, is given in Fig. 10. COLUMNS I..14 \\ PROCESS STATE DETERMINATION BASED ON PREVIOUS ACTION. \\ \\ MANUFACTURING PLAN = 900, 1150,1050,950, 1050, 1200, 1300, 1500, 1050,900,1000,1000 DISPATCHED MATERIAL = MANUFACTURING PLAN l MATERIAL UNIT CONSUMPTION RECEIVED MATERIAL = PREVIOUS ORDER Pl : MATERIAL INVENTORY = INITIAL INVENTORY - DISPATCHED MATERIAL, ’ PREVIOUS + RECEIVED MATERIAL - DISPATCHED MATERIAL \\ PROJECT FUTURE STATE. \\ \\ P2: MATERIAL REQUIREMENT GROSS = MANUFACTURING PLAN l MATERIAL UNIT CONSUMPTION \\ DECISION BY DETERMINISTIC ANTICIPATION-ALGORITHMIC. \\ \\ CTE=2’A/IC T= IF (PREVIOUS Pl: + PREVIOUS ORDER) z = (P2: + FUTURE 1 P2:)’ THEN 0 ELSE IF (MAXIMUM (0, FUTURE 1 P2: -MAXIMUM (0, PREVIOUS Pl : - P2: + PREVIOUS ORDER))) > = CTE THEN XPOWERY (CTE I MAXIMUM (0, FUTURE 1 P2: - ’ MAXIMUM (0, PREVIOUS Pl : - P2: + PREVIOUS ORDER)), 0.5) ’ ELSE IF (4’ MAXIMUM (0, FUTURE 2 P2: -MAXIMUM (0, PREVIOUS Pl: +’ PREVIOUS ORDER - P2: - FUTURE lP2: ))) > - CTE THEN XPOWERY (CTE I MAXIMUM (0, FUTURE 2 P2: - MAXIMUM (0, ’ PREVIOUS Pl : + PREVIOUS ORDER - P2: - FUTURE 1 P2: )), 0.5) ELSE 2 Q = IF T < = 1 THEN MAXIMUM (0, FUTURE 1 P2: - MAXIMUM (0,’ PREVIOUS Pl: - P2: + PREVIOUS ORDER)), ELSE IF T < = 2 THEN MAXIMUM (0, FUTURE 1 P2: - ’ MAXIMUM (0, PREVIOUS Pl: - P2: + PREVIOUS ORDER )) + ’ (T-l) l MAXIMUM (0, FUTURE 2 P2: - MAXIMUM (0, ’ PREVIOUS Pl : + PREVIOUS ORDER - P2: - FUTURE 1 P2: )) ELSE 0 ORDER = Q \\ SUMMARY OF EFFECTIVENESS. :\\ AVERAGE INVENTORY = MEAN (MATERIAL INVENTORY [l], MATERIAL INVENTORY) NUMBER OF ORDERS - IF ORDER > 0 THEN PREVIOUS + 1 ELSE PREVIOUS TOTAL BACKORDERS - IF MATERIAL INVENTORY < 0 THEN PREVIOUS + 1’ ELSE PREVIOUS TOTAL COST = NUMBER OF ORDERS l A + AVERAGE INVENTORY l IC \\ :\\

PARAMETERS.

A = 500 INITIAL INVENTORY = 0 ss-0 LEADTIME = 1 ALPHA = 0.70 MATERIAL UNIT CONSUMPTION IC = 2

= 0.1

w”I*cT”R*NG PLhN OIIIITCHID R&CLIvED LUmRmL LUTm*AL MmRmL INnHlORY WTEmILFqUIRmsNT :m,, m%MCE IHVLWTORI TOTLL01 ORDERS Hwm‘R BICI(ORD~RI TmmLeom

900 1 1130 1 1,s:

nAN”PICT”RINC slul DIIPITCHID MTERIAL RECIIVED PUTERI*I MTIRILL INVCWTORI: MTElIIL RIcl”IREnENT

so 0 220 111 101 1,s -90 9s 90 1:: 103 a.0 2.0 0.0 120 191 -eo -31 !: : 2 1 320 9,: Id ‘I . lJP0 Illlo 1a:a 130 150 101 254 131 121 10s i 130 IS0 101

:m,, ~VEIUCC INVEWTOR’I N”almR 01 oFxmFc?i TOTAL mcmnDm.s TO?&L COST

I.0 111 0.0 190 2.0 38 a”, I, : : 6 2575 2192 xl,:

Fig. 9. Model of an alternative--material management problem.

Modeling

and evaluation

of alternatives

553

COLUMNS I..12

;

PROCESS STATE DETERMINATION BASED ON PREVIOUS ACTION

COLLBCTIONS=l200, PREVIOUS * 1.05 PAY~S~1000,1500.15~,1~,1ooo.1 BANE ACCOUNT BALANCE=INITIAL BAB -PAYMENTS - SHORT TBRM lNW%TMENT+’ COLLECTIONS + LOAN, PREVIOUS - PAYMENTS - SHORT TERM INVESTMBNT + PREVIOUS SHORT TERM INVESTMENT * (l+lRz) + COLLBCIIONS + LOAN _PREVIOUS LOAN l (l+BR)

i

DECISION BY DETERMINISTIC!

ERROR CORRECTION.

LOAN = MAXIMUM (- PREVIOUS BANE ACCOUNT BALANCE, 0)

SHORT TERM INVBSTMENT= MAXIMUM (PREVIOUS BANE ACCOUNT BALANCE, 0)

;

SUMMARY

OF EFFECTIVENESS

INTBREST INCOME = 0, PREVIOUS SHORT TERM INVESTMBNT * INVESTMENT RATE INTEREST EXPENSE = 0. PREVIOUS LOAN * BORROWING RATE NET INTEREST INCOME = INTEREST INCOME - INTEREST EXPENSE CUMULATWB NET INTEREST INCOME = NET INTEREST INCOME. PREVIOUS + NET INTEREST INCOME

\ \

PARAMETERS

1R:INVESTMENT RATE = 1.3% INITIAL BAB 40 BR: BORROWING RATE = 2.4%

1 COLLFZTIONS 1200.0 PAYMENTS 1ooo.0 BANK ACCOUNT BALANCE 290.0 LOAN 0.0 SHORT TERM INVESTMENT 0.0 INTEREST INCOME 0.0 INTEREST EXPENSE 0.0 NET INTEREST INCOME 0.0 CUMULATIVE NET INTEREST 0.0

COLLECTIONS 16uS.l PAYMENTS 1500.0 BANK ACCOUNT BALANCE 692.5 LOAN 0.0 SHORT TERM INVJiSTMENT 371.2 INTEREST INCOME 7.5 INTEREST EXPENSE 0.0 NET INTEREST INCOME 7.5 CUMULATIVE NET INTEREST 3.3

Fig. 10. Model

for current

2 1260.0 1500.0 -240.0 0.0 290.0 0.0 0.0 0.0 0.0

3 1323.0 1500.0 116.8 240.0 0.0 3.8 0.0 3.8 3.8

4 1389.2 1300.0 -156.6 0.0 116.8 0.0 5.8 -5.8 -2.0

5 1458.6 1cco.o 576.9 156.6 0.0 1.5 0.0 1.5 -0.5

8 1688.5

9 1772.9 2m.o 474.5 0.0 64.5 9.0 0.0 9.0 17.1

10 1861.6 1500.0 426.9 0.0 474.5 0.8 0.0 0.8 17.9

11 1954.7 1700.0 735.3 0.0 426.9 6.2 0.0 6.2 24.1

management

problem.

64.5 0.0 692.5 4.8 0.0 4.8 8.1

situation-cash-flow

6 1531.5 1cm.o 371.2 0.0 576.9 0.0 3.8 -3.8 -4.2 12 2052.4 684.9 0.0 735.3 5.6 0.0 5.6 29.7

The alternative for the cash flow management problem, given in Fig. 7, attempts to forecast cash flow based on invoices, purchases and “other expenses” information. The forecast, which is assumed known for the current period, is then used to anticipate the need for a loan and negotiate it in advance, hopefully with a smaller borrowing rate than in the current situation. For simplicity we assume that “short-term investment” is made just to allocate the current period’s surplus with maturity at the next period. Thus we ignore for the time being the obvious possibility of making “short-term investment” for more than one period, with possibly better rates, based on the knowledge of a cash flow forecast. A model that simulates this alternative together with the results for 12 periods is given in Fig. 11. The slack resource affected in this case is clearly “cumulative net interest income” (expense). Notice that this is greatly improved with respect to current situation. As a further example of modeling we consider an extension of the cash flow management problem shown in Fig. 7, leading to a second more refined alternative. This assumes that the cash flow forecast is known for the whole horizon and that investment and borrowing decisions will be made based on such a forecast. Also, investments and loans can be taken optionally for maturity at one, two or three periods in the future. Due to the large number of variables that these assumptions generate, the only way to make decisions in this case is to use an LP-based algorithm. In Fig. 12, we show an LP-based model for the problem just described, using the format of OPTIMUM, the IFPS’s optimizer package [23,27] and the results for the same 1Zperiod data used in the alternative in Fig. 11. Notice that optimized decisions lead to a better “cumulative net interest income”.

OSCARBARROS

554

\ COLUMNS 1..12 PROCESS STATE DETERMINATION BASED ON PREVIOUS ACTION. \ COLLE~ONS=l2~~~VIOUS ’ 1.05 PAYMENTS=1~,1~,1500,1300,1000,1~,1500~~~~,1500,1700,1800 BANK ACCOUNT BAI.ANCE+&NITIAL BAB - PAYMENTS - SHORT TERM mTMENT ’ + COLLECTIONS + LOAN, PREVIOUS - PAYMENTS - SHORT TERM INVESTMENT - PREVIOUS LOAN * (l+BR:) + PREVIOUS SHORT TERM INVESTMENT * (l+IR:) + COLLECTIONS + LOAN \ \ PROJECT FUTURE STATE. INVOICES = 1260,PREVIoUS + 1.05 PURCHASES = 1300,1300,1100,800,800,1300,1800,1800,1300,15~.1~ OTHER EXPENSES=‘LDO COLLECTION FORECAST = INITIAL COLLECTION, PREVIOUS INVOICES PAYMENTS FORECAST = INITIAL PAYMENTS, PREVIOUS PURCHASES + OTHER EXPENSES CASH FLOW FORECAST= COLLECTION FORECAST _ PAYMENTS FORECAST

\

DECISION BY DETERMINISTIC

ANTICIPATION-SIMPLE.

LOAN REQUISITION = LOAN LOAN = ~I~{ 0, - INITIAL BAB - CASH FLOW FORECAST ) , ’ MAX~( 0, PREVIOUS LOAN * (l+BR:) - PREVIOUS BANK ACCOUNT BALANCE - ’ PREVIOUS SHORT TERM INVESTMENT * (I+fR:) - CASH =OW FORECAST ) SHORT TERM INVESMNT = MAXIMUM( 0. INITIAL BAB + CASH FLOW FORECAST),’ MAXIMUM( 0, PREVIOUS BANK ACCOUNT BALANCE + CASH FLOW FORECAST ’ + PREVIOUS SHORT TERM INVESTMENT * (l+lR:) - PREVIOUS LOAN * (l+BR:) ) \ \ SUMMARY OF EFFECTIVENESS. INTEREST INCOME = 0, PREVIOUS SHORT TERM INVESTMENT * INVESTMENT RATE INTEREST EXPENSE= 0, PREVIOUS LOAN * BOGOTA RATE NET INTEREST INCOME= INTEP.EST INCOME - IN’IEREST EXPENSE CUMULATIVE NET INTEREST INCOME = NET INTEREST INCOME, PREVIOUS + NET INTEREST INCOh@ 2 COuEcnONS GYm.0lada0 \ Ium.0 lxQoo.O LANCB 0.0 0.0 \ PARAMETERS. 1260.0 1373.0 ._..-_-1yxI.o Lxa.0 UmaRBXWNSBS zaac! aa. INITIAL BAB=90 couacn0~FuPEcAsT fM(i.O 1zw.o PAYMBHT FQRachs~ KcuJ lsoall 1R:INVESTMENT RATEel. CMXFl.0~PoRaGwc am.0 -3uw BR:BORROWING RATE=2.0% 0.0 0.0 lz.5 ~~~Q~moN lJ.0 ,225 INITIAL CO~E~ION=l2~ s-~~~~ & 53.8 0.0 BmRHsTZNcM(B 0.1 0.0 3.8 INITIAL PAY~~S=l~ WruIgsZ-~a 0.0 0.0 mwIsRBsTlNcQbm clmlnATMiNaTLMEREsT

0.0 0.0

3.8 3.8

23 -2.5 2.0

9 17n.9 XX33.0 0.0 1861.6 ,300.O ua.0 nn.9 2.cm.o -227.1 0.0 0.0 583 LO.2 0.0 10.2 43.3

Fig. 11. Model of an alternative-cash

4 1389.2 1300.0

10 wiL.6 SW.0 0.0 19Y.7 15W.O ZO.0

1861.6 1309.0

Ml.6 0.0 0.0

9343 73 0.0

7.3 M.8

5

6

irm

141.6 1oma 0.0 1331.5 800.0 Ua.0 1,586 lUB.0 cd.6 0.0 0.0 432.1 0.0 0.7 4, 13

IWO.0 01) lwI1.1 1300.0 2m.o 15315 Km.* 5x5 0.0 OJI 959.1 55 0.0 53 6.8

11 1954.7 1700.0 0.0 m52.4 KWJ.0 lcao 1954.7 1706.0 w.7 0.0 0.0 ,**l.l 12.1 0.0 121 63.0

12 20524 mm0 0.0 z155.0 16m.O 203.0 am* l&-O.0 zz.4 (1.0 00 1469.1 IS.6 OAl 156 la.6

flow management problem.

There are many other ways in which the above models can be extended. For example, considering some variables as stochastic, such as lead time and consumption in the material management system and coflections and purchases in the cash flow management problem. Also, rigorous statistical procedures, e.g. replication, can be used to obtain valid estimates of effectiveness improvements over the life of the system. 4.4. Cost -effectiveness analysis It is clear that by modeling a given alternative structure, according to ideas in Sections 4.1-3, we can measure the organizational effectiveness, e.g. slack reduction, the alternative will induce in

Modeling and evaiuation of alternatives

555

practice under given conditions. This will allow us to compare effectiveness of this alternative with respect to the current situation or to other alternatives. The difference of effectiveness of an alternative with respect to a given base situation represents the value the alternative has with respect to that situation. This is what Marschack [12] calls information value. Of course, this value is not only associated to the (possibly new and better) information the alternative is using, but also to, as Marschack’s theory makes clear, the decision rules the alternative uses. Hence, our method COLUMNS INlTIAL, L.12 \PROCESSSTATEDETERMINATXONBASED ONPRRVIOUS ACTION COLLHCI7ONS=0,L?OOJ'RFiVIOUS* LO5 PAYMENTS=0,1000,!500,r500,1300,1000,1000.1500,2000,2000,1500,1700,1800 \PROJECT FUTURESTATE INVOICES= O,l~~~IOUS * 1.03 PURCHASES = 0,1300,1300,1100,800,800.1300.1800,18~~1300,15~,1~0 O~~S~=O~~ COI&ECl-IONFORECAST =O,INlTIALCOLLECI7ON,FRRVlOUS INVOIPAYMENTS FORECASTI 0, INRUL COLLECITON. PREVOIUS FURCHASES+ GTRERHXFEWHS CASH FLOW FORECASI’ = 0,COLLECTIONFORECAST - PAYMENTS fQRfXAfl \ BANK ACCOUNT BALANCE = BAB PROJECEDBAB=BAB,PREVIOUS+CASHFLOW FORHCAST+LGANl+LGANZ+ LOAN3-FREVIOUSLGAN*(l+BRI:) PREVIOUS2 LOANZ+ (1 + BR2+ PREVIOUS 3 LOAN3 * (1 + IR3:) - STII - ST12 - STI3’ ; PREVIOUS sTI1* (1 + IRl:) + PREVIOUS 2 sn2 * (1 + IR2$+ PRHVIous3 sT13* (1 + rR3:) \ DECISION BY DETERMINISTIC ANTICIPATION- ALGORITHMIC (LINEARPROGRAMMING). \LOAN DECISION FOR PAYMENT ONE,Two.THREE PREXIODS IN THE FUTURE LOANl= 0 LOAN2==0 LOAN3=0 \ SHORT TERM INVESTMEIM (STD DECISION-MATURE-Y AT ONI%TWO,THREEPERIOLXS. STII=O sTI2=0 ST13=0 \ SUMM~YOFE~~I~~~ INTEREST INCOh4E=O,O,PRE’.‘IOUSSTII l IRl: + PREVIOUS 2 STI2 * lR2: + PREVIOUS 3 STI3 l IR3: fN-IERESTEXPENSE=O,O,PREVIOUSLOAN1 l BRI: + PREVIOUS 2 LOAN2 * BRZ.z+ PREVIOUS3 LOAN3 * BR3:

NHTIN%BESTlNCGNEE~INCGME-RHERHSTHXPENSE CuMuLATlvENErI~

INCOMFJ=NETINXRESTINCOME,

PREVIOUS+NEH'IMERESTlNCDMH

\ PARAMETERS. IRl: INVESTMENTRATE ONE PERIOD = 1.3% IR2: INVESTMENT RATE TWO PERIOD = 27% IR3: INVESTMENT RATE THREE PEIRIOD = 4.2% =2% BRl:BORROWINGRATEONEPFRIOD BRZ: BORROWING RATE TWO PERIOD = 3.9% BR3: BORROWING RATE THREE PERIOD = 5.8% = 1200 INfTIAL COLLEXXION = 1000 INITIAL PAYMENTS BAB =W OB~C~~ MAIM ~~A~~ NET WTEREST INCOME (12) DECKSIONS LOANlfl) LOAN2(2) LOANl(12) LOAN2(1) LOANZ(2) LOXNZ( 12) LOAN3( 1) LOAN3(2) LOiN3(12) ST11(1) ST1 l(2) STI2f I) STI2f2) STI2(12) STI3(1) STI3(2) STi3( 12)

CONSTRAINTS PROJECTEDBAR( 1) .GE 0 PROJECXED BAR(2) .GE 0 PROJECXED BAB(l2) .GE. 0 Fig.

It-continued

overleaf:

OSCARBARROS

556 PROCXSS STATE DPTXRYIWATION COLLRTIONS PAYMXNTS PROJECT FUTURX STATX. INVOICES PURCHASxS OTHEREXPENSES COLLECTION FORECAST PAYMENTSFORECAST CASH FLOW FORECAST PROJECTED BAB

BASXq~o~,gRXV:~~~,cACTi~N. 1500.0

1323.0 1500.0

1389.2 1300.0

1458.6 1000.0

153 I.5

1000.0 1260.0 t300.0 200.0 1200.0 1000.0 200.0 .O

1323.0 1300.0 200.0 1260.0 1500.0 -240.0 .O

1389.2 1100.0 200.0 1323.0 1500.0 -177.0 .O

1458.6 800.0

153 I.5

1608.1 1300.0 200.0

ioo.0 1389.2 1300.0 89.2 0

800.0 200.0 1458.6 1000.0 458.6 .O

DILCIS~ON BY DXTHRMINISTICANT~CIPAZIOW-A~~~XITHMI~ ~LLNXAR PROQRAMMINOI. LOAN DEClSION FOR pAYMENTONR.TWO.TH~P~lODS INTHHPUTURB. .O .O 87.4 .O .O LOAN1 .O 0 .O .O LOAN2 34:: .O .O .O .O LOAN3 SHORTTERM INVESTMEN (STI)DEClSIONS-MATURITYATONE,TWO,THREXPXRIODS. ST11 202.8 .O 0 0 123.1 ST12 87.2 .O 0 .O .O ST13 0 .O 0 .O 298.8 SUMMART

1000.0

%:i!

531.5 .O

.O .O .O .O 656.:

OX BPFRCTIVXNXSS.

INTEREST INCOME INTEREST XXPXNSX NST INTXRSST INCOME CUMULATIVBNXT INTEREST I

PROCESS STATS DXTSRMINATION COLLECTIONS PAYMXNTS PRO JXCT FUTURX

:8 .O .O

2.6 .O 2.6 2.6

2.4 .O 2.4 5.0

7 8 9 BASXD ON PRBVIOUS ACTION. 1608.1 1608.5 1772.9 1500.0 2000.0 2000.0

.O

1.7 -1.7 3.2

.O 2.0 -2.0 1.2

1.6 I::

2.8

10

11

12

1861.6 1500.0

1954.7 1700.0

2052.4 1800.0

1954.7 1500.0 200.0 1861.6 1500.0 3b1.6 .O

2052.4 1600.0 200.0 1954.7 1700.0 254.7 .O

2155.0 1600.0 200.0 2052.4 1800.0 252.4 1473.3

STATX.

INVOICES PURCHASES OTHEREXPENSSS COLLECTION PORXCAST PAYMENTS FORSCAST CASH FLOW PORSCAST PROJECTSDBAB

1688.5 1800.0 200.0 1608.1 1500.0 108.1 .O

1772.9 1800.0 200.0 1688.5 2000.0 -311.5 .O

1861.6 1300.0 200.0 1772.9 2000.0 -227.1 0

DXCISION BY DRTRR~INISTIC ANTICIPATION-ALGORITHMIC (LINXAR PRO~RA~MIN~~. LoANDECIS~ONPAYM~TONE,TWO, THRRX PERIODSINTN3FU~RE. LOAN1 .O 0 .O 0 .O LOAN2 .O .O 0 .O .O LOAN3 .O 0 .O .O .O SHORTTERM INVXSTMBN (STI)DBClSIONS-MATURITY ATONE, TWO,THREHPXRIODS. ST11 .O .O 0 0 254.7 ST12 .O 0 0 474.2 0 ST13 108.1 0 456.7 0 .O SUMMART OF SFFBCTIVXNXSS. INTEREST INCOME INTEREST EXPENSE NW INTEREST INCOME CUMULAT~VSN~ INTERXST I

.O .O .O 2.8

12.6 .O 12.6 15.4

Fig. 12. Model of alternative ‘L-cash-flow

276 0 27.6 43.0

4.5 O 45 47.5

.O .O .O 47.5

.O .O .O .O .O .O

35,3 35:; 82 8

management problem.

represents a practical way to measure “information value” and implement the ideas proposed by Marschack. Information value or the value associated to a given alternative should then be compared to the cost of going from the base situation to the alternative. Cost determination includes the specification of the regulation and information processing functions that will be computerized and the estimation of development cost by a suitable methodology, such as function points [28]. If cost is less than a discounted value generated over the life of the system, then the alternative is cost-effective and is worth implementing. Otherwise it should be discarded. As an example of alternative evaluation we consider the cash-flow management case modeled in Section 4.3. We take as a base case the current situation presented in Fig. 10 and as alternatives 1 and 2, situations modeled in Figs 11 and 12, respectively. In order to simplify the analysis we assume that “cumulative net interest income” figures for period 12 in the base case and alternatives are representative of what will occur in any given year over the life of the system. Then we easily calculate the alternative’s marginal value with respect to the current situation, which is shown in Fig. 13. In the same figure we also show the equivalent marginal annual costs over the life of the

Modeling and evaluation of attematives ilAL1 JE (COST)

20 30 40 50 60

ALTERNATIVE

1

ALTERNATIVE

2

70 80 90 100

Fig. 13. Marginal value and cost for alternatives.

system for each alternative with respect to the current situationt and net marginal value (marginal value less marginal cost). It is obvious that the values in Fig. 13, based on the specific data presented, lead to recommend alternative 1 over alternative 2. 5.

SUMMARY

AND

CONCLUSIONS

In this work we have tried to show that IS exists in organizations to regulate their processes. This point of view has allowed us to define what an IS should be in generalized terms to accomplish such a regulation. Thus, we have developed a model that includes generalized functions to regulate generalized processes through flows of info~ation, which can serve as a basic pattern to approach the design of any IS. From the generalized IS model we conclude that the design problem in IS consists of searching for alternative structures, including policies and rules for decision functions and associated information processing and flows; i.e. the design of the organizational components within the scope of the IS. We call this design external design, to distinguish it from the problem of internal or software design. fWe do not gjve the technical details of these estimations, but they are based on the function points methodology [28]. IS

w-o

558

OSCARBARROS

Linking the problem of external design to the ideas of modern organization theory has also provided us with criteria to generate alternative structures that most likely would increase organizational effectiveness. This raises the problem of evaluating such alternatives. The modeling approach explained in Sections 4.1-4 provides a practical way to quantitatively perform this evaluation and assure that the external design we arrive at is cost-effective. This type of evaluation also changes the focus for IS justification in practice, from efficiency improvements-usually coming from personnel reductions-to effectiveness improvements. The latter comes from better process regulation which is, according to our approach, the real purpose of IS. Acknowledgement-Support

of this work from FONDECYT

is appreciated.

REFERENCES S. M. McMenamin and J. F. Palmer. Essential Systems Analyses. Yourdon Press, New York (1984). R. Hull and R. King. Semantic database modeling: survey, application and research issues. ACM Comput. Surv. 19, 201-260 (1987). J. Peckham and F. Maryansky. Semantic data models. ACM Cornput. Surv. 20, 153-190 (1988). J. Martin and C. Finkelstein. Information Engineering. Savant Institute, London (1981). M. Page-Jones. The Practical Guide to Structured Systems Design. Yourdon Press, New York (1980). J. D. Couger, M. A. Colter and R. W. Knapp. Advanced System Development/Feasibility Techniques. Wiley, New York (1982). I71 G. B. Davis. Strategies for information requirements determination. IBM Sysr. J. 21, 4-30 (1982). PI B. Ives and G. P. Learmonth. The information system as a competitive weapon. Commun. ACM 27,1193-1201 (1984). [91 G. W. Astley and A. H. Van de Ven. Central perspectives and debates in organization theory. Adhrin. Sci. Q. 28, 245-273 (1983). UOI R. L. Daft and R. H. Lengel. Organizational information requirements, media richness and structural design. Mgr Sci. 32, 554571 (1986). Th. W. Malone. Modeling coordination in organizations and markets. Mgr Sci. 33, 1317-1332 (1987). iti; J. Marschack. Economics of inquiring communicating and deciding. Am. Econ. Rev. 58, l-8 (1968). [I31 J. L. Peterson. Petri Nets. Comput. Surv. 9, 224-252 (1977). P41 B. C. Reiman and A. D. Waren. User-oriented criteria for the selection of DSS software. Commun. ACM 28,1966-1980 (1985). P51 0. Barros. Information requirements and alternatives in information system design. Information Systems 12, 125-136 (1987). WI J. R. Cameron. An overview of JSD. IEEE Trans. Software Engng. SE-12, 222-240 (1986). v71 A. A. Verrijn-Stuart. Themes and trends in information. Trends in Information Systems (Edited by B. Langefors et al.) North Holland, Amsterdam. I181 0. Barros. Modeling and evaluation of alternative information system structures. Proc. Ninth Inl. Conf. on Information Sysfem. (Edited by J. I. De Gross and M. H. Olson) (1988). J. Forrester. Principles of Systems. Wright-Allen, New York (1970). g; 0. Barros. A pragmatic approach to computer assisted system building. Proc. Twenty-Third Annual Hawaii Int. Conf. on Systems Sciences (1990). WI 0. Barros. Alternative architectures in information system design. Beyond Producfivity: Information Systems for Organizational Effectiveness (Edited by Th. M. A. Bemelmans). North Holland, Amsterdam (1984). WI Execucom Systems Co. IFPS User’s Manual, Rel. 11.0 Austin, Texas (1987). ~31 Execucom Systems Co. IFPSiOptimum User’s Manual. Rel. 3.0 Austin. Texas 11983). ~41 A. Roy, L. s. Lasdon and I. ioideman. Extending planning languages to includk opilmization capabilities. Mgl Sci. 32, 360-373 (1986). ~51 L. A. Johnson and D. C. Montogomery. Operations Research in Production Planning Schedule and Control. Wiley, New York (19741. WI E. Silvkr ad H. Meal. A simple modification of the EOQ for the case of varying demand rate and discrete opportunities for replenishment. Prod. Znvtry Mgmt 14, 64-73 (1969). decision support system for a product mix problem. v71 A. Roy, E. E. De Falomir and-L. Lasdon. An opt&&ion-based Interfaces 12, 26-33 (1982). VI A. J. Albrecht and J. E. Gaffney. Software functions, source lines of code, and development effort prediction: a software science validation. IEEE Trans. Software Engng SE-g, 639-648.