THE VEST POCKETGUIDE TOINFORMATION TECHNOLOGY 2nd phần 7 doc

❍ Eight powerful methods for more accurate forecasts ❍ Spreadsheet models, complete with graph, ready-to-usewith your numbers The Sales & Market Forecasting Toolkit offers a variety of f

to develop pricing policies that will maximize total salesrevenues This is usually a function of price elasticity If theproduct is highly price sensitive, a reduction in price cangenerate a substantial increase in sales, which can result inhigher revenues A product that is relatively insensitive toprice can have its price substantially increased without alarge reduction in demand Exhibit 13.4 shows the relation-ships between price elasticity (ep) and sales revenue (S),which can aid a firm in setting its price.

Computer programs exist that help determine price ticity and various pricing policies With the aid of computersoftware for spreadsheets and statistical packages, the mar-keting managers can typically develop what-if scenarios inwhich they can alter factors to see price changes on futuredemand and total revenues

elas-Price e p > 1 e p = 1 e p < 1

Price rises S falls No change S rises Price falls S rises No change S falls

One of the widely used pricing methods, especially inlarge corporations, is pricing to achieve a targeted rate

of return on investment (ROI) Furthermore, there is anincreasing tendency among firms to adopt some form

of target ROI pricing This is mainly due to a growingawareness of the need to integrate pricing policy withthe objective of achieving a satisfactory rate of return oncapital invested ROI pricing is certainly the mostwidely used pricing method today The use of spread-sheet software and what-if analysis can be readily applied

to the area of product pricing

The conventional ROI pricing technique is generallyalong the following lines: a standard volume of produc-tion is estimated; the variable cost per unit is calculatedfor this level of production; and fixed factory overhead,selling, and administrative expenses are allocated overthe number of units at standard volume of production.Depreciation on assets is included in the fixed costs Therate of depreciation is either an estimated rate, which inthe opinion of the management reflects the fall in thevalue of assets, or more likely, the depreciation rateallowed under the tax law is generally adopted Themarkup per unit is arrived at by calculating the desireddollar return (on the total capital invested—i.e., debt aswell as equity) and dividing by the number of units

at standard volume The return on investment rateexpected is determined by management according to its

Comprehensive Sales Planning 209

Sales Analysis

Sales analysis assists managers in identifying products, salespersonnel, and customers who are contributing to profitsand those who are not Several reports can be generated tohelp marketing managers make good sales decisions The

sales-by-product report lists all major products and their salesfor a period of time, such as a month This report showswhich products are doing well and which ones need

improvement or should be discarded altogether The salesperson report lists total sales for each salesperson foreach week or month This report can also be subdivided byproduct to show which products are being sold by each

sales-by-salesperson The sales-by-customer report is a useful way to

identify high- and low-volume customers

expectations of what constitutes a fair return Tax aspectsare generally ignored The outline of an ROI pricingmodel (with assumed figures) is presented below:

Substituting the assumed figures in the above formula:

The spreadsheet contains parameters for what-if tivity) analysis on three levels: normal, optimistic, andpessimistic Consequently, the template generates prod-uct prices under optimistic, pessimistic, and normalexpectations of the person making the pricing decision

(sensi-A printout of the worksheet with assumed figures isshown in Exhibit 13.5

SP RR+FC–( )t [FC+D+INT+OI–CI]+VC unit⁄

1–t( ) ×X -

=

SP 30000 20000+ – (.40) [ 20000 + 3000 + 5000 45000 100000 + – ] 5

1 40 –

( ) ( 100000 ) -

=

$11.83 per unit

=

Popular Forecasting and Statistical Software 211

POPULAR FORECASTING AND

STATISTICAL SOFTWARE

There are numerous computer software packages that areused for forecasting purposes They are broadly dividedinto two major categories: forecasting software and general-purpose statistical software Some programs are stand-alone,while others are spreadsheet add-ins Still others are tem-plates A brief summary of some popular programs follows

Sales & Market Forecasting Toolkit

It is a Lotus 1-2-3 template that produces sales and marketforecasts, even for new products with limited historical data

❍ Eight powerful methods for more accurate forecasts

❍ Spreadsheet models, complete with graph, ready-to-usewith your numbers

The Sales & Market Forecasting Toolkit offers a variety

of forecasting methods to help you generate accurate ness forecasts even in new or changing markets with lim-ited historical data The forecasting methods include:

You can use the built-in macros to enter data into yourforecast automatically For example, enter values for thefirst and last months of a 12-month forecast The com-pounded-growth-rate macro will automatically computeand enter values for the other 10 months

Forecast! GFX

Forecast! GFX is a stand-alone forecasting system that canperform five types of time-series analysis: seasonal adjust-ment, linear and nonlinear trend analysis, moving-averageanalysis, exponential smoothing, and decomposition Trendanalysis supports linear, exponential, hyperbolic, S-curve, andpolynomial trends Hyperbolic trend models are used to

analyze data that indicates a decline toward a limit, such asthe output of an oil well or the price of a particular model ofpersonal computer Forecast! GFX can perform multiple-regression analysis with up to 10 independent variables.

ForeCalc

ForeCalc, a Lotus add-in, uses nine forecasting techniquesand includes both automatic and manual modes, and elimi-nates the need to export or reenter data In automatic mode,just highlight the historical data in your spreadsheet, such

as sales, expenses, or net income; then ForeCalc tests severalexponential-smoothing models and picks the one that bestfits your data

Forecast results can be transferred to your spreadsheetwith upper and lower confidence limits ForeCalc generates

a line graph showing the original data, the forecasted ues, and confidence limits

val-ForeCalc can automatically choose the most accurateforecasting technique:

❍ Simple one-parameter smoothing

❍ Holt’s two-parameter smoothing

❍ Winters’s three-parameter smoothing

❍ Trendless seasonal models

❍ Dampened versions of Holt and Winters’s smoothingForeCalc’s manual mode lets you select the type of trendand seasonality, yielding nine possible model combinations.You can vary the type of trend (constant, linear, or damp-ened), as well as the seasonality (nonseasonal, additive, ormultiplicative)

StatPlan IV

StatPlan IV is a stand-alone program for those who stand how to apply statistics to business analysis You canuse it for market analysis, trend forecasting, and statisticalmodeling

under-StatPlan IV lets you analyze data by range, mean,median, standard deviation, skewdness, kurtosis, correlationanalysis, one- or two-way analysis of variance (ANOVA),cross-tabulations, and t-test

The forecasting methods include multiple regression,stepwise multiple regression, polynomial regression, bivari-ate curve fitting, autocorrelation analysis, trend and cycleanalysis, and exponential smoothing

The data can be displayed in X-Y plots, histograms, series graphs, autocorrelation plots, actual versus forecastplots, or frequency and percentile tables

time-Geneva Statistical Forecasting

Geneva Statistical Forecasting, stand-alone software, canbatch-process forecasts for thousands of data series, pro-vided the series are all measured in the same time units (days,

Popular Forecasting and Statistical Software 213

weeks, months, and so on) The software automatically triesout as many as nine different forecasting methods, includingsix linear and nonlinear regressions and three exponential-smoothing techniques, before picking the one that best fitsyour historical data

The program incorporates provisions that simplify andaccelerate the process of reforecasting data items Once youcomplete the initial forecast, you can save a data file thatrecords the forecasting method assigned to each line item.When it is time to update the data, simply retrieve the fileand reforecast, using the same methods as before

SmartForecasts

SmartForecasts, a stand-alone forecasting software program,features the following:

❍ Automatically chooses the right statistical method

❍ Lets you manually adjust forecasts to reflect yourbusiness judgment

❍ Produces forecast results

SmartForecasts combines the benefits of statistical andjudgmental forecasting It can determine which statisticalmethod will give you the most accurate forecast and handlesall the math Forecasts can be modified using the program’sEyeball utility You may need to adjust a sales forecast toreflect an anticipated increase in advertising or a decrease

in price SmartForecasts summarizes data with descriptivestatistics, plots the distribution of data values with histo-grams, plots variables in a scattergram, and identifies lead-ing indicators

You can forecast using single- and double-exponentialsmoothing, and simple- and linear-moving averages Iteven builds seasonality into your forecasts using Winters’sexponential smoothing, or you can eliminate seasonality byusing time-series decomposition and seasonal adjustment

In addition, SmartForecasts features simultaneous series forecasting of up to 60 variables and 150 data pointsper variable, offers multivariate regression to let you relatebusiness variables, and has an Undo command for mistakes

multi-Tomorrow

Tomorrow, a stand-alone forecasting package, uses an mized combination of linear regression, single exponentialsmoothing, adaptive rate response single exponential smooth-ing, Brown’s one-parameter double exponential smoothing,Holt’s two-parameter exponential smoothing, Brown’s one-parameter triple exponential smoothing, and Gardner’s three-parameter damped trend Some of the main features include:

opti-❍ There is no need to reformat your existing sheets Tomorrow recognizes and forecasts formula

spread-cells (containing totals and subtotals, for example) Ithandles both horizontally and vertically orientedspreadsheets It accepts historical data in up to 30 sep-arate ranges.

❍ Allows you to specify seasonality manually or lates seasonality automatically

calcu-❍ Allows you to do several forecasts of different timeseries (for example, sales data from different regions)

❍ Undo command restores original spreadsheet

❍ Browse feature allows you to look at any part of thespreadsheet (including the forecast) without leavingTomorrow

❍ Checks for and prevents accidental overlaying ofnonempty or protected cells

❍ Optional annotation mode labels forecast cells, lates MAPE, and, when seasonality is automaticallydetermined, describes the seasonality

calcu-❍ Comprehensive context-sensitive online help

Forecast Pro

Forecast Pro, a stand-alone forecasting program, is the ness software that uses artificial intelligence A built-inexpert system examines your data Then it guides you toexponential smoothing, Box-Jenkins, or regression—which-ever method suits the data best

busi-MicroTSP

MicroTSP is a stand-alone software package that providesthe tools most frequently used in practical econometric andforecasting work It covers the following:

1 Descriptive statistics

2 A wide range of single-equation estimation techniquesincluding ordinary least squares (multiple regression),two-stage least squares, nonlinear least squares, andprobit and logit

Forecasting tools include exponential smoothing ing single exponential, double exponential, and Winters’ssmoothing) and Box-Jenkins methodology

(includ-Sibyl/Runner

Sibyl/Runner is an interactive, stand-alone forecasting tem In addition to allowing the usage of all major forecast-ing methods, the package permits analysis of the data,suggests available forecasting methods, compares results,and provides several accuracy measures in such a way that

sys-Popular Forecasting and Statistical Software 215

it is easier for the user to select an appropriate method andforecast needed data under different economic and environ-mental conditions For details, see Makridakis, S., Hodgsdon,and S Wheelwright, “An Interactive Forecasting System,”

American Statistician, November 1974

Other Forecasting Software

There are many other forecasting software programs such

as Autocast II, 4 Cast, and Trendsetter Expert Version

General-Purpose Statistical Software

There are numerous statistical software programs that can

be utilized in order to build a forecasting model Some ofthe more popular ones include:

❍ SAS Application System

is regularly called upon to provide forecasts A personalcomputer with a spreadsheet is a good beginning, but thestand-alone packages currently available provide the mostaccurate forecasts and are the easiest to use In addition,they make several forecasting models available and canautomatically select the best one for a particular data set

D ECISION S UPPORT S YSTEMS

DISTINGUISHING AMONG TPS, MIS, EIS, DSS, AND ES

As discussed in Chapter 1, information systems are guished by the type of decisions they support, the operatorwho uses the system, the management control level of thesystem, the function of the system, and its attributes (seeExhibit 1.1) There are information systems to supportstructured decisions, unstructured decisions, and anything

distin-in between At the strategic level of management, decisionsare unstructured, and decision styles may differ signifi-cantly among managers Furthermore, a specific decisionproblem may occur only once Thus, information systemsdeveloped for this level often are decision specific Once thedecision is made, the information system used for it is nolonger applicable in its current form For subsequent deci-sions, the system must be modified or discarded—a devel-opment that has major implications for the design ofinformation systems Whereas executive information sys-tems and decision support systems aid in decisions that areunstructured, transaction processing systems and expertsystems aid in decisions that are structured

The manager who uses the information system helpsdistinguish the system Transaction processing systems(TPSs) are used at the operational level of an organizationsuch as by clerks or secretaries Executive information sys-tems (EISs) are used specifically by personnel at the seniormanagement level such as vice presidents or presidents of

an organization Decision support systems (DSSs) are used

by middle management such as managers of the accountingdepartment Expert systems (ESs) are used by personnel atall levels of an organization

Another factor that distinguishes information systems isthe function of the systems Transaction processing systemswere established to computerize manual systems Execu-tive information systems (EISs) were designed to aid seniormanagers in decision making Decision support systems

Decision Support Systems (DSSs) 217

were designed to aid middle managers in decision making,and expert systems (ESs) were designed to aid all personnel

in decision making

The final distinguishing factor of information systems isthe attributes of the system Transaction processing sys-tems are used to handle day-to-day transactions such as theaccounts payable system of an organization Attributes ofexecutive information systems include visual summaries offorecasts and budgets of an organization Decision supportsystem attributes include visual displays of the sales,income or interest estimates for the day, month, or year Expertsystem attributes include systems that assess bad debts orauthorize credit

DECISION SUPPORT SYSTEMS (DSSs)

A DSS is a computer-based information system that assistsmanagers in making many complex decisions, such as deci-sions needed to solve poorly defined or semistructuredproblems Instead of replacing the manager in the decisionprocess, the DSS supports the manager in his or her applica-tion of the decision process In other words, it is an auto-mated assistant that extends the mental capabilities of themanager Most authorities view the DSS as an integral part

of the MIS, in that its primary purpose is to provide making information to managerial decision makers A DSSallows the manager to change assumptions concerningexpected future conditions and to observe the effects on therelevant criteria As a result of these direct benefits, a DSSenables the manager to gain a better understanding of thekey factors affecting the decision It enables the manager toevaluate a large number of alternative courses of actionwithin a reasonably short time frame

decision-A DSS summarizes or compares data from either or bothinternal and external sources (see Exhibit 14.1) Internalsources include data from an organization’s database such

as sales, manufacturing, or financial data Data from nal sources includes information on interest rates, popula-tion trends, new housing construction, or raw materialpricing

exter-DSSs often include query languages, statistical analysiscapabilities, spreadsheets, and graphics to help the userevaluate the decision data More advanced decision supportsystems include capabilities that allow users to create amodel of the variables affecting a decision With a model,users can ask what-if questions by changing one or more ofthe variables and seeing what the projected results would

be A simple model for determining the best product pricewould include factors for the expected sales volume at eachprice level Many people use electronic spreadsheets for

simple modeling tasks A DSS is sometimes combined withexecutive information systems (EISs) DSS applicationsused in business include systems that estimate profitability,plan monthly operations, determine the source and applica-tion of funds, and schedule staff.

PALISADE’S DECISIONTOOLS SUITE

Palisade’s DecisionTools Suite is a DSS tool in the area ofrisk and decision analysis It includes such programs as

@RISK, @RISK for Project, TopRank, PrecisionTree, BestFit,and RISKview These programs analyze risk, run MonteCarlo simulations, perform sensitivity analyses, and fit data

to distributions

❍ @RISK is a risk analysis and simulation add-in forMicrosoft Excel and Lotus 1-2-3 It is the risk analysistool Replace values in your spreadsheet with @RISKdistributions to represent uncertainty, then simulateyour model using powerful Monte Carlo simulationmethods @RISK recalculates your spreadsheet hun-dreds (or thousands) of times The results: distribu-tions of possible outcome values! Results are displayedgraphically and through detailed statistical reports

❍ @RISK for Project adds the same powerful MonteCarlo techniques to Microsoft Project models, allow-ing users to answer questions such as, What is thechance the project will be completed on schedule?

❍ TopRank is a what-if analysis add-in for eitherMicrosoft Excel or Lotus 1-2-3 for Windows Take anyspreadsheet model, select the cells that hold yourresults, and TopRank automatically determines whichspreadsheet values affect your results the most.TopRank then ranks the values in order of impor-tance Your results can be displayed in Tornado, Spi-der, and Sensitivity high-resolution graphs, allowing

Graphical

Large database

Integrates many sources of data

Report and presentation flexibility

Geared toward individual decision-making styles

Modular format

Optimization and heuristic approach

What-if and simulation

Goal-seeking and impact analysis

Performs statistical and analytical analysis

SYSTEM (DSS)

DSS Applications 219

you to easily understand the outcome at a glance.TopRank works easily and effectively with @RISK byidentifying the critical cells that users should concen-trate on when running Monte Carlo simulations

❍ PrecisionTree is a powerful, innovative decision sis tool You can enter decision trees and influence dia-grams directly in your spreadsheet models, and detailall available decision options to identify the optimaldecision Your decision analysis factors in your atti-tudes toward risk and the uncertainty present in yourmodel Sensitivity analysis identifies the critical fac-tors that affect the decision you’ll make PrecisionTree

analy-is a real plus for outlining all available options for adecision or identifying and presenting the best course

up to 26 distribution types using advanced tion algorithms Results are displayed graphically andthrough an expanded report that includes goodness-of-fit statistics BestFit distributions can be used directly

optimiza-in @RISK for Excel, Lotus 1-2-3, and Microsoft Projectmodels

❍ RISKview is the distribution viewing companion to

@RISK, @RISK for Project, or BestFit It is a powerfultool for viewing, assessing, and creating probabilitydistributions

Manage-2 Manufacturing Decision Support System (MDSS),

devel-oped at Purdue University to support decisions inautomated manufacturing facilities, is especially use-ful for CAD/CAM operations

3 RCA has developed a DSS to deal with personnel

problems and issues The system, called Industrial Relations Information Systems (IRIS), can handle prob-lems that may not be anticipated or that may occuronce, and can assist in difficult labor negotiations

4 The Great Eastern Bank Trust Division developed a

DSS called On-line Portfolio Management (OPM) that

can be used for portfolio and investment ment The DSS permits display and analysis of vari-ous investments and securities

manage-5 ReolPlan, a DSS to assist with commercial real estate

decisions, is useful for various decision aspects ofpurchasing, renovating, and selling property

6 EPLAN (Energy Pion) is a DSS being developed by the

National Audubon Society to analyze the impact ofU.S energy policy on the environment

7 The Transportation Evacuation Decision Support System (TEDSS) is a DSS used in nuclear plants in Virginia Itanalyzes and develops evacuation plans to assistmanagers in crisis management decisions regardingevaluation times and routes and the allocation of shel-ter resources

8 The U.S Army has developed an enlisted manpowerDSS to help with recruitment, training, education,reclassification, and promotion decisions It encom-passes simulation and optimization to model person-nel needs and requirements It interacts with anonline database and other statistical analysis softwarepackages

9 Voyage Profitability Estimator is a DSS used by a

ship-ping firm to compute the income from decisionsaffecting charter rates to be charged for particulartrips The system saves time and makes it possible toevaluate trade-offs between speed and fuel usage.The analysis involves ship and voyage characteris-tics such as tonnage, rate of fuel consumption, andport cost

10 Monthly Plan Calculations serves as a corporate

bud-geting tool to measure the levels of manpower needed

to perform various functions, to calculate costs, and ingeneral to evaluate the adequacy of proposed opera-tional plans Using simple formulas, this system cal-culates the cost of materials and inventory, amongother items, based on input that consists of monthlyproduction and shipment plans Typically, the system

is used iteratively in an attempt to generate a planthat is sufficiently profitable and that meets the com-pany’s goal of maintaining reasonable-level produc-tion in spite of the seasonal nature of the product

11 Source and Application of Funds is an online budget of

source and applications of funds that has been usedfor operational decision making and financial plan-ning in an insurance company to provide monthlycash flow figures The DSS “output” is used atweekly meetings of an investment committee to help

Executive Information Systems (EISs) 221

in allocating funds across investment areas and tominimize the amount of cash that is left idle in banks

12 Interactive Audit Staff Scheduling Systems, an integer

programming model, was designed by Balachandranand Zoltners to assist public accounting firms inscheduling their audit staff in an optimal and effec-tive manner The computerized management supportsystem for scheduling staff to an audit can includethe basic model along with a judgmental schedulingsystem and a scheduling information database Moti-vation, morale, turnover, and productivity of theaudit staff can all be affected by scheduling In thescheduling process, the audit firm needs to considerits audit philosophy, objectives, staff size, rotationalplans, and auditor evaluation Many feasible auditstaff schedules may fill these needs, but the firmneeds to select the schedule that best meets its ownobjectives

EXECUTIVE INFORMATION SYSTEMS (EISs)

An executive information system (EIS) is a DSS made cially for top managers and specifically supports strategicdecision making An EIS is also called an executive supportsystem (ESS) It draws on data not only from systems inter-nal to the organization but also from those outside, such asnews services and market research databases The EIS userinterface often uses a mouse or a touch screen to help execu-tives unfamiliar with using a keyboard One leading systemuses a remote-control device similar to those used to control

spe-a television set An EIS might spe-allow senior executives to cspe-all

up predefined reports for their personal computers, whetherdesktops or laptops They might, for instance, call up salesfigures in many forms—by region, by week, by fiscal year,

by projected increases The EIS includes capabilities for lyzing data and doing what-if scenarios

ana-Another aspect of the EIS user interface is the graphicpresentation of user information The EIS relies heavily ongraphic presentation of both the processing options and data.Again, this is designed to make the system easier to use.Because executives focus on strategic issues, the EISoften has access to external databases such as the Dow JonesNews/Retrieval service Such external sources of informa-tion can provide current information on interest rates, com-modity prices, and other leading economic indicators.Exhibit 14.2 presents the attributes of an executive informa-tion system

A popular EIS software is Xecutive Pulse developed byMegatrend System, Inc It is a Windows-based executiveinformation system The software interfaces with many

popular LAN-based accounting applications It providesdecision makers with easy access to financial and salesinformation including trend analysis using drill-down anddrill-across technology Hundreds of charts, graphs, andviews are available with a mouse click The system extractsdata from accounting history files, builds a database, andstores up to three years of information for each accountingperiod Users can drill down through five organizationallevels, compare actual versus history and actual versusbudget, and display report or graphic results XecutivePulse features extensive sales, cash flow, and humanresource analysis, plus daily trends for accounts receivable,accounts payable, margins, sales, and inventory.

Limitations of Current EISs

Although they offer great promise, many EISs have notbeen successfully implemented and many executives havestopped using them A common reason cited in severalfailed attempts is the mistake of not modifying the system

to the specific needs of the individual executives who willuse the system For example, many executives prefer tohave information presented in a particular sequence withthe option of seeing different levels of supporting detailedinformation such as cost data on a spreadsheet The desiredsequence and level of detail varies for each executive Itappears that an EIS must be tailored to the executives’requirements or the executives will continue to managewith information they have obtained through previously

established methods This limitation can be corrected by loring the software based on the particular needs of themanagers within the specific company After the softwarehas been appropriately modified, it will have significantpractical applications

tai-EIS Applications

There are many EIS applications for managers, includingthose described in the following sections

EIS IN MEASURING PRODUCTIVITY

This application bears on management’s concern over ductivity Management may use both internal and external

pro-Graphical

Easy-to-use interface

Broad, aggregated, perspective

Different data sources

Optionally expand to detail level

Provide context

Timeliness crucial

SYSTEMS (EISS)

Executive Information Systems (EISs) 223

information extracted from the EIS to show how ity in an organization has declined in recent years Financialdata can be retrieved from the EIS database to demonstratehow increases in unit labor costs over time have been pri-marily responsible for significant increases in the product’sunit cost and have been damaging to the company’s com-petitiveness by forcing increases in the product’s sellingprice Executives can also compare company sales (internaldata) to industry sales trends (external data) from the EIS toproject market share changes in response to changes in sell-ing price

productiv-External information may also be extracted from the EISdatabase to indicate how competitors achieve greater effi-ciency by using less labor and more advanced technology tomanufacture a quality product at a materially lower unitcost As a result, management may demonstrate that thecompetition is able to sell greater quantities of their prod-ucts at lower prices This information may provide justifica-tion for closing the unprofitable plant and opening amodern facility that will enable a company to be more com-petitive in the industry

EIS IN PRODUCT COSTING DECISIONS

Resolving the conflict between profitability in the short runand increasing market share in the long run requires a mix

of both external and internal data for a rational decision.Executives need information on product demand and elas-ticity, competing products and strategies, the economy, andother factors such as the cost of manufacturing the productand trade-offs that exist relative to different product qualitylevels under different cost assumptions Some questionsexecutives may raise are:

❍ What is the current level of quality and how does thelevel differ from the desired level?

❍ What is the current full cost of producing a unit andhow does the amount differ from the full cost at thedesired level of quality?

❍ What costs are variable over different levels of uct quality?

prod-❍ What costs are controllable relative to producing andselling the products?

EIS can provide data for solutions to some of these tions by computation Many internal decisions depend onassumptions and measurements that require judgment andmay be subject to different interpretations In product cost-ing decisions, issues involving appropriate cost and prod-uct quality trade-offs are equally subjective and unlikely tohave a unique interpretation

A RTIFICIAL I NTELLIGENCE (AI) AND

E XPERT S YSTEMS (ES S )

WHAT IS ARTIFICIAL INTELLIGENCE (AI)?

Artificial intelligence is the application of human ing techniques to machines Artificial intelligence systemsuse sophisticated computer hardware and software to simu-late the functions of the human mind Expert systems arethe most promising applications of artificial intelligenceand have received the most attention

reason-Expert systems are computer programs exhibiting ioral characteristics of experts Expert systems involve thecreation of computer software that emulates the way peoplesolve problems Like a human expert, an expert systemgives advice by drawing upon its own store of knowledgeand by requesting information specific to the problem athand Expert systems are not exactly the same thing as deci-sion support systems A DSS is computer-based softwarethat assists decision makers by providing data and models

behav-It performs primarily semistructured tasks, whereas anexpert system is more appropriate for unstructured tasks.Decision support systems can be interactive just like anexpert system But, because of the way decision supportsystems process information, they typically cannot be usedfor unstructured decisions that involve nonquantitativedata Unlike expert systems, decision support systems donot make decisions but merely attempt to improve andenhance decisions by providing indirect support withoutautomating the whole decision process

Some general characteristics indicate whether a givenbusiness application is likely to be a good candidate for thedevelopment of an expert system For example, the appli-cation must require the use of expert knowledge, judg-ment, and experience The business problem must have aheuristic nature and must be defined clearly The area ofexpertise required for the application must be well defined

Expert Systems 225

and recognized professionally, and the organization ing the expert system must be able to recruit an expert who iswilling to cooperate with the expert system’s developmentteam The size and complexity of the application must bemanageable in the context of organizational resources, avail-able technical skills, and management support

develop-EXPERT SYSTEMS

An expert system (ES), sometimes called a knowledge tem, is a set of computer programs that perform a task atthe level of a human expert Expert systems are created onthe basis of knowledge collected on specific topics fromhuman experts, and they imitate the reasoning process of ahuman being Expert systems have emerged from the field

sys-of artificial intelligence, which is the branch sys-of computerscience that is attempting to create computer systems thatsimulate human reasoning and sensation We describe arti-ficial intelligence in more detail later in the chapter.Expert systems are used by management and nonman-agement personnel to solve specific problems, such as how

to reduce production costs, improve workers’ productivity,

or reduce environmental impact Based on methodicallyusing a narrowly defined domain of knowledge that is builtinto computer programs, the expert system comes up with asolution to a problem much the same way an expert would.The key to the definition is that the domain must be nar-rowly defined An expert system cannot (at this point) bedeveloped to give useful answers about all questions—it islimited, as a human expert is limited, to a particular field.For example, one expert system would not tell the control-ler both whether to lease or buy a piece of equipmentbased on the tax differences and also whether a pendingbusiness combination needs to be treated as a pooling or as

a purchase

How Expert Systems Work

Expert systems are usually considered to have six majorcomponents The relationships of these components areillustrated in Exhibit 15.1 Based on the relationships illus-trated in Exhibit 15.1, it is apparent that expert systemsmust work interactively with system users to help themmake better decisions The system interacts with the user bycontinuously asking for information until it is ready tomake a decision Once the system has sufficient informa-tion, an answer or result is returned to the user It is essen-tial to note that not only must the system assist in making

1 Knowledge rule database: Contains the rules,

problem-solving knowledge, and cases used when makingdecisions

2 Domain database: The set of facts and information

rel-evant to the domain (area of interest)

3 Database management system: Controls input and

management of both the knowledge and domaindatabases

4 Inference engine (Processing System): Contains the

inference strategies and controls used by experts tomanipulate knowledge and domain databases It isthe brain of the expert system It receives the requestfrom the user interface and conducts analysis, rea-soning, and searching in the knowledge base Theinference engine aids in problem solving such as byprocessing and scheduling rules It asks for addi-tional information from the user, makes assump-tions about the information, and draws conclusionsand recommendations The inference engine mayalso determine the degree to which a recommenda-tion is qualified and in the case of multiple solutionsrank them

E XHIBIT 15.1 EXPERT SYSTEM RELATIONSHIPS

DatabaseManagement

System

InferenceEngine

DomainDatabase

KnowledgeAcquisitionFacility

User

Inference

Knowledge

Database