INFSY540.1 Information Resources in Management Lesson #4 Chapters 8 Models and Decision Support Copyright 1998 by Jerry Post Information Systems & Technology An information system (IS) is an arrangement of people, data, processes, communications, and information technology that interact to support and improve day-to-day operations in a business as well as support the problem-solving and decision making needs of management and users. Information technology is a contemporary term that describes the combination of computer technology (hardware and software) with telecommunications technology (data, image, and voice networks). A practical way of making data useful. What is an information system? What is an information system? Information System

Transaction Processing System Decision Support System Data-Driven DSS Model-Driven DSS Information Systems Transaction Processing Systems aka Data Processing Systems Decision Support Systems Executive Information Systems

Management Information Systems Expert Systems Office, Workgroup, Personal Information Systems Our text does not have any of these being DSS subsets Data-Driven Decision Support Using Transaction Processing Systems for anything but processing transactions is hard: Not easily accessible Mainframes Cost Mainframe Complexity Mainframes open to many users is risky

Data spread to many databases and computers But users now have powerful PCs with user friendly analysis tools & they want to use them Data-Driven Decision Support History: On Line Transaction Processing (OLTP) DataBase Management System (DBMS) Indexed Sequential Access Method (ISAM) Relational DataBase Management System (RDBMS) Structured Query Language (SQL) Executive Information Systems (EIS) Data Warehouse On Line Analytical Processing (OLAP)

Front- and Back-Office Information Systems Front-office information systems support business functions that reach out to customers (or constituents). Marketing Sales Customer management Back-office information systems support internal business operations and interact with suppliers (of materials, equipment, supplies, and services). Human resources Financial management

Manufacturing Inventory control What is a model? Websters New American Dictionary (1995) One who poses for an artist. An example for imitation or emulation A miniature representation A structural design Model ( verb): to shape, fashion, construct A model is a simplification of something else. Bob Kilmer Models and Analysis INPUTS MODEL OUTPUTS

ASSUMPTIONS 1 Assumptions and Conclusions The aviation instructor had just delivered a lecture on the use of parachutes. And if it doesnt open? someone asked. If it doesnt open? replied the instructor, Well, ... that is whats known as jumping to a conclusion. 1 GIGO INPUTS MODEL OUTPUTS ASSUMPTIONS INPUTS

Constants Parameters Variables OUTPUTS Criteria or MOE Additional Statistics 1 Types of Models Mental Symbolic Mathematical Computer Physical 1 Sample Model

$ Determining Production Levels in Perfect Competition Marginal cost Average total cost price Q* Quantity 1 Order Model vice-presidents Decide if we should produce warehouse manager marketing

manager sales manager sales staff summarize sales orders review sales orders receive sales orders customer check stock to match order production manager decide steps to produce

accounting manager review costs add fixed costs compute costs to produce engineers bill customers Simple Model of Evaluating Custom Orders 1 Models of Physical Items: CAD Computer-aided design. Designers traditionally build models before attempting to create a physical product. CAD systems make it easier to create diagrams and share them with

multiple designers. Portions of drawings can be stored and used in future products. Sample products can be evaluated and tested using a variety of computer simulations. 1 Statistical Decision Models Strategy Decision 100 80 60 40 20 0 1st Qtr 2nd Qtr Actual 3rd Qtr

4th Qtr Forecast Output 1 f ( x) 2 1 x 2 exp 2 Model Data Tactics Operations Company 1 File: C08Fig08.xls Why Build Models?

Understand the Process Prediction Optimization Simulation To conduct "What If" analysis Dangers 1

Human Biases Acquisition/Input Data availability Selective perception Frequency Concrete information Illusory correlation Processing

Inconsistency Conservatism Non-linear extrapolation Heuristics: Rules of thumb Anchoring and adjustment Representativeness Sample size Justifiability Regression bias Best guess strategies Complexity Emotional stress Social pressure Redundancy Output

Question format Scale effects Wishful thinking Illusion of control Feedback Learning on irrelevancies Misperception of chance Success/failure attribution Logical fallacies in recall Hindsight bias 1 File: C08Fig09.xls Prediction 25 20

Economic/ regression Forecast Output 15 10 5 Moving Average Trend/Forecast 0 Q1 Q2 Q3 Q4 Q1 Q2 Q3 Q4 Q1 Q2 Time/quarters 2 File: C08Fig10.xls Simulation Goal or output

variables 25 Output 20 15 Results from altering internal rules 10 5 0 1 2 3 4 5 6

7 8 9 10 Input Levels 2 File: C08Fig08.xls Optimization Maximum Goal or output variables 25 Output 20

Model: defined by the data points or equation 15 10 5 5 3 0 1 2 3 4 5 Input Levels 6 7 8

9 10 1 Control variables 2 Figure 10.2 2 Simulation Websters New American Dictionary (1995) An object that is not genuine The imitation by one system or process of the way in which another system or process works. Simulate (verb): imitate, create the effect or appearance of Handbook of Systems Analysis (1985), E. S. Quade The process of representing item by item and step by step the essential features of whatever it is we are interested in.

2 Bob Kilmers Simple Definitions: Model: simplified representation of something else.* Simulation: means of using or operating a model.** * Something else = a real or proposed entity or system ** Must have inputs and outputs. 2 Building Models Input Process Equation: Output output = f(input,time) Define System

Input - Process - Output Simplifying assumptions System boundary Build Equations Identify parameters (variables you can control) Identify variables you cannot control Define equations for the variables Estimate parameters from data Use Model to transform Inputs into Outputs 2 Modeling Limitations Model complexity Cost of building model Errors in model Data Equations

Presentation and interpretation 2 Models are for... Models are for thinking with. -- Sir M. G. Kendall Models are for experimenting with. Models are for communicating with. Models always have assumptions. (Even though they might not be stated) Models are always wrong. They always have error.

(Question: Is the level of error acceptable?) 2 EOQ Model 2 Appendix: Forecasting Uses Marketing Future sales Consumer preferences/trends Sales strategies Finance

Interest rates Cash flows Financial market conditions HRM Labor costs Absenteeism Turnover Strategy Rivals actions

Technological change Market conditions 3 Forecasting Methods Structural Models Derive underlying models Estimate parameters Evaluate model Focus on explanation and cause Time Series

Collect data over time Identify trends Identify seasonal effects Forecast based on patterns sales P trend S D D Increase in income Q time 3

Structural Equations Demand is a function of Price Income Prices of related products Model QD = b0 + b1 Price + b2 Income + b3 Substitute Data Estimate QD = 1114 - 0.1 Price + 1.2 Income - 1.0 Substitute Forecast 33318 = 1114 - 0.1 (155) + 1.2 (20000) - 1.0 (160)

Need to know (estimate) future price, income, and substitute price. 3 Time Series Components sales Seasonal Trend Dec 1. Trend 2. Seasonal 3. Cycle 4. Random Dec Dec Dec time A cycle is similar to the seasonal pattern, but covers a time period longer than a year.

3 Exponential Smoothing Exponential Smoothing 1600 1500 1400 1300 Raw Data 1200 1100 1000 Smooth:0.20 900 800 1 3 5

7 9 11 13 15 17 19 21 St = Yt + (1 - ) St-1 S is the new data point is the smoothing factor Use Excel: Tools, Data Analysis Exponential Smoothing 3

Exponential Smoothing Choosing the smoothing factor (): It is usually between 0.01 and 0.20 Test multiple values and compare errors: (actual - smooth) * (actual - smooth) Compute the sum. Choose the factor with the least total sum-of-squared error. Larger factors place more importance on recent data, which results in less smoothing. (A2-D2)*(A2-D2) Sum 929,916 Sum 848,686 Sum 769,265 3

Smoothing with Trends Double Exponential Smoothing 34000 32000 30000 28000 Raw Data 26000 Smooth:0.20 24000 22000 20000 1 3 5 7 9

11 13 15 17 19 Apply exponential smoothing and choose smoothing factor (). Apply exponential smoothing a second time to the smoothed data. 3 Forecasting with Exponential Smoothing Forecast for time T+ [ 2] yT 2 ST 1

ST 1 1 T = 20 =1 = 0.2 S20 = 32,064 last of the raw data forecast one period ahead smoothing factor (value at time 20, after one smoothing) S[2] = 33,141 (value at time 20, after second smoothing) Y21 = (2.25)32,064 - (1.25)33,141 = 30,718 3 Time Quantity Trend Difference 1 24917 24484

432 2 26152 24983 1169 3 27297 25482 1816 4 26157 25980 177 5 26710 26479 231 6 26103 26977 -874 7 27981 27476 505 8 26327 27975 -1647 9 24913 28473 -3560 10

28524 28972 -448 11 29774 29470 303 12 29136 29969 -833 13 29332 30468 -1136 14 30306 30966 -660 15 32133 31465 669 16 33329 31963 1366 17 34522 32462 2060 18 34769 32961 1808

19 33355 33459 -104 20 32684 33958 -1274 21 34456 22 34955 23 35454 24 35952 Estimating Trend Yt = b0 + b1(t) Use regression to estimate b0 and b1. Intercept Time Coefficients Std Error t Stat P-value 23985.81 652.48 36.76 2.2E-18

498.60 54.47 9.15 3.4E-08 Plug t into equation to estimate new value (on trend): Y21 = 23,986 + 498.6 * (21) = 34,456 Result is the prediction on the trend, with no random factors and no cycles. 3 An Overview of Decision Support Systems File: C08Fig11.xls DSS: Decision Support Systems Sales and Revenue 1994 300 Model 250 Legend

200 t da a to a ly na ze sales 154 163 161 173 143 181 revenueprofit prior 204.5 45.32 35.72

217.8 53.24 37.23 220.4 57.17 32.78 268.3 61.93 47.68 195.2 32.38 41.25 294.7 83.19 67.52 su re lts 150 Sales Revenue Profit Prior 100 50 0 Jan Feb Mar

Apr May Jun Output Database 4 Characteristics of Decision Support Systems Handle lots of data from various sources Report & presentation flexibility Text and graphics capabilities Support drill down analysis Complex analysis, statistics, and forecasting

Optimization, satisficing, heuristics Simulation What-if analysis Goal-seeking analysis 4 Figure 10.14 4 Capabilities of a DSS Support all problem-solving phases Support different decision frequencies Support different problem structures Support various decision-making levels

4 The Model Base Financial models Statistical analysis models Cash flow Internal rate of return Averages, standard deviations Correlations Regression analysis Graphical models

Project management models 4 Table 10.3 4 Group Decision Support Systems Characteristics of a GDSS Special design Ease of use Flexibility Decision-making support 4 Characteristics of a GDSS

Anonymous input Reduction of negative group behavior Parallel communication Automated record keeping 4 Figure 10.18 4 Executive Support Systems (ESS) Tailored to individual executives Easy to use Drill down capabilities

Access to external data Can help when uncertainty is high Future-oriented Linked to value-added processes. 5 Capabilities of an ESS Support for defining an overall vision Support for strategic planning Support for strategic organizing & staffing Support for strategic control Support for for crisis management 5

Easy access to data Graphical interface Non-intrusive Drill-down capabilities EIS: Enterprise Information System EIS Software from Lightship highlights easeof-use GUI for data look-up. 5 Enterprise IS Sales Production Costs Distribution Costs Fixed Costs Executives Central Management

Production Costs South North Overseas 5000 4500 4000 3500 3000 2500 2000 1500 1000 500 0 r EIS o f a t Da South North

Overseas 1993 1994 1995 1996 Production: North Data Data Sales Data Distribution Data Item# 1995 1994

1234 2938 7319 542.1 631.3 753.1 442.3 153.5 623.8 Production 5