Final Research (2)

Chapter 1
Introduction
1.1. Rationale
The year 1988 marked a round turn in the Vietnamese economy, which concerned initiating
the economic transformation from the former centrally planned to a market economic oriented
system with state management.
To cope with the new environment, the banking sector started its deregulation. This is featured
in the transformation from a uni-type to two-tier system, wherein the State Bank of Vietnam
play the role of cash issuer and controller of commercial banking, and other banks work
toward establishing a true business platform.
During 1988-1990, the commercial banking business was ‘booming’ in terms of the number of
new banks established and in lending activities. However, due to the poor quality in operation
and insecurity, nearly all credit cooperatives and joint stock banks came to bankruptcy. Billions
of dong were frozen in bad debts within state owned banks and as assets in shareholding
banks which still operated.
To overcome this situation, one requires a suitable credit policy to:
• Extend credit to every economic sector, public and private; including households.
• Diversify funding instruments in light of market developments.
• Restructure bank loans with further emphasis on medium and long term credit rather than
short term lending.
• Enhance credit quality, and strictly manage credit risks.
• Diversify the credit market for banks.
In order to perform this credit policy, Vietnamese banking system initiated computerization,
and currently, payments and statistics have been computerized. By the end of 1993, banks set
up LAN internally. In the coming years, the following steps will be taken:
• Computer link between State Bank of Vietnam and banks.
• Computer link among banks.
• Computer link between banks and customers.
Along with computerization, different types of advanced payment facilities are gradually
introduced; like credit cards, and ATM installation to serve residents and foreign travelers.
1

In that situation, efficiency and effectiveness are critical for the success of the banking
industry. They are not only dependent upon managerial skills but also the adoption of new
technology.
To control credit quantity and quality, it is necessary to have a new technology to enhance
credit quality, and manage credit risks. Determining which loan applicant should be extended
credit, as well as the amount of credit, are major practical decisions that confront commercial
bank lending officers, credit analysts, and loan committees. These decision makers must
assess the financial health of an applicant, which requires analysis of both quantitative and
qualitative information on the outlook of the company. To correctly perform the analysis,
commercial loan officers need to have an understanding of the primary Cs of lending: credit,
collateral, capital, capacity, and character. As commercial loan officers evaluate a company,
they examine numerous financial ratios, percentages, and trends, and perform many
interrelated analyses. Because the complex nature of the problem requires experience and
precludes the use of simple algorithms, expert systems are appropriate for this type of loosely
structured, complex problem.
An expert system is a software system that imitates the reasoning results of human experts in
a well defined domain. It aims to generate advice about problems in the domain comparable to
the advice that a human expert would deduce for the same problems. In recent years,
practitioners have become more familiar with the technology and the technology has
advanced to become more supportive of business applications.
Recent financial applications of expert systems include:
• Decision analysis in securities trading,
• Cashflow analysis, and
• Venture capital analysis for small telecommunication companies.
Another problem area that can benefit from this technology is commercial loan analysis, the
process of evaluating a company’s financial strengths and weaknesses.
1.2. Objective
The purpose of this research is to design an expert system that helps commercial bank
lending officers, credit analysts, and loan committees to reduce the time devoted to the
analysis in evaluating loan applicants and to improve the quality of the evaluation.

1.3. Scope of the research
The system is designed to analyze commercial loans for industrial or retail borrowers. The
system only focuses on the credit granting decision and does not consider other aspects of
the credit granting decision process such as keeping track of collateral maturities, collection
follow-up, etc. In addition, the system does not provide a specific evaluation of economic or
competitive factors within a given industry.
Chapter 2
2
Literature Review
2. 1. Financial Perspective
Automation in Banking procedures has been evenly spread out over the last decades with the
advent of fast computing machines. Automation in banking began with the use of computers
for performing rote clerical tasks and moved on rapidly to automatic transaction processing.
Soon, a need was felt for using computers for making managerial decisions, or assist the
upper management in making such decisions. One of the departments i.e. credit department
felt the urgent need for the same and systems based on statistical, expert systems began to
develop and are currently in use. The search for better and efficient methods for making the
credit decisions continues in the pursuit of perfection. The dream of all lending institutions for a
completely reliable, efficient automated procedure for evaluation of prospective credit
applicants continues to remain a dream.
Tamisin (1991) has created a model of the loan negotiation process based on case-based
reasoning process. The prototype system designed by the author makes use of previous
experiences to guide the problem solving process. The author gave an introduction to the loan
negotiation process by first defining clearly the meaning of negotiation and then described the
phases or the life cycle of the negotiation process as divided into three stages. Loan
negotiation consists of three phases (Marsh, 1984) which are shown in the figure below:
PREPARATION and SUBMISSION
CREDIT CONDITION NEGOTIATION
ANALYSIS and EVALUATION
Figure 2.1 Loan Negotiation Phases

Preparation and Submission Phase: This phase involves the preparation of the applicant's
application using the project proposal, financial report and loan requirements.
Analysis and Evaluation phase: This phase takes the output of the last phase, the financial
application and uses it as the input. It gives the overall loan application's status as output. This
phase involves the analysis and evaluation of the financial report, project proposal and the
loan requirements. After this, the credit history of the applicant is thoroughly investigated and
the overall application status is provided. The status signifies the decision for approval,
rejection or negotiation of the loan.
Credit Condition Negotiation Phase: If the applicant is a corporate customer, a negotiation is
generally required. This phase is invoked when a talk between the two parties is called for.
The inputs to this phase are the analysis status and loan requirements. During the negotiation
3
process, one negotiating agent proposes a set of conditions. If the second agent agrees on
this set of conditions, then an agreement is reached. Otherwise, the second agent proposes
its own set of conditions. This process is iterated until an agreement or a deadlock situation is
reached.
Bryant (1962) gives the entire working of the mortgage department of a lending institution. All
the terms and paraphernalia related to the Credit Division are outlined brilliantly. However, the
drawback with the book was that of it’s outdation. As it was written in the early 1960’s, most of
the policies and workings of the credit department have undergone appreciable changes and
modifications, as of today. The second drawback of the book is that, it covers only Home Bank
Loans, and strictly remains in that domain.
Zinkhan (1990) earlier study had indicated the principles of credit require only 'Five Cs' i.e.
capacity, capital, character, collateral, and conditions - in relationship to the evaluation of a
given firm's credit risk. This paper suggests the 'Sixth C' of Credit and calls it the 'Customer
Profitability Analysis'. A number of efforts have been undertaken to quantify and summarize
two of these five C's for the purpose of estimation of the bankruptcy of firms: capacity and
capital. Notable among these efforts are the Z-score model of Altman (1968), the Zeta
analysis model of Altman et al. (1977), and the application of a discriminant analysis model to
small businesses by Edminister (1972) . In addition, a credit-scoring model was developed by

Chesser (1974) to determine the creditworthiness of a potential business loan customer.
According to the author, the above mentioned quantitative models ignore the character,
collateral, and conditions dimensions of credit risk analysis . The purpose of this technique is
to jointly evaluate the objective and subjective estimations of the six C's in order to generate
an overall indicator of the relative attractiveness of a given potential business loan. A
hierarchical model of the business loan evaluation process was also suggested as shown in
Figure 2.2.
Acharya (1990) has designed and tested a Loan Negotiation System based on a rule based
system. “Negotiation is a dynamic process of adjustment by which two or more parties, each
with his own objectives, confer together to reach a mutually satisfying agreement on a matter
of common interest. The process of negotiation may end with or without a consensus”.
According to the author, Negotiation is oriented toward the future, its progress is defined by
the negotiating agents. It is also oriented with the sole goal that each party tries to obtain
maximum payoff.
Overall Indicator of Loan Attractiveness
Indicator of Expected
Customer Profitability
Indicator of Credit Risk
Indicator of
Expected
Profitability
on Current
Transaction
Indicator of
Expected
Profitability
on Future
Transaction
Indicator of
Capacity and

Capital Risk
Indicator of
Character Risk
Indicator of
Collateral Risk
Indicator of
Conditions Risk
4
Figure 2.2 Hierarchical Model of the Loan Evaluation process
The five phases through which every negotiation proceeds are: exploration, bidding,
bargaining, settling, and ratifying. Usually, they may not follow sequentially and the negotiators
may alter the sequence of the phases.
The negotiators may follow the sequence on one aspect of the deal and then start all over
again on a second aspect.
Duchessi et al. (1988) describe a knowledge-Engineered system for Commercial Loan
Decisions. The paper initially describes the Commercial Loan Analysis process which covers a
general and a detailed analysis. This is followed by a complete description of the expert
system 'Commercial Loan Analysis Support System (CLASS)'. This expert system is
described as one which is designed to evaluate a company's financial posture, recommend
commercial loan decisions and pertinent covenants, and document the loan analysis.
Commercial Loan Analysis process as described by the authors examines numerous financial
factors to uncover a company's financial weaknesses. Their evaluation begins with a general
analysis involving an examination of key financial trends and factors. When one of those
factors does not meet the industry norm, commercial loan officers must perform a more
detailed analysis to uncover the causes.
General Analysis: It consists of trend analysis, and separate analyses of credit, collateral,
capital and capacity. General trend analysis provides loan officers with a quick indication of a
company's performance in several key areas: sales, operating income, net income, selling and
administrative expenses, working capital, and cash flow. Credit analysis measures a
company's ability to repay its short and long-term obligations. The credit analysis really

consists of efficiency, profitability, and liquidity analyses. Inventory turnover, receivables
turnover, fixed asset turnover, and total asset turnover are the primary efficiency measures.
Profitability analysis considers operating margin, profit margin, return on assets, and return on
equity, while liquidity analysis includes the current and quick ratios. Collateral analysis
examines the relationship between the value of all assets and pledged assets. To estimate
their value, loan officers appraise book value, age, and condition of assets. Capital analysis
provides an indication of a company's leverage position. Long term debt to total assets, total
debt to total assets, interest coverage, and fixed charge coverage are the primary factors. If all
four measures are above industry standards, a company's capital position is strong. If they
are all below, capital is poor. Capacity analysis measures the degree to which a loan can be
supported by a company, using the same ratios as capital analysis.
Detailed Analysis: This is performed whenever a general analysis indicates a weakness. As
problem areas are identified and examined, the loan officers accumulate loan covenants, or
restrictions, which become part of the final loan agreement.
Gilliam (1990) describes the financial analysis procedure for the estimation of lending risk
which can be calculated using tangible factors or ratios . The terms and ratios described
correspond to the derivable ratio's that are related to the 'Five Cs' of credit. The author states
that the financial analysis can provide valuable information about why a company needs to
borrow, what the loan will fund, whether operations will be able to generate sufficient cash flow
to repay debt, and whether company assets will be available as collateral. If the borrower has
met the initial criteria for a desirable customer through interviews and credit investigation, the
5
next place to begin risk analysis is with the company's financial statements. The degree of
reliance on financial analysis is directly related to how they are prepared. In defining lending
risk, there are three general areas of investigation relating to financial statements.
Economic Condition, which measures the company's leverage, liquidity, and activity positions;
Profitability, which addresses break-even analysis and other trends in company sales and
expenses; and
Cash Flow, which assists in determining sources and uses of cash to pinpoint why a company
is borrowing and whether the loan can be repaid.

Lenders have access to several tools to analyze these areas. Trend analysis focuses on a
company's direction regarding assets, liabilities, revenues, and expenses. Trend analysis
focuses on a company's direction regarding assets, liabilities, revenues, and expenses.
Comparing management's plan to actual performance highlights management's ability to
forecast and plan for future events.
The lender's primary purpose is to make loans that can be repaid while minimizing the bank's
exposure to loans with poor credit quality. To be successful, it is important for lenders to
assess areas of relative strength and weakness by reviewing the economic condition,
profitability, and cash flow of a company using tools such as trend analysis and industry
comparison.
Harris (1994) gives the fundamentals of trading describing it as a zero-sum game when
measured relative to underlying fundamental values. This paper classifies traders into three
categories winning, utilitarian and futile . This paper then throws light on the origins of trading
profits , how the contributions of the three categories affect the price efficiency and market
liquidity.
2.2. Existent Systems
Holsapple et al. (1988) throw light on the adaptation of expert system technology to financial
management. The paper describes the rudimentary architecture of expert systems,
recognition of the potential of expert systems by the financial markets and their acceptability
and impact. The conclusion made by the paper is that current technology is inadequate for
applications requiring insight, creativity, and intuition. According to the author, an expert
system is a software system that imitates the reasoning results of human experts in a well
defined domain. It aims to generate advice about problems in the domain comparable to the
advice that a human expert would deduce for those same problems.
Table 2.1 Expert Systems for Financial Applications of ' Commercial Loan Analysis'
Expert System Company Function
Authorizer's Assistant American Express Credit Authorization
Financial Analyzer Athena Group Commercial Loan Approval
Lending Advisor Syntelligence Credit Analysis
Mortgage Loan Analyzer Arthur Anderson Mortgage Loans Evaluation

Underwriting Advisor Syntelligence Commercial Underwriting
6
Marble (1988) ( Managing and Recommending Business Loan Evaluation ) : is an expert
system with inductive learning to evaluate Business Loans. The paper first describes the loan
evaluation process and the basic difficulties that confront a loan decision. The paper then
describes the design and construction of Marble.
Countrywide Loan-Underwriting Expert System (CLUES) (1991): The paper gives the life cycle
of a loan and then describes the planning, construction and working principles of CLUES. An
important aspect covered in this paper is the enumeration of the reasons as to why the rule
based methodology was used in CLUES. The drawbacks and advantages of the various AI
technologies were also touched upon.
Mortgage Risk Evaluator (MRE) (1962): Nestor, one of the earliest neural network companies,
has a product that appraises mortgage applications. The system was trained on several
thousand actual applications, about half of which were accepted and the other half of which
were rejected by the human underwriters. Learning from the successes and failures of this
body of experience , the system looks for patterns in the data to determine what constitutes a
bad risk. AVCO Financial Services , in Irvine, California, uses this neural network system for
Credit Risk Analysis. The system was trained on more than 10,000 credit case histories. A
New York Times edition reported that one test indicated there would have been a 27 percent
increase in profits if the neural network system had been used instead of the computerized
evaluation system previously used by AVCO. Besides credit risk analysis , other commercial
applications of neural networks in Finance include Identification of forgeries, interpreting
handwritten forms, rating investments and analyzing portfolios .
Neuroforecaster is an advanced windows based, user-friendly, intelligent, neural network
forecasting tool. Made by Accel Infotech (S) Pte Ltd., Singapore. It is packed with the latest
technologies including neural network, fuzzy computing and non-linear dynamics. Besides
performing the Loan Analysis,. it can also be used for other financial applications some of
which are enumerated below.
Stock Price Prediction, Stock market six monthly return forecast, Stock selection based on
company ratios, Stock market index forecast, National GDP forecast, Sales Forecast, US$ to

Deutshmark exchange rate forecast, Fraud Detection and Fault Diagnosis, Air passenger
arrival forecast, Credit rating of bank loan applications, Property price valuation, Bond Rating,
Construction demand forecast, XOR-A classical problem.
2.3. Relative Comparison between Methodologies
Digiammarino et al. (1991) elucidate on the gradual evolution of Artificial Intelligence
technology in the automation of loan underwriting. The paper claims that several applications
of information technology are now experiencing rapid growth as they penetrate the middle-size
and smaller-size ranks of the industry. The author first describes the ScoreCard, a statistical
technique for credit evaluation, followed by the comparison between the Statistical, AI and
Neural Network methodologies for Loan underwriting.
As per the author, the Scorecard is the most familiar kind of decision support system in
consumer credit and dates back to the 1960's or earlier. Derived from statistical analyses that
correlate application and credit report data to each lender's actual loan losses, these models
accurately measure the probability of each applicant defaulting. The score is calculated by the
application processing system and used along with collateral and credit policies to recommend
7
a decision to the reviewing credit officer. The degree to which these recommendations are
followed varies widely from one organization to another. Innovative financial institutions'
current involvement in advanced research in decision support systems is currently moving in
two dimension: profit-driven objectives and more sophisticated modeling techniques, including
artificial intelligence. The issue of profitability is so important, that leading organizations are
beginning to look for ways to address it head-on in decision systems. In addition to sharpening
the objective of decision support models by incorporating profitability, innovators are applying
more advanced techniques to build the models themselves.
The author claims that classical statistical techniques are capable of producing more powerful
models than are commonly used today, but the improvement is only marginal compared to the
difficulty of implementing more complex equations. As a result, a great deal of attention is
being devoted to the use of artificial intelligence in consumer credit.
Neural networks are potentially very valuable tools for the credit industry in situations where
these factors occur together: large amounts of data, complex interactions, quick feedback on

the results of each decision, and a lack of human constraints. Expert systems take the
opposite approach, eschewing hard data in favor of the judgment of human experts. A
weakness is that expert systems do not make use of hard data even when such information
could enhance human judgment. In these instances ( for example, credit application screening
) , the expert system can be modified to include a risk measure as part of the final
recommendation. The key difference between a classical statistical model and a neural net is
that the net typically analyzes a larger number of mathematical forms for the relationships
between predictive variables. This procedure can lead to better predictive ability when the
predictors work in complex but stable and well-defined ways.
Eyden (1994) This paper compares the use of artificial neural networks (ANNs) and multiple
discriminate analysis (MDA) in the prediction of credit risk. It clearly brings out the comparison
between statistical modeling versus neural networks in financial decision making. MDA is
traditionally regarded as the most applicable statistical method for the prediction of credit risk
and is widely used by financial institutions and other organizations. MDA is based on a linear
equation and has the limitation of not being appropriate for non-linear data. ANN's on the other
hand, incorporate both linear and nonlinear components and therefore, according to the paper,
prove that they are more suitable for different data types. The paper presents a case study of
financial application (credit scoring problem) to measure the comparative performances of the
two models. The findings of the paper indicate that ANN's outperform MDA in the forecasting
of credit risk.
Costantino (1991) gives out a very useful eight-point procedure for determining whether expert
systems and neural networks are in fact superior to current credit-scoring processes. The
paper says that, although the framework of this procedures is loosely defined, it addresses the
full range of issues that confront creditors in the adoption of new technologies. The paper
draws a clear cut method of evaluating the superiority or inferiority of the AI technology over
the statistical methods for credit risk analysis. According to the author, the evaluation of new
technologies requires an information base that considers basic business components-
economic returns, people, systems, and control. The impact on these basis entities can be
best understood by addressing the eight issues as given below:
The cost of new technologies vs. the old: This calculation depends on the number of credit-

scoring systems requiring redevelopment and the configuration of each of the new
technologies.
8
Thorough understanding of the current process: Answers to questions like "Which applicants
are rejected/ accepted at high rates? " ; " Which applicant profiles are accepted as often as
they are rejected? "; for each credit-scoring system in use helps to understand the nature of
the incoming populations and how the current process selects new accounts.
Relationship of the approval rate to the bad rate: This relationship can be developed by
analyzing a sufficiently large random sample of good performers, bad performers, and rejects.
Knowing the relationship between approval rates and bad rates helps to determine the size
and potential risk associated with the applicants most likely to be affected by the new
technology.
Long term cost-return tradeoffs: To evaluate this, a summary table is drawn from the last
three evaluation steps and ratios are calculated.
Service Level maintenance or enhancement: Whether the new technology will cause decision
turnaround times to increase to unacceptable levels or improve. This can be tested during the
test-benchmark step. Real-time environments can significantly degrade the performance of
technologies that were designed as batch systems.
The Readiness of organization to manage the development and implementation of the new
technology: This step is the most qualitative one to evaluate.
The firm's data processing capacity: This can be found by the answers to questions like "Are
there enough programmers available to code and test the required programs ?"; "Will new
hardware have to be acquired in order to implement the new technology ?" ; " Will new
hardware and interfaces need to be acquired ?"
The technology's auditability and security capability: Auditability comes in two flavors-applicant
and technology. Auditability is key to maintaining system controls. Security is the key to
maintaining competitive advantage and to preventing fraudulent transactions.
9
Chapter 3
Theoretical consideration

3.1. Credit analysis: what make a good loan?
Credit Department must satisfactorily answer three major questions regarding each loan
application:
1. Is the borrower creditworthy? How do you know?
2. Can the loan agreement be properly structured and documented so that the bank and its
depositors are adequately protected and the customer has a high probability of being able
to service the loan without excessive strain?
3. Can the bank perfect its claim against the assets or earnings of the customer so that, in
the event of default, bank funds can be recovered rapidly, with low cost and low risk?
3.1.1. Is the borrower creditworthy?
The question that must be dealt with before any other is whether or not the customer can
service the loan - that is, pay out the credit when due, with a comfortable margin for error. This
usually involves a detailed study of six aspects of loan application - character, capacity, cash,
collateral, conditions, and control. All must be satisfactory for the loan to be a good one from
the lender’s point of view.
• Character. The loan officer must be convinced that the customer has a well-defined
purpose for requesting bank credit and a serious intention to repay. Once the purpose is
known, the loan officer must determine if it is consistent with the bank’s current loan policy.
Even with a good purpose, however, the loan officer must determine that the borrower has
a responsible attitude towards using borrowed funds, is truthful in answering the bank’s
questions, and will make every effort to repay what is owed. Responsibility, truthfulness,
serious purpose, and serious intention to repay all moneys owed make up what a loan
officer calls character.
• Capacity. The loan officer must be sure that the customer requesting credit has the
authority to request a loan and the legal standing to sign a binding loan agreement. This
customer characteristic is known as the capacity to borrow money.
• Cash. Does the borrower have the ability to generate enough cash, in the form of income
or cash flow, to repay the loan? In general, borrowing customers have only three sources
to draw upon to repay their loans: (a) cash flows, (b) the sale or liquidation of assets, or (c)
funds raised by issuing debt or equity securities. However, bankers have a strong

preference for cash flow as the principal source of loan repayment because asset sales
can weaken a borrowing customer’s balance sheet, while additional borrowing by a loan
customer can make the bank’s position as creditor less secure. Moreover, shortfalls in
cash flow are common indicators of failing businesses and troubled loan relationships. The
loan officer’s evaluation of a borrower’s cash involves asking and answering such
10
questions as: Is there a history of steady growth in earning or sales? Is there a high
probability that such growth will continue to support the loan?
• Collateral. Does the borrower posses adequate net worth or own enough quality assets to
provide adequate support for the loan? The loan officer is particularly sensitive to such
features as the age, condition, and degree of specialization of the borrower’s assets.
Technology plays an important role here as well. If the borrower’s assets are
technologically obsolete, they will have limited value as collateral because of the difficulty
of converting them into cash if the borrower’s income falters.
• Conditions. The loan officer and credit analyst must be aware of recent trends in the
borrower’s line of work or industry and how changing economic conditions might affect the
loan. A loan can look very good on paper, only to have its value eroded by declining sales
or income in a recession or by the high interest rates occasioned by inflation. To assess
industry and economic conditions, most banks maintain files of information - newspaper
clippings, magazine articles, and research reports - on industries represented by their
major borrowing customers.
• Control. Whether changes in law and regulation could adversely affect the borrower and
whether the loan request meets the bank’s and the regulatory authorities’ standards for
loan quality.
3.1.2. Can the loan Agreement be properly structured and documented?
The loan officer is responsible to both the customer and the bank’s depositors and
stockholders and must seek to satisfy the demand of all. This requires, first of all, the drafting
of a loan agreement that meets the borrower’s need for funds with a comfortable repayment
schedule. The borrower must be able to comfortably handle any required loan payments,
because the bank’s success depends fundamentally on the success of its customers. If a

major borrower gets into trouble because it is unable to service a loan, the bank may find itself
in serious trouble as well.
A properly structured loan agreement must also protect the bank and those it represents -
principally its depositors and stockholders - by imposing certain restrictions on the borrower’s
activities when these threaten the covery of bank funds. The process of recovering the bank’s
funds - when and where the bank can take action to get its fund returned - also must be
carefully spelled out in the loan agreement.
3.1.3. Can the bank perfect its claim against the borrower’s collateral?
The collateral pledged behind a loan and the other assets that a borrower may own are the
second line of defense against loan default, after the borrower’s cash flow. When the
borrower’s cash flow or income falters, the lender must look to the borrower’s assets.
Therefore, the key issues for any lender include whether or not the bank can get clear title to
any assets that are available to backstop the loan, which creditors have priority of claim if a
borrower’s assets must be liquidated to cover a loan, and whether the borrower has assigned
the bank exclusive interest in certain assets or has pledged those assets to someone else. An
important technical issue here - crucial in mortgage lending - is whether deeds to property
have been properly filed with local governmental authorities so that the bank knows for sure
who currently has title to the property. If a home owner is borrowing money and using his or
11
her home as collateral, the loan officer must verify not only that the customer has title to the
home but also whether other lenders have legitimate claims against that property.
3.2. Financial ratio analysis of a customer’s financial statements
The calculation of financial ratios provides the basis of most technical, quantitative credit
analysis. Information from balance sheets and income statements is typically supplemented by
financial ratio analysis. By careful selection of items from a borrower’s balance sheets and
income statements, the loan officer can shed light on such critical areas in business lending as
1. A borrowing customer’s ability to control expenses;
2. A borrower’s operating efficiency in utilizing resources to generate sales and cash flow;
3. The marketability of the borrower’s product line;
4. The coverage that earnings provide over a business firm’s financing cost;

5. The borrower’s liquidity position, indicating the availability of ready cash;
6. The borrower’s track record of profitability or net income;
7. The amount of financial leverage a business borrower has taken on; and
8. Whether a borrower faces significant contingent liabilities that may give rise to substantial
claims in the future.
3.2.1. The business customer’s control over expenses
How carefully a business firm monitors and controls its expenses is a barometer of the quality
of its management and how well its earnings are likely to be protected. Selected financial
ratios usually computed by loan analysts to monitor a firm’s expense control program include
the following:
Wages and salaries / net sales
Overhead expenses / net sales
Depreciation expenses / net sales
Interest expense on borrowed funds / net sales
Cost of goods sold / net sales
Selling, administrative, and other expenses / net sales
Taxes / net sales
3.2.2. Operating efficiency: measure of a business firm’s performance
How effectively are assets being utilized to generate sales and cash flow for the firm and how
efficiently are sales converted into cash? Important financial ratios here are
Annual cost of goods sold / average inventory (or inventory turnover ratio)
Net sales / total assets
Net sales / net fixed assets
Net sales / accounts and notes receivable
average collection period = Accounts receivable / Annual credit sales / 360
12
In general, the higher a firm’s inventory ratio, the better it is for banks and other creditors,
because this ratio shows the number of times during a year that the firm turns over its
investment in inventories by converting those inventories into goods sold. When the inventory
turnover ratio is too low, it may indicate poor customer acceptance of the firm’s products or

ineffective production control and inventory control policies. Too high an inventory turnover
ratio could reflect underpricing of the firm’s product or inadequate stocks of goods available for
sale, with frequent stockouts, which drives customers away.
The ratio measuring turnover of fixed assets indicates how rapidly sales revenues are being
generated as a result of using up the firm’s plant and equipment to produce goods or services.
If the fixed-asset turnover ratio falls, this may indicate the firm has invested too heavily in plant
and equipment, given the strength of current market demand for its product, and thus has
substantial productive capacity that isn’t being used. Alternatively, a fixed-asset ratio that is
too high would lead the analyst to believe the firm has not devoted enough of its resources to
increasing or upgrading its physical plant in order to achieve greater efficiency and
productivity.
The collection period ratio reflects the firm’s effectiveness in collecting cash from its credit
sales and provides evidence on the overall quality of the firm’s credit accounts. A lengthening
of the average collection period suggests a rise in past-due credit accounts and poor
collection policies.
3.2.3. Marketability of the customer’s product, service, or skill
In order to generate adequate cash flow to repay a loan, the business customer must be able
to market goods, services, or skills successfully. A bank can often assess public acceptance
of what the business customer has to sell by analyzing such factors as the growth rate of
sales revenues, changes in the business customer’s share of the available market, and the
gross profit margin (GPM), defined as
GPM = (Net sales - cost of goods sold) / Net sales
A closely related and somewhat more refined ratio is the net profit margin (NPM):
NPM = Net income after taxes / Net sales
The GPM measures both market conditions - that is, demand for the business customer’s
product or service and how competitive a marketplace the customer faces - and the strength
of the business customer in its own market, as indicated by how much the market price of the
firm’s product exceeds the customer’s unit cost of production and delivery.
The NPM, on the other hand, indicates how much of the business customer’s profit from each
unit of sales survives after all expenses (including taxes) are deducted, reflecting both the

effectiveness of the firm’s expense-control policies and the competitiveness of its pricing
policies.
3.2.4. Coverage ratio
Coverage refers to the protection afforded creditors of a firm based on the amount of the firm’s
earnings. The best-known coverage ratios include the following:
13
Interest coverage = Income before interest and taxes / Interest payments
Coverage of interest and principal payments = Income before interest and taxes /
(Interest payments + principal repayments/(1 - Firm’s marginal tax rate))
Coverage of all fixed payments = (Income before interest and taxes + Lease
payments) / (Interest payments + Lease payments)
The interest coverage ratio indicates the margin of safety that earnings provide creditors in
relation to interest charges.
The coverage of all fixed payment simply extends the interest coverage ratio to account for
contractual commitments under leasing agreements.
3.2.5. Liquidity indicators for business customers
The borrower’s liquidity position reflects his or her ability to raise the cash in timely fashion at
reasonable cost, including the ability to meet loan payments when they come due. Popular
measures of liquidity include the following:
Current ratio = Current assets / Current liabilities
Acid-test liquidity ratio = (Current assets - Inventories) / Current liabilities
Net liquid assets = Current assets - Inventories of raw materials or goods -
Current liabilities
Net working capital = Current assets - current liabilities
An individual or institution is liquid if it can convert assets into cash or borrow immediately
spendable funds precisely when cash is needed. Liquidity is, therefore, a short-run concept in
which time plays a key role. For that reason, most measures of liquidity focus on the amount
of current assets (cash, marketable securities, accounts receivable, inventory, prepaid
expenses, and any other assets that normally roll over into cash within a year’s time) and
current liabilities (accounts payable, notes payable, taxes payable, and other short-term claims

against the firm, including any interest and principal payments owed on long-term debt that
must be paid during the current year).
The current ratio indicates the extent to which the claims of short-term creditors are covered
by assets that can be readily converted into cash without loss. High current ratios suggest a
high margin of safety for short-term creditors. However, the ratio does not consider differences
in the quality of receivables and inventories.
Concern over the quality of liquidity of inventories is purged in the quick ratio. Only the “quick”
assets of cash, marketable securities, and receivables are included. For many industries in
which inventory values may be suspect, the quick ratio is a more reliable measure of liquidity
than the current ratio.
3.2.6. Profitability indicators
The ultimate standard of performance in a market-oriented economy is how much net income
remains for the owners of a business firm after all expenses (except stockholder dividends)
are charged against revenue. Most loan officers will look at both pre-tax net income and after-
tax net income to measure the overall financial success or failure of a prospective borrower
14
relative to comparable firms in the same industry. Popular bottom-line indicators of the
financial success of business borrowers include the following:
Before-tax net income: Total assets, net worth, or total sales.
After-tax net income: Total assets, net worth, or total sales.
Return on equity = Net income available to common stock / common stock equity
Return on assets = Net income after tax /Average total assets
Profit margin = Net income after tax / Net sales
Return on equity is a summary measure of how effectively common stockholders’ funds have
been employed, including the effectiveness of the use of financial leverage.
Return on assets indicates the efficiency with which management employed the total capital
resources available to it. It is a better measure of operating performance than return on equity
because the latter is effected by the degree of financial leverage.
Profit margin measures the profit per currency unit of net sales. Its complement (1 - profit
margin) indicates the expense incurred to generate one currency unit of revenue and reveals

the effectiveness of cost controls and pricing policies.
3.2.7. The financial leverage factor as a barometer of capital structure
The term financial leverage refers to the use of debt in the hope that the borrower can
generate earnings that exceed the cost of debt, thereby increasing the potential return to a
business firm’s owners. Key financial ratios used to analyze any borrowing business’s credit
standing and use of financial leverage are as follows:
Leverage ratio = Total liabilities / Total assets
Capitalization ratio = Long-term debt / Total long-term liabilities and net worth
Debt-to-sales ratio = Total liabilities / Net sales
The greater the amount of indebtedness a borrowing customer has already taken on, other
factors held equal, the less well secured is any particular lender’s position. The higher the
leverage ratio becomes, the less likely it is that additional loans will be granted to a customer
until he or she pays down some of the outstanding indebtedness. Moreover, if a loan is
granted to a highly leveraged borrower, it is likely to carry a higher interest rate plus a
requirement that more collateral be pledged.
The capitalization ratio focuses upon the business customer’s use of permanent financing,
essentially comparing the degree to which the firm is supported by long-term creditors as
opposed to its owners’ equity capital (net worth). Business debt can also be linked to business
sales, because those sales ultimately provide the funds needed to retire the debt. If a firm’s
liabilities increase relative to its sales, management will have to compensate for the heavier
debt burden by either finding less expensive sources of credit or lowering expenses so that
more sales revenue reaches the firm’s bottom line of net income.
3.2.8. Contingent liabilities
Usually not shown on customer balance sheets are other potential claims against the
borrower that the loan officer must be aware of, such as
15
1. Guaranties and warranties behind the business firm’s products.
2. Litigation or pending lawsuits against the firm.
3. Unfunded pension liabilities the firm will likely owe to its employees in the future.
4. Taxes owed but unpaid.

5. Limiting regulations.
These so-called contingent liabilities can turn into actual claims against the firm’s assets at a
future date, reducing available funds to repay the loan. Usually the loan officer’s best move in
this circumstance is, first, to ask the customer about pending or potential claims against the
firm and then to follow up with his or her own investigation, checking courthouse records,
public notices, and newspapers. In this instance, it is far better to be safe and well informed
than to repose in blissful ignorance.
3.3. Commercial Loan Decision Making Process
In practice, the evaluation of a loan application is based on the information presented in
financial statement plus any qualitative information, such as the quality of management, the
ability to repay the loan, and the availability and value of collateral. Frequently the qualitative
information is of greater value in the lending decision than the financial statement analysis.
Exhibit 1 presents the decision-making process for evaluating commercial loans. Exhibit 1
represents a generic overview of the lending process and was an underlying framework in the
designing of the Expert System. The evaluation of a firm’s credit worthiness is a score that
weighs each of characteristics presented in Exhibit 1, Block . When the credit risk score is
calculated, the risk classification of the applicant is established by comparing it to an
objectively determined standard.
If the loan is approved, the bank establishes the terms of the loan with the customer in order
to assure repayment. The final phase of the process involves organizing all the data and
information used in the decision process and storing it in the loan documentation file. This file
is the basis for future performance reviews.
3.4. Architecture Of Expert Systems
Keeney (1988) suggested a concept called value-driven Expert System concept which is
implements for multiattribute utility decision making in developing countries. However, this
concept requires strong background of utility function.
On the other hand, Olave et al. (1988) focused on the side of Expert System that can solve
problems based on symbolic and qualitative data which often requires human thinking.
Furthermore, Hopsapple et al (1988) showed that Expert System can generate advice on the
same problem or similar problems which human experts have deduced before.

The Expert System was stated by Ignizio (1990), shows the role of model rather than reporting
medium (i.e. computer). Furthermore, Ignizio in the same paper emphasized an algorithmic
method.
16
According to Ignizio, an Expert System can be defined as a model and associated procedures,
that exhibits the degree of expertise in problem solving, within a specific domain.
 Yes
No
17
START
IS
THE FIRM A PRESENT
CUSTOMER OF THE
BANK?
IS AN
EXTENSIVE CREDIT
CHECK ON THE FIRM
REQUIRED?
EVALUATE THE
POTENTIAL OF A NEW
CUSTOMER
RELATIONSHIP
INVESTIGATE THE CREDIT-
WORTHINESS OF THE
PROPOSED LOAN BY
ANALYZING THE FIRM’S...
- Quality of Financial
information
- Economic Characteristics
(Size, Market Share,

Diversification)
- Competitive Position in
Industry
- Financial Characteristics
(Profitability, Liquidity,
Leverage, Growth)
- Management (Quality,
Experience, Depth)
- Availability of Funds
(Equity or Debt Markets)
- Ability to Repay Loan
(Cash Flow Analysis,
Security)
- Supplier Experience
- Experience at Previous
Bank
- Value of Collateral
... DETERMINE A CREDIT
RISK SCORE
DETERMINE THE CREDIT
NEEDS OF THE CUSTOMER
(PURPOSE OF LOAN,
AMOUNT, MATURITY)
DETERMINE THE CREDIT
NEEDS OF THE CUSTOMER
(PURPOSE OF LOAN,
AMOUNT, MATURITY)
CHECK LOAN
APPLICATION AGAINST
LEGAL AND POLICY

RESTRICTIONS



No
Yes
On the other hand, Expert System can solve problem based on symbolic and qualitative data
which often required human thinking.
In an Expert System generally the knowledge is obtained from an authority in a specific,
narrow field of activity. This field is called domain and the authority is called the domain expert.
The program-developer or knowledge engineer interviews the domain expert and enters the
factual, judgmental and procedural knowledge into the Expert System program as a
knowledge base.
The conventional architecture of an expert system is shown in Figure 3.2.
Figure 3.2. Generic Architecture of Conventional Expert System
It consists of :
18
SHOULD LOAN
BE RECOMMENDED?
LOAN MONITORING PROCESS
• Timing of Payments
• Value of Collateral
• Compliance with Covenants
• Periodic Financial Reports
LOAN COMMITTEE
APPROVAL
COMPLETED
DOCUMENTATION FILE
RECORD ANALYSES AND
RECOMMENDATIONS

No
Knowledge
System:
- Rule sets
- Variables
UserUser InterfaceInference Engine



Figure 3.1. Commercial Loan Decision Making Process
RECOMMENDATIONS
CONCERNING TERMS OF
THE LOAN (Type of
Financing Amount, Interest
Rate, Collateral, covenants,
Repayment)


END
• A knowledge system,
• An inference engine, and
• A user interface.
The latter two are software portions of an expert system and form its problem processing
system, which can accept requests from users and then process them by drawing on available
knowledge system contents.
The knowledge system stores application-specific reasoning knowledge about a particular
domain. Each piece of reasoning knowledge specifies what conclusion is valid when a
particular situation exists. Such a fragment of knowledge is commonly represented as some
variant of a production rule. A set of rules pertaining to a defined problem area is called a rule
set and the collection of rule sets available for dealing with all problems in the expert system’s

domain is sometimes called “rule-based systems”.
In its most rudimentary form, a rule is composed of 2 parts:
• The premise consists of one or more conditions, and
• The conclusion is composed of a series of one or more actions that are to be taken as
valid if the premise is satisfied.
C
1
C
2
... C
n
A
1
A
2
... A
m
n,m ≥ 1
C
i
is essentially a predicate, whose value at any moment is either true, false, or unknown.
Here, we assume that these conditions are conjunctively related, although disjunctions are
also permissible in a rule’s premise.
A
j
can also be denoted as a predicate. When the premise is true, the truth of A
j
is likewise
asserted. That is, it represents an action that can be legitimately performed.
There are many diverse, yet equivalent, syntactic conventions for formally specifying a rule.

Each rule in a rule set encapsulates some fragment of reasoning knowledge that can, in
principle, be developed and modified independently of other rules.
An important feature of many expert systems is their ability to cope with uncertain situations or
inexact reasoning. One way to support dealing with uncertainty involves stating a confidence
factor (CF) for each rule, depicting the degree of belief or weight an expert ascribes to a rule.
A confidence factor may be linearly related to the degree of certainty asserted by the experts
on a rule. A low value, say 0.2, is interpreted as a less valid rule while a high value, say 0.8, is
viewed as a more valid one. Much more flexible and sophisticated CF means for dealing with
reasoning under uncertainty exist.
Rules, such as those above which form the core of a knowledge system, are acquired by a
knowledge engineer. Knowledge engineers use techniques such as interviewing, videotapes,
protocol analysis, and personal observations to understand and capture an expert’s reasoning
knowledge. The rules elicited from one expert may be very different from those acquired from
another, which reflects the phenomenon of differences among experts. Care must be
exercised in constructing a rule set.
19
How well an Expert System performs depends on the extent to which a knowledge engineer
succeeds in formalizing an expert’s reasoning knowledge in a rule set.
In addition to a rule set, an Expert System must be able to keep track of the current state of
the world. For instance, the Expert System may need to know the most recent budget deficit
to use it as a basis for rules determining whether the impending deficit can be inferred to be
large. Thus, the knowledge system would need a state variable for the budget deficit figure. An
Expert System’s state variables are sometimes characterized as being attribute/value pairs.
The individual predicates of the previous rule is example of attribute/value pairs. State
variables can be organized to form frames. A frame consists of a number of attributes that
describe a concept or object. The value of each attribute can either be specified, determined
by default, or inherited from other frames.
A more natural way of handling state descriptions for business applications is to use such
common business computing techniques as database management. The database concept is
essentially a prefabricated frame.

The inference engine is the control mechanism of an Expert System. It is activated when the
user initiates a consultation session with the Expert System by requesting advice about a
specific problem. The problem statement identifies a goal to be sought and may specify initial
values for some state variables. A goal is a variable whose value is to be deduced by the
inference engine. Moreover, some of the other state variables may also have unknown values.
Beginning with known state variable values (if any), the inference engine applies rules that
allow it to infer values for unknown variables. It may also prompt the user to establish values
for unknown variables. The inference process moves through a series of states with more and
more variables’ values becoming known until the goal state is reached (i.e., the goal variable’s
value becomes known). The inferred result is then reported to the user as advice.
There are many dimensions along which inference engines can be characterized and
differentiated. Many of these are concerned with how an engine applies rules in making the
unknown known. For instance, inference engines can differ in terms of:
• whether rules are processed in “forward” or “backward” fashion;
• The order for selecting candidate rules
• The way in which confidence factors are combined; and
• The language of implementation
Forward and backward inference refer to the way in which the engine examines a rule
premise first or conclusion first. In the latter case, an inference engine looks at rule
conclusions to identify those rules that have actions affecting the goal variable (i.e., that could
give it a known value). The premise of such a candidate rule can then be examined. If current
state variable values make the premise true, the rule is “fired” by taking the actions stated in
its conclusion. However, if those variables currently have unknown values, then they become
subgoals as the basis for further backward inference.
In forward reasoning, an inference engine examines each rule’s premise with respect to the
current state variable values. If such values make its premise true, the rule is fired causing
20
variable values to change as directed by the conclusion’s actions. The inference engine
continues this non-goal-directed reasoning until either the goal variable value is established or
no more rules can be fired. The goal is reached in the first case, but fails to be attained

otherwise.
For both forward and backward inference, as well as hybrids thereof, several rules can be
candidates for immediate processing. The approach that an inference engine uses to choose
the order for processing candidate rules is called a selection strategy. It can influence how
rapid the consultation will be and possibly what its results will be.
Inference engines can also differ in terms of their approaches to combining confidence factors
as rules are fired. The alternative approaches are often called certainty algebras. They range
from conservative to venturesome methods of propagating certainties about rules and
variables’ values along to the result, allowing the expert system to express a degree of
confidence in the advice it deduces. Some inference engines support no certainty algebras.
Others support only one. Sometimes a programmer must be employed to code an algebra. In
other cases, an inference engine has built-in switches for easy selection of a desired algebra.
It is important to understand that there is a major distinction between the language used to
state rules and the language used to implement an inference engine that processes those
rules. Modern tools for developing expert systems provide a high-level language for
developers to formally state human expertise in the form of rules. Users and developers are
primarily interested in the flexibility and power of knowledge representation provided by such a
language and in the knowledge processing muscle of the corresponding inference engine.
An expert system’s user interface supports the interaction between a user and the inference
engine during a consultation session. Common interfacing techniques include commands,
forms, icons, menus, and their combinations. The general principle of designing a user
interface is that it should match what users of the noncomputer system have been
accustomed to. Dual communication is provided by an interface: users ask the inference
engine for advice and users are asked for specific data by the inference engine during
consultations. In addition, a user can ask to explore the inference engine’s line of reasoning
after the deduced advice is presented. The ability to explain the reasoning behind a
recommendation enables both naive users and experts to understand the rationale underlying
a piece of advice.
3.5. Applying Expert System To Commercial Loan Decisions
The user interface allows users to supply facts about the specific problem, execute the

knowledge system, and obtain reports. Written reports are especially important for loan
analysis because loan decisions often must be justified. A loan decision is also subject to
review in future periods if the borrower requests another loan or defaults on the current loan.
The knowledge system contains expert knowledge regarding the kinds of problems the system
is designed to solve. For commercial loan analysis, this knowledge is acquired from experts in
lending and includes procedures, rules, heuristics, and general facts.
The inference engine applies the knowledge of the expert to the facts that describe a specific
situation. For commercial loan analysis, these facts come from loan application, the borrower’s
financial statements, and average financial ratios for the borrower’s industry.
21
The inference engine uses the knowledge system to conclude new facts from the large
number of detailed facts supplied through the interface. The final conclusion of the commercial
loan system is advise on whether or not to grant the loan in question.
3.5.1. Knowledge Engineering
The process of building the knowledge system of an expert system is called knowledge
engineering and consists of the following activities: knowledge elicitation, knowledge
representation, and knowledge programming.
During knowledge elicitation, the knowledge engineer extracts from the expert the concepts,
rules, and procedures normally used.
Next, the engineer represents the knowledge in the expert system software.
Finally, the knowledge engineer constructs and revises the system to reflect the changes
(misfits and enhancements) that result from testing the system.
3.5.2. Tools
Expert System can be developed from scratch using high level symbolic programming
languages. However, recently powerful software environments exist for knowledge
representation and processing on general purpose computers, PC’s.
Today, over 60 PC based Expert System Development software tools are available in the
market. These tools are designed specifically to provide a framework to represent the
knowledge system and to provide built-in inference engine.
These Expert System tools have some advantages:

• The built-in Inference engine provides a strong mechanism for searching solution.
• Provides convenient interface with the user and other external programs.
• Most of them are suitable for popular computers.
Expert system shells and other software tools make it possible to built expert systems more
quickly and inexpensively than if general purpose programming languages were used. Some
tools, often called shells, provide all the expert system components except the knowledge
system. The knowledge engineer has only to fill the knowledge system with expertise about a
specific kind of problem, such as commercial loan analysis. If a more general purpose
programming language were used, the system developers would also have to code the
inference engine and user interface.
22
Chapter 4
Design and methodology
4.1. Methodology
4.1.1. Research Methodology
Data relevant to the research will be collected from both primary and secondary sources. This
requires working with some banks in order to conduct interviews/discussions for collecting the
rules and experience in commercial loan decisions in the actual conditions of Vietnam.
The bank head offices and branches in HCMC. will be chosen to visit. Publications of the bank
will also be collected at the time of visiting. Commercial bank lending officers, credit analysts,
and loan committees will be interviewed and requested to test the software after designing.
4.1.2. Research Framework
The process of building the expert system includes the following steps as follows:
• Defining the problem
• Discovering the basic concepts and their interrelationships
• Developing the relevant rules and procedures
• Building an initial prototype and testing it
• Revising and expanding in a series of iterations until the system solves the overall problem
as originally intended.
4.2. Design

4.2.1. Overall Structure Design
The system is a microcomputer-based credit analysis tool in commercial lending. It includes
three parts:
• Install the system information
It provides the means to the users to install or revise the system information as desired such
as which files the system needs and content of these files, which outputs the system
generates and how to calculate them.
• Database management
It provides the friendly user interface to the users to interact with the database and update the
data. In addition, it processes data provided by users to prepare historical and analytical
23
Background
Rules
Credit Risk
Rating Rules
Overall
Financial Rules
Rulebase
information, as well as proformas projections. The program will generate the following on
either an historical or proformas basis:
Balance Sheet
Income Statement
Cash Flow Analysis
Financial Ratios
Figure 4.1. Overall Structure Design
USER
• Consultation
It uses the data provided by database management module and some added information
provided by the users to consult a loan should be granted or not.
4.2.2. Implementing Knowledge Base Design

The knowledge base for the expert credit granting system prototype consist of two
components: a customer database and a rulebase.
4.2.2.1. Customer Database
The customer database contains all available data pertaining to the customers/borrowers. The
structure of the database is shown in Figure 4.2. The various blocks of data (rectangular
24
INSTALL SYSTEM
INFORMATION
DATABASE
MANAGEMENT
CONSULTATION
Structure of
Files
List of Files
Format of
Outputs
List of Outputs
System Information Database
Financial
Statement Files
Customer
Files
Reports
Files
Projections
Files
Customer Database
boxes in Figure 4.2) are ‘files’, and each data item belonging to a file is a ‘field’. The database
is logically divided into many files according to the attribute group to which a specific data item
might belong. The files are related to each other through appropriate key fields.

Each of the files in the database has at least one unique index key and may have multiple
index keys to efficiently perform various search and retrieval tasks. The unique index keys
may be single fields or a combination of several data fields belonging to the file. For example,
a unique record in the income statement database is a combination of two fields: a customer’s
code and the year-end.
Figure 4.2. Database Structure
• Company/borrower Information
Before entering actual financial information, it is necessary to first record the borrower’s name
and provide information about the borrower.
• Financial Information
Financial information consists of Balance Sheet and Income Statement at least one full year.
In Balance Sheet, the data must be balanced; that is, assets must equal liabilities plus net
worth in all historical periods.
• Setting up Annual Projections
There are only a few steps involved in setting up annual projection. Program requires at least
two full years of historical data. In addition, the data must be balanced; that is, assets must
equal liabilities plus net worth in all historical periods.
25
Customer
Information
Key: Cust Code
Annual Projections
Key: Cust Code
Year-end
Cash Flow
Key: Cust Code
Year-end
Information for
Consultation
Key: Cust Code

Balance Sheet
Key: Cust Code
Year-end
Forecast Variables
Key: Cust Code
Financial Ratios
Key: Cust Code
Year-end
Income Statement
Key: Cust Code
Year-end