How to Make the Right Decision , livre ebook

Random House Publishers India Pvt. Ltd. - K Laha Arnab

Découvre YouScribe en t'inscrivant gratuitement

Je m'inscris

Obtenez un accès à la bibliothèque pour le consulter en ligne
En savoir plus

85 pages

English

Vous pourrez modifier la taille du texte de cet ouvrage

Obtenez un accès à la bibliothèque pour le consulter en ligne
En savoir plus

A propos
Informations
Extrait

Description

FROM THE ACCLAIMED IIMA BUSINESS BOOKS SERIESHow can you better manage your inventory by looking at the past movement of your stock? How can you ensure that your customer mailings target the right people to make the most impact?How do you go about hiring the appropriate people for a job profile?Business analytics, the method by which data around a business is analysed to better determine the choices we make, is your answer. In this accessible, user-friendly book, Professor Arnab Laha explains the relevance of this growing field in business and looks at its uses in marketing, finance, operations and HR. He also devotes a special section to setting up business analytics for your workplace. With examples and case studies, How to Make the Right Decision is the most useful book you can buy for yourself and your business.

Sujets

Economics

Informations

Publié par	Random House Publishers India Pvt. Ltd.
Date de parution	24 mai 2015
Nombre de lectures	0
EAN13	9788184007053
Langue	English

Informations légales : prix de location à la page 0,0400€. Cette information est donnée uniquement à titre indicatif conformément à la législation en vigueur.

Extrait

ARNAB K. LAHA

HOW TO MAKE THE RIGHT DECISION
RANDOM HOUSE INDIA
CONTENTS
A Note on the Author
1. In the Eyes of the Lender
2. Where the Money Lies
3. In the World of Risks
4. Putting Your Eggs in Different Baskets
5. Assessing a Customer s Value
6. Segmenting the Market
7. Predictive Modelling
8. The Marketing Mix
9. Too Much or Too Little?
10. Is the Process under Control?
11. Investing in Talent
12. Understanding Why Employees Leave
Notes
Footnote
4. Putting Your Eggs in Different Baskets
Acknowledgements
Follow Random House
Copyright
A NOTE ON THE AUTHOR
Prof. Arnab Kumar Laha is a member of the faculty at the Indian Institute of Management, Ahmedabad. He received his PhD degree from the Indian Statistical Institute. He has a strong interest in the fields of business analytics, quality management and risk management and teaches several courses on these subjects at IIMA. He has published several research papers in national and international journals of repute.
Prof. Laha was featured among India s best business school professors by Business Today in 2006 and Business India in 2012. In August 2014, he was named as one of the 10 Most Prominent Analytics Academicians in India by Analytics India Magazine . He is the convener of the IIMA series of conferences on Advanced Data Analysis, Business Analytics and Intelligence, which is held biennially since 2009. He has been associated as a consultant with several reputed organizations, both in the private and public sectors.
CHAPTER 1
IN THE EYES OF THE LENDER
Sachin is a chartered accountant by profession and works for a reputed multinational firm in India. Recently, he has placed a request for a credit card with a leading Indian bank. After a long wait of six weeks, he receives a notification from the bank that his application has been rejected. Sachin is taken aback by the decision. He earns a decent salary and has furnished all the financial details to the bank, and yet his application has been rejected.
In another interesting instance, Apoorva Bank has recently approved a factory loan to a small-sized firm in Gurgaon while it declined any credit to a mid-sized business operating across India and with revenues twenty-five times the revenues of the small-sized firm.
In the two instances, the banks have used some criteria for selection in granting or declining the credit card/loan to their customers. But what are those criteria and how does a bank decide that it should or should not grant credit to a particular individual or business?
The answer lies in Credit Scoring, which is at the heart of any credit decision-making. A good selection criterion aims at avoiding losses and hence tries to assess the borrower s risk. By selecting customers who are likely to repay the debt, a bank reduces its chances of suffering a loss on the loans granted.
The practice is not new. In the olden days, when a village moneylender would grant credit to a farmer, he would ask for some collateral (crop produce, house papers, jewellery, and so on). In case these were not available, he would assess the farmer s repaying capacity based on his past experience with him, the size of his land, be it owned or rented, the size of his family to assess his expenses, etc. Let s call these factors decision variables . Then, based on his expert judgement he would decide to grant or not grant him a loan.
Now consider a hypothetical case where instead of a few villagers, he has to deal with thousands of borrowers at his shop every day asking for credit. Expert decision-making on a case-by-case basis in this scenario would be inefficient and may generate loans of poor quality. He may also lose some borrowers to other moneylenders, in case he takes too much time to take the decision. What can a moneylender do? Can he improve his decision-making capabilities? Yes, if he can somehow give a ranking to each borrower, based on certain fixed decision variables, and then approve the loan to only those who are above a certain cut-off. To illustrate, if he has ten borrowers lined up for credit, he may number them 1 to 10, 1 being the best and 10 being the worst in terms of potential repayment behaviour. Now, if the moneylender wants to give credit to only 30 per cent of the potential borrowers, he will pick the customers numbered 1, 2 and 3 for the same.
This is what banks do when they have to scan thousands of credit applications month after month.
THE CREDIT SCORE
Let s bring our friend Sachin back into the picture. Assume that the bank uses the following decision variables (characteristics or factors) in assessing a borrower s repayment capacity: age, income and employment status. TABLE 1.1: A Sample Scorecard Decision Variable Range Score Points Age Upto 25 70 25-40 100 Above 40 120 Income (per month) Upto 20,000 50 20,000-50,000 80 Above 50,000 100 Employment Status Unemployed 20 Self-Employed 50 Private 80 Government 100 Others 60
If Sachin s age is 28 and he draws a monthly income of Rs 30,000 working in a private sector, he will get a score of 260 (100 + 80 + 80 as mapped from Table 1.1 ). The higher the score, the better the chances that he will repay the debt. Similarly, the lower the score, the lower his chances of being able to repay the loan. Hence, the bank will compare the score against its predetermined cut-off and will decide on whether credit should be granted or not. The cut-off is decided based on the bank s appetite for risk and its strategy in terms of growth, targeting, and so on.
The next question to ask is how does a bank come up with the score points for particular decision variables and how does it choose the score cut-off?
This is where data analytics and statistical techniques play a significant role.
THE CREDIT SCORE PROCESS
A three-stage process is followed in assessing the borrower s chances of defaulting on the loan and this assessment is then used for granting him a loan or not.
Stage 1: Preparing the Data
Data lies at the heart of Credit Score development. In its simplest form, data includes a list of the existing customers (see tip below on reject inference ), together with the customers characteristics as of the date when their loan/credit was approved. In addition, an indicator is created to flag the customer who has been behaving well on repayment terms unlike another who has defaulted on the loan.
Figure 1.1 highlights the data preparation for the model development process starting in, say, January 2010. Here, one has chosen to track all the approved loans of a bank from 2007 for a period of twenty-four months. The period from which approved customers are chosen is called the observation period and the period for which the repayment behaviour is tracked is called the performance period . A performance tag of 1 (Good) or 0 (Bad) is calculated for each customer during the performance period.

FIGURE 1.1: Data Preparation Period
The definition of a performance tag depends upon the characteristic of the loan product. For shorter-term loans, say car loan or student loan, one may decide to use the shorter window of performance and tag a borrower as Bad if he misses two consecutive payments during the performance period. For some other products, say, home loan, one may decide to tag a borrower as Bad only if he misses six consecutive payments.
One can also exercise flexibility in choosing the starting date and the length of the observation period. It all depends upon how much data one wants to track and how recent the data should be.

While looking at only existing customers is convenient, one should not forget the accounts rejected in the past since it will bring in the bias due to past strategy through which only a subset of accounts had been approved.
To overcome this, a technique called reject inference is used to estimate the behaviour of rejected accounts. Ideally, both existing and rejected accounts should form part of the modelling dataset.

Please ensure that the performance period is used only to tag good or bad behaviour. No other loan characteristic should come from this period.
Stage 2: Developing the Model
Model development forms a critical step as the collected data in Stage 1 is used to generate the score points for each loan characteristic that is important in predicting repayment behaviour.
While various techniques are available to model such behaviour, the logistic regression method is used most often. The name derives from the use of the logistic function in the Regression Technique. 1 The technique is used primarily to predict the categorical behaviour (behaviour or outcome with a limited number of possibilities) using the independent factors that one considers critical to such behaviour. In our case, it is being used to predict the repayment behaviour of a borrower (taking two values: Good behaviour and Bad behaviour) based on the borrower s characteristics at the time of loan application. The underlying assumption is that the past predicts the future. If, in the past, certain characteristics have led to a particular behaviour or outcome, the occurrence of those characteristics in the future may lead to the same behaviour.
Once the logistic regression technique is applied to the data at hand, one may get the result that only age, income and employment status matter in borrowers behaviour to repay the loan over time (these are termed statistically significant variables ), and that other characteristics such as monthly expenses, location, etc., can be ignored.
The model also assigns the relative importance (weight) to each significant characteristic. For example, age may get a weight of +0.9. A positive sign implies that the higher the age, the better the chances of the borrower not defaulting. If the sign were negative, the opposite relationship would hold.
Weights for each of the significant variables can be converted into an easy-to-use scorecard as depicted in Table 1.1 . The