Contents 1. Executive Summary: ............................................................................................4 2. Objective ..............................................................................................................5 3. Company Background .........................................................................................5 3.1. Solutions ........................................................................................................5 3.2. Services..........................................................................................................6 3.3.

Vision, Mission and Values ..........................................................................7 3.4. Success Stories ..............................................................................................8 4. Segmentation .......................................................................................................9 4.1. Types of segmentation.................................................................................10 4.2. Usefulness of segmentation .........................................................................10 4.3. The Business Challenge ..............................................................................10 4.4. The Solution ................................................................................................11 4.5. Benefits of Segmentation ............................................................................11 4.6. Customer Segmentation Applications .........................................................13 Determination of customer value .............................................................................13 5. Fractals Segmentation Framework ...................................................................14 5.1. Fractals Value Based Segmentation vs. plain vanilla segmentation .........17 5.2. Characteristics of Fractals Framework Based Segments ...........................19 5.3. Key Parameter Based Findings ...................................................................22 6. Modeling ............................................................................................................23 6.1. Steps: ...........................................................................................................23 7. Credit Card.........................................................................................................27 7.1. Entities Involved..........................................................................................27 7.2. Benefits of Credit Card................................................................................27 7.3. A typical credit card transaction..................................................................29 7.4. What does a customer have to pay? ............................................................29
1|Page

7.5. Understanding a credit card statement ........................................................31 7.6. Various payment options .............................................................................31 7.7. Revolving credit ..........................................................................................32 7.8. Revenue Streams .........................................................................................32 7.9. Expense Streams..........................................................................................32 7.10. 7.11. How do the Banks make Profit ................................................................33 Reasons for Profit Decline .......................................................................33

8. Segmentation of Credit Card Customer for a Bank...........................................34 8.1. The Workflow .............................................................................................34 8.2. Distribution of Raw Data ............................................................................35 8.3. Categorization of Raw data ............................................................................36 8.4. Preparation of DQ-report.............................................................................40 8.5. Missing Value Treatment: ...........................................................................41 8.6. Rolling Up data in Customer level: .............................................................42 8.7. Outliers operation: .......................................................................................43 8.8. Analysis considerations and assumptions ...................................................44 8.9. Segmentation Framework ...........................................................................44 8.9.1. Population, Revenue and Risk Distribution ..........................................45 9. Segment Analysis & Recommendations ...........................................................56 9.1. Assumptions: ...............................................................................................56 9.2. Segment selected for profiling ....................................................................56 9.3. SWOT Analysis:..........................................................................................56 10. 11. Final Recommendation for Different Promotional Program: ........................62 References ......................................................................................................65

2|Page

Table of Figures:

Figure 1: Segmentation example. ............................................................................12 Figure 2: Fractal's Segmentation Framework ..........................................................14 Figure 3: Typical Value Based Segmentation .........................................................15 Figure 4: Comparison of parameters across segments between two approaches ....18 Figure 5: Typical Credit Card Transaction ..............................................................29 Figure 6: The Workflow of Project ..........................................................................34 Figure 7: Distribution of Raw Data .........................................................................35 Figure 8: Segmentation Framework ........................................................................44 Figure 9: Cross Tabulation Result ...........................................................................45 Figure 10: Vintage Distribution ...............................................................................46 Figure 11: Distribution of Customers According to Credit Limit ...........................47 Figure 12: Calculated Derived Variables for Every Segment .................................48 Figure 13: Distribution of Attrition Score ...............................................................49 Figure 14: Distribution of Behavioral Score............................................................50 Figure 15: Distribution of Revolvers for each Segment ..........................................51 Figure 16: Distribution of Cash Advance Customer For Each Segment .................52 Figure 17: Distribution of MCC Transaction ..........................................................53 Figure 18: Distribution of Age for Each Segment ...................................................54 Figure 19: Distribution of Different Types of Cards for Each Segment .................55

3|Page

1. Executive Summary:
The project name Segmentation of Credit Card Customers for a Bank has been carried out for a particular bank in customer. The main objective of the study is to segment the total credit card customer base provided by the bank and find out the target segment and suggest required promotional activities which can lead to an increment in revenue of the bank. Here actual client is Visa who has given the project to solve their dilemma that what are the different segments they should target and what promotion activities they should run to increase their revenue. Before continue with the analysis we needed to understand the total credit card operations with different charges which are applied in general credit card transactions. We also needed to know the different revenue streams for bank, customer, and merchants .Here in this report I have given a brief overview of credit card operations. After that I needed to know the different segmentation approaches which are being followed by Fractal Analytics Ltd. So in the report I have discussed about segmentation and benefit of that along with Fractals segmentation framework and difference between plain vanilla based segmentation and value based segmentation. Here visa has given eight datasets comprising near about 2000 variables with 1 million records. These datasets gives different information like Balances, Delinquency information, billing information, Demographic information, Attrition details, and transaction details of 3 years. After getting these datasets we calculated some variables i.e. number of missing values, maximum, minimum, p1, p25, p95, p99 for some variables and also calculated frequency, cumulative frequency and percentage for some variables which are categorical type variables. After doing this we need to roll up the data into customer level as the bank had given the data in card level. To get the data into customer level we have executed SAS code which directly converted the data into customer level. After getting the data into customer level it was need to do the outlier operation. The outlier operation comprises of Capping and Flooring. Capping means removing the very high values and Flooring means removing very low values. After doing the outlier operation we have done segmentation on the basis of two dimensions i.e. Expected revenue and Risk score. Here we have segmented into a 3X3 matrix. After segmenting the customers we calculated different derived variables for each and every segment which helped us to analyze the problem and give recommendation which may lead to revenue.
4|Page

At the end I have suggested some marketing strategies for targeted segments i.e. three promotional campaign such as Credit line increase program, Card up gradation program and MCC promotional program which may lead to increase revenue.

2. Objective
The objectives of this project mentioned below. i) ii) iii) Segment the total customer base. Analyze the segments and find out the targeted segments. Suggest Marketing strategies for the targeted segments.

3. Company Background
Fractal is a leading provider of advanced analytics that helps companies leverage data driven insights in making better decisions. Our analytics solutions help companies across the globe enhance profitability by powering their customer management efforts with scientific decision making. We have deployed analytics to drive significant value for many Fortune 1000 companies in the areas of customer acquisition, retention, value enhancement, risk management and marketing optimization.

We have successfully leveraged a Global Delivery Model in analytics to provide high impact and lasting solutions in a highly cost-effective manner ensuring maximum value to our clients. We have delivery centers in Mumbai, New Delhi and Kolkata in India, and consulting offices in the USA and Singapore.

We work with clients in multiple geographies including USA, Canada, Latin America, India, South Korea, Taiwan, and Australia. Our clients include Fortune 500 companies in financial services, insurance, retail and CPG domain.

3.1. Solutions
With information and data becoming more crucial each day, analytics is evolving to be an inherent part of all business processes. Fractal understands this, offering you analytics solutions that are unmatched in quality and content, developed through years of expertise. Each solution carries with it a history of successes, each time making clients wiser, helping them take more intelligent decisions.
5|Page

CRM ANALYTICS Analytic solutions deployed to lower the cost of customer acquisition, optimize customer retention and revenues per customer MARKETING MIX MODELING Analytic solutions deployed to improve brand performance, optimize marketing spends, enhance sales volumes and efficiently use of media budgets BUSINESS ANALYSIS Analytic solutions for multi-dimensional reporting, analysis and insight on business, consumer and competition

CONSUMER INSIGHTS Analytic solutions to understand consumer behavior, attitudes and needs for different segments and then devising effective growth strategies RISK ANALYTICS Analytic solutions geared towards controlling default and bad debt as well as improving collections CUSTOMER LIFETIME VALUE Analytic solutions to derive a single, financial score to each customer allowing for the evaluation of the relative importance of each customer as they contribute to the profits of an organization

3.2. Services
Fractal offers services to span the entire cycle of analytics and knowledge research. Right from text mining and data collection to the final predictive modeling, Fractal offers completely integrated as well as modular approaches to client's requirements.

Modeling and Analysis

Our modeling and analysis services enable businesses to address many critical issues such as credit scoring, behavioral scoring, pricing, valuations and predicting customer behavior accurately and well in advance.
6|Page

Data Management
Our data management services include building custom data models, performing data transformation and cleansing, and creating transformation engines.

Text Mining
Our text mining services help bring structure to unstructured text data and then use it to understand and predict customer behavior.

3.3. Vision, Mission and Values

Fractal Vision To be the most respected provider of Analytics across the globe To create wealth through breakthrough contribution to analytics To nurture and groom an exceptional set of professionals Fractal Values Client Value Creation It is the basis of our existence and the driver of our growth. We will strive to create enduring value for our clients through excellence in service delivery, win-win relationships and a relentless focus on our clients business issues. Excellence We will work relentlessly to improve every aspect of our service delivery and through this endeavor be the best in our field. There will be no place for mediocrity in our actions, delivery and results. Speed We understand that speed is a competitive weapon for our clients and for us. It is our commitment to act with speed and reduce turnaround times in everything we do. Innovation We will innovate to stay at the leading edge of our industry and constantly enhance our capability to add more value to our clients and shareholders. We will always challenge the status quo in everything that we do. Respect and fairness We will treat all individuals and stakeholders with respect and be fair in our transactions. We will ensure that anyone transacting business with us will do so with the comfort and confidence that we will be open and fair manner. Integrity We will be transparent and open in the way we conduct ourselves. By always being transparent we will develop enduring trust in our words and actions. Professionalism We will promote a work culture where commitment, responsibility and accountability are integral and non-negotiable components of our work ethic.
7|Page

3.4. Success Stories

Financial Services Fractal builds the analytical function to drive acquisition, retention & value management programs Fractal helps leading company determine the correlation between verification and customer repayment behavior and optimize its overall verification methodology Retail bank reduces loss provisioning requirement by 46% deploying Fractal's Mortgage Metrics application scorecard Fractal's risk scorecards helped an NBFC reduce delinquency levels by 76% Card issuer increases collections by USD 4 million per annum through use of Risk Trees Leading Asian credit card issuer reduces default rates ICICI bank increases approvals without increasing default rates Fractal's CRM analytics solution helped a large bank reduce retention efforts down to 40% Credit card issuer increases spend per customer through advanced spend analytics Predictive models help card issuer reduce attrition by 30 basis points Fractal enables a young dynamic bank grow asset book through effective crosssell programs Consumer bank increases cross-sell conversion by 30% using Fractal's models

Retail Leading apparel retailer analyzes customer behavior & store performance Leading movie rental client enhances performance of recommendation engine by 4% Leading CPG company identifies the most effective in-store promotion vehicle for various brands Auto part distributor reduces defaults through behavioral risk models Known Value Items analysis helps retailer in pricing strategy Drivers analysis enable a retailer to understand drivers of quality perception CPG Promotions analysis enables CPG company to increase promotions effectiveness by 25% CPG brand uses marketing mix modeling to increase revenues by 5% while staying spend neutral Affinity analysis increases effectiveness of cross brand promotions Brand attrition analysis enables CPG company to enhance market share
8|Page

 Price models enable CPG company to achieve right relative price across SKUs Drivers analysis enables leading skin care brand to identify drivers of brand preference Insurance Leading insurance company increases cross-sell conversion by 40% and annual profit by $600,000 Fractal helps a leading insurance company reduce cost of acquisition by capturing 78.5% contacting just 40 % of the population Fractal provides automated data aggregation and attrition analysis for a leading insurance firm Telecom Telecom operator manages customer churn through Fractal's models De-duplication and campaign reporting for a telecom operator

4. Segmentation
Segmentation is an effective technique that helps card issuers to understand their portfolio and identify key customer segments to enhance the effectiveness of their activation and usage programs. Customer segmentation is the subdivision of a market into discrete customer groups of individuals that are similar in specific ways relevant to marketing. Customer Segmentation is a powerful means to target groups effectively, and allocate marketing resources to best effect. Traditional segmentation focuses on identifying customer groups based on demographics and attributes such as attitude and psychological profiles. Value-based segmentation, on the other hand, looks at groups of customers in terms of the revenue they generate and the costs of establishing and maintaining relationships with them. Value Based Customer Segmentation Customer segmentation is the practice of dividing a customer base into groups of individuals that are similar in specific ways relevant to marketing, such as age, gender, interests, spending habits, and so on. The customer base can be credit card base, savings account base, debit card base etc. Using segmentation allows us to target groups effectively and allocate marketing resources to best effect. Traditional segmentation focuses on identifying customer groups based on demographics and attributes such as attitude and psychological profiles. Value-based segmentation, on the other hand, looks at groups of customers in terms of the revenue they generate and the costs of establishing and maintaining relationships with them. Data repository of banks store huge amounts of detailed information of their every interaction with all customers which can help them understand their credit card portfolio better in terms of
9|Page

customer behavior, growth opportunities and problem trends. Segmentation essentially means slicing/ dicing/ grouping data on the basis of various parameters so as to unfurl the knowledge hidden in it. There are many ways of segmenting and extracting information out of data. Based on marketing objectives the customer portfolio can be compartmentalized on the basis of psychographic, behavioral and demographic dimensions.

4.1.

Types of segmentation

Based on statistical techniques there are two different ways of segmenting: Directed Segmentation When the dimensions for the segmentation process are known then it is called Directed Segmentation, for e.g. segmentation against behavioral data, psychographic data etc. Undirected Segmentation When the dimensions of segmentation are not known at the start of the segmentation process i.e. it is mainly data driven it is referred to as Undirected Segmentation. Techniques like clustering are used to perform this kind of segmentation process.

4.2.
1. 2. 3. 4. 5.

Usefulness of segmentation

Provides a periodic health check of your portfolio. Helps manage portfolio growth better. Helps in designing and running campaigns on different segments. Credit limit utilization analysis can be used for limit renewal of segments. Identification of customer characteristics of customers acquired from different channels. 6. Identification of triggers for attrition which can be used for proactive retention. 7. Know when its time for changes in marketing and risk practices.

4.3.

The Business Challenge

The vision is to analyze and understand the credit card portfolio in depth to facilitate developing profit driven and risk-averse business strategies. This included knowing the weak points of the business and to improve on that. Some of the frequently faced card business questions include:
10 | P a g e

Which customer segment(s) should be targeted for activation campaigns? Which electron card customers can be targeted for card upgrades to classic/gold/platinum? Which customers are likely to respond to different types of co-brand and loyalty programs? Which segment(s) are likely to respond to credit card cross sell offers? Which customer segment(s) are likely to be part of Less Cash Society (LCS) and more inclined to take services like Bill Pay? Which customers segments have a higher proportion of international transactions? How to incentivize customers to increase POS usage (and generate more interchange revenue) on the credit card? The challenge is to come up with a composite solution suite that readily provides a transparent insight into the state of matter. Drilling down deep into different problems simultaneously to find the root causes and identify patterns which in turn can be acted upon to resolve the issue Recognizing the sweet-spots of business, wherein more efforts can be put to push the profit margin up. This requires an extensive and multifarious review of the business, analysis and interpretations. To find a wholesome solution package is the actual challenge essentially.

4.4.

The Solution

Customer Segmentation is most effective when a company tailors offerings to segments that are the most profitable and serves them with distinct competitive advantages. This prioritization can help companies develop marketing campaigns and pricing strategies to extract maximum value from both high- and low-profit customers. Customer segmentation procedures include: deciding what data will be collected and how it will be gathered; collecting data and integrating data from various sources; developing methods of data analysis for segmentation; establishing effective communication among relevant business units (such as marketing and customer service) about the segmentation; and implementing applications to effectively deal with the data and respond to the information it provides.

4.5.

Benefits of Segmentation

Segmentation improves a companys focus on the following dimensions: Profitability: All customers are not created equal; segmentation helps to align expense base with potential profits Customer: Customers want to be recognized as different with distinct needs; segmentation results in improved retention of key accounts Product differentiation: Segmentation helps identify the attributes of a successful product and also aids in new and differentiated product development. Competition: Your best competitors will pick-off your best customers; banks need to improve retention
11 | P a g e

Resources: Segmentation allows companies to align the right resources with the right opportunities

As illustrated in the figure below, Segmentation leads to differentiated marketing approaches.

Figure 1: Segmentation example.

Steps to Increased Profitability Better retention

Examples of Segmentation-based Initiatives: Product/Pricing/Service Assigning bankers, targeting high-value at-risk customers Creating loyalty/affinity relationships to stimulate account purchase/sales Differentiating service levels by segment Creating propensity models that incorporate prediction for profitable sales Designing products for cross-selling based on preferences of profitable customers (pull versus push) Re-pricing deposits and loans, including selective pricing decreases Mapping customers in current and grandfathered products to better fitting options Adjusting funds availability Identifying high opportunity loan and deposit prospects Focusing on total lifecycle profitability (not maximum short-term margin) Refining measures and incentives to be aligned with: Differentiated service, especially for target customer High priority sales Retention

Improved cross-sell

Targeted pricing / product design High-impact prospecting

Enhanced metrics

12 | P a g e

4.6.

Customer Segmentation Applications

Suggested Approach Movie festivals, and other similar Value added service offers to VIP customers Trip Wire Mechanism Impact Increased Loyalty & Retention of high spends form these segments Retention of the best customers Compensate for risk & reward the non risky customers Up sell in these low risk segments using differential pricing High increase in spends due to high retail spend/ transaction for these segments

Recommended Solution Retention program

Differential Pricing

Risk Based Pricing

Differential Pricing Test Marketing Loyalty up-sell program Loyalty up sell promos targeting increased transactions (e.g. double rewards points threshold of 10 transactions/ month) Departmental Stores/ Supermarket Discounts & Promotions (e.g. Threshold based entries in lucky draws held by Company)

Usage/Spend Increase program

Significant Spend Increase achieved due to presence of a large card base in these segments

Determination of customer value

A Spends vs. Transactions matrix incorporates the average amount of spend on the card, and compares it with other customers. This gives an idea of the probable revenue potential of the customer, thus clearing the way for value migration. Determination of customer value proves very useful in target marketing. For example, high spend low transaction customers probably have the potential to make more high value transactions, and thus can be targeted. Products can also be customized based on customer need and spend profile, thus leading to higher response rates.

13 | P a g e

5. Fractals Segmentation Framework

Figure 2: Fractal's Segmentation Framework

In the above picture depicts Fractals segmentation framework. Here after getting the raw data the overall customer portfolio is made based on Frequency Monetary methodology. In frequency monetary method it is analyzed how many times one customer makes transactions. Based on that a primary customer portfolio is prepared. After preparing customer portfolio actual segmentation is done as per their framework i.e. matrix form. The matrix might be 4X4 or 3X3 depending upon the business understanding. In general there are 4 major segments i.e. for example High Spending High Transaction, High Spending Low Transaction, Economy spenders, and Low Spending Low transaction. Here Spending and Transaction are the two dimensions taken for segmentation. Now we have to proceed to Targeted mass marketing. Here we suggest what are the different types of campaign can be run on segments based on their behavioral history. Inactive customers can be given offer to active them etc. In this way the whole segmentation project is done.
14 | P a g e

The below chart illustrates a typical value based segmentation of spends vs. transactions

Figure 3: Typical Value Based Segmentation

Amount of spend (Increasing) D1 C1 B1 A1

Frequency of transaction

D2 D3

C2 C3

B2 B3

A2 A3

(increasing) Only user Inactive ATM

D4

C4

B4

A4

15 | P a g e

The above mentioned Segmentation is done by identifying customers who are as much homogeneous as possible within segments and as much heterogeneous as possible across. In such a well-designed segmentation framework, customers within a segment have similar interests, need and behavior and respond similarly to a marketing effort. The process involves:

Segmenting and profiling across business dimensions Identifying segment level challenges Identifying and quantifying business opportunities Analyzing the trends and recommending marketing strategies to enhance profitability

This provides the much required insight starting from a macro (portfolio) level view and gradually reaching to the micro level. The above mentioned Segmentation is done by identifying customers who are as much homogeneous as possible within segments and as much heterogeneous as possible across. In such a well-designed segmentation framework, customers within a segment have similar interests, need and behavior and respond similarly to a marketing effort. The process involves:

Segmenting and profiling across business dimensions Identifying segment level challenges Identifying and quantifying business opportunities Analyzing the trends and recommending marketing strategies to enhance profitability

This provides the much required insight starting from a macro (portfolio) level view and gradually reaching to the micro level.

Fractals Segmentation Framework Rationale Segmentation is done across the dimensions of retail spends and number of retail transactions Based on this basic information, other important cardholder behavioral and demographic characteristics and risk are analyzed

16 | P a g e

 Retail spends and transactions are the exhibitors of customer behavior and capture other information such as o Customer Loyalty (Higher number of transactions implies higher loyalty) o Revenues (higher spends and transactions implies higher revenue levels) o Profitability (In an environment with a mass marketing focus, revenues and profitability are correlated) Key Advantages o Pareto principle achieved o Simple and intuitive o Achieves the key objectives of segmentation arriving at distinct segments.

5.1.

Fractals Value Based Segmentation vs. plain vanilla segmentation

Essentials of a Segmentation Strategy o A good segment strategy involves four basic concepts: o Definition Each segment must have definable characteristics in terms of behavior and demographics so that a customized marketing strategy can be developed and implemented for that segment. o Strategy The marketing budget must be allocated and driven by the segment strategy. Each segment will usually need its own marketing strategy with different messages and rewards. o Infrastructure An analytical and campaign infrastructure to support the segment strategy. The infrastructure should measure success and support automated segment campaigns. The infrastructure will involve automated communications and a feedback loop to learn what is working. o Action Plan There should be an action plan with definite goals, milestones, timetables and budgets Characteristics of an Ideal Segment o An ideal Segment Has definable characteristics in terms of behavior and demographics Is large enough in terms of potential sales to justify a custom marketing strategy with appropriate rewards and budget Makes efficient use of available data to support segment definition and marketing efforts Can be measured in performance, with control groups

17 | P a g e

Clients Segment Definition

A1 A2 B1 B2 B3 C1 C2 C3 D1 D2 D3

New Cust Long term dormant VIP Pt 500/700 VIP Pt1200 Dormant VIP Heavy GP user General GP user GP user showing signs of dormancy in past 3 months Heavy CA user Gen CA user CA user showing signs of dormancy

Comparison of parameters across segments between two approaches:

1.
Average Revenue/ Customer/ month

Average Revenue
Average Revenue: Value Based Segmentation
200,000 150,000 100,000 50,000 0
A3 A A 1 2 B 4 C C 4 C1 C C D1 D D D 3 2 A 2 3 4

Average Revenue: Existing Client Segmentation

Average Revenue/ Customer/ month

160,000 140,000 120,000 100,000 80,000 60,000 40,000 20,000 0

A1 A2 B 1 B2

B3

A4

D2 D1 C1 C2 C3 D3

B B B 3 1 2

Segment s

Segment s

2.

Cash Revenue Share

Cash Revenue Share: Existing Client Segmentation 70 % 60 % 50 % 40 % 30 % 20 % 10 %0 %
D 1 D 3 A 2 A 1 D 2 C 2 C 3

Cash Revenue Share: Value Based Segmentation

70 % 60 % 50 % 40 % 30 % 20 % 10 % 0 % C A D 4

Cash Revenue Share

A 1

A 2

A 3

A 4

B 1

B 2

B 3

B 4

C 1

C 2

C 3

C 4

D 1

D 2

D 3

Cash Revenue Share

B 1

B 2

B 3 C 1

Segment

Segment

Figure 4: Comparison of parameters across segments between two approaches

18 | P a g e

5.2.
i)

Characteristics of Fractals Framework Based Segments

Characteristics of Segments A1 and A2:

A1 A2

1. 2. 3. 4. 5. 6.

Very High Volumes & Good Revenues The customers activate in a short period of time & have almost Zero Delinquency Comprise of KEB veterans who are internally sourced & loyal. Convenience Users (General Purchase) but averse to Cash Advance & Installments These customers have high large credit limit but low credit utilizations Highest Merchant spends for A1 & A2 at : Restaurants, Departmental Stores, Gasoline & Airlines 7. Top International Transaction Countries for A1 & A2 : US, Japan, China & Canada ii) Characteristics of Segments A3, B4 and A4:

A3 B4 A4

1. 2. 3. 4.

Most Profitable Segments with the Highest Revenue High ticket size per transaction Broad-based Revenue Profile - Avail Cash Advance & Installment facilities as well The profile in terms of sourcing, vintage and affinity cards pattern resembles those of A1 & A2

19 | P a g e

5. Highest Merchant spends for A3 at : Restaurants, Airlines, Retail-Fashion & Gasoline 6. Highest Merchant spends for A4 at : Retail-Fashion, Gasoline & Auto-Services 7. Highest Merchant spends for B4 at : Retail-Fashion, Restaurants, Dept. Stores & AutoServices 8. Top International Transaction Countries for A3, A4 & B4 : US, Japan, China, UK, Italy & Australia iii) Characteristics of Segments B1 and B2:

B1 B2

Very similar to A1, A2 but lower spends The average age for these two buckets is around 40 Years 9% of total revenues contributed by these segments, while exhibiting no perceptible risk threat 4. Highest Merchant spends for B1 at : Departmental Stores, Restaurants, Gasoline & Others 5. Highest Merchant spends for B2 at : Restaurants, Departmental Stores, Gasoline & Others
1. 2. 3.

iv)

Characteristics of Segments C2 and C3:

C2 C3

1. Most of the customers in these two segments can be directly mapped to the current high and moderate GP usage while the rest can be mapped to the moderate CA usage 2. Provincial youth (under 35) constitute a major proportion of these buckets

20 | P a g e

3. Significant percentage of these customers have been sourced internally and exhibit good vintage 4. Very little foreign transaction and very little secondary card penetration 5. Highest Merchant spends for C2,C3 at : Restaurants, Departmental Stores, Gasoline & Others v) Characteristics of Segments D1, D2 and D3:

D1 D2 D3

1. 2. 3. 4.

Satisfactory activity and delinquency levels Low credit limits may be one of the cause of lower transaction levels Government payments stand out as a transaction category Highest Merchant spends for D1 at : Departmental Stores, Restaurants, Gasoline & Others 5. Highest Merchant spends for D2 at : Restaurants, Gasoline, Departmental Stores & Others 6. Highest Merchant spends for D3 at : Gasoline, Restaurants, Departmental Stores & Others vi) Characteristics of Segment D4:

D4

21 | P a g e

1. Biggest segment comprising of around 40% of all customers 2. Sourcing and vintage suggest predominantly external customers, with relatively new vintage 3. Behavior score distribution, high proportion of Classic Cards, delinquency patterns imply greater concentration of Sub-Prime customers 4. KEB appears to be cautious with this segment for limit allocations 5. Abysmal levels of retail activity, and a near total dependence on cash 6. High concentration of self employed customers 7. Highest Merchant spends for D4 at : Restaurants, Departmental Stores, Gasoline, & Others

5.3.

Key Parameter Based Findings

The holistic understanding of portfolio can be achieved by checking the distribution of the a few key business parameters, how they vary across segments. The totality of composite factors defines every segment uniquely. The key parameters are:

Activation: Generally 50% of newly acquired customers activate their cards within 15 - 25 days. The initial activity behavior is indicative of the eventual profitability. Expenses on activating a customer can be decided with the help of ACTIVATION PREDICTION models

Cash Penetration: As we move from high spend to low spend buckets the cash advance ratio increases. It is generally observed in a portfolio that D4 bucket is generating more than half of the cash volume. With the highest risk also in this bucket, these customers need to be closely monitored

Pay down Ratio: The buckets where the pay down ratio is less & balance at risk is high are the ones that are of concern. There is a collections problem in such buckets. A collections scorecard to help in optimizing collections strategy is recommended.

Revenue-Interest vs. Interchange: A1 bucket has the highest revenue equivalent. Customers with high spend & high transactions are more profitable. The interest revenue is not a significant part of the total revenue in the high spend buckets. This can be increased by inducing revolve from these segments. Further around 65% of interest revenue is contributed by D4, CA & I buckets. Also it is generally observed that high spending buckets exhibit higher Interchange revenue whereas low spending buckets (Cash Needy) exhibit higher interest revenue
22 | P a g e

Revolve: The low spending buckets have a much higher percentage of revolvers as compared to the high spending buckets. The revenue per revolving customer is starkly higher than the revenue per transacting customer in the low spend buckets

Risk: The low spend volume segments like D4, C4 & CA have higher delinquency ratios, vindicating the high risk in cash withdrawing customers which is a cause for concern. Converting the outstanding into a card loan and repayment through an installment is a viable option as many of these people may not be transacting afresh because of delinquency related issues

6. Modeling
Model A model is linear or nonlinear combination of predictor variables. The objective of a model is to use available historical data predictor variables to predict the event we are trying to model Various techniques to build models Parametric techniques Non-parametric techniques Recursive partition techniques Target Variable The business objective is converted into the variable we are trying to predict which is called Target Variable The target variable is also called dependent variable / outcome variable

6.1.
I. II. III. IV. V. VI.

Steps:

Business understanding Data Understanding Data Extraction Data Preparation Model Development Evaluation & optimization

There are well defined deliverables and activities for each phase of the process.
23 | P a g e

i)

Business understanding. Activity.

o Establishing business and modeling objectives o Familiarization with client business o Identifying misclassification costs o Series of business meetings over 1 week with client teams for e.g. o In a Marketing Customer Acquisition exercise we typically meet the Campaign Managers, Product Team, and IT

Process

Deliverables
o Business understanding report

ii)

Data Understanding

Activity
o Identify key data sources available o Identify in detail data fields that will be available and can be used for building model. o Business meetings with personnel in IT department In an asset to asset cross sell model, we study systems for the availability of demographic data, asset details, transaction data, details of liability relationship In an asset to liability cross sell model, we look at liability account data, transaction behavior, offer details etc.

Process

Deliverables
o Data on electronic media.

24 | P a g e

iii)

Data Extraction

Activity
Process
o Client extracts data based on data extraction request o Provides data to Fractal on electronic media. o Clients IT team extracts data based on specifications in the data extraction request. o Data is transferred to Fractal on electronic media such as CD, tape or via secure online transfers.

Deliverables
o Data on electronic media.

iv)

Data Preparation

Activity
o Clean the data and format the data o Create modeling dataset o Create appropriate target variable o o o o o o Clean the data Transform raw data into meaningful variables Perform missing value imputation Perform outlier detection and treatment Class continuous variables where required Create target variable

Process

Deliverables

o Prepared data ready for modeling

v)

Data Visualization

Activity
o Provides quick visual patterns in data o Provides an intuitive framework to understand the model o Helps is assessing differences in impact of variables on different segments thereby validating or negating the need for segment specific models

25 | P a g e

 Process Deliverables
o Generate univariate and multivariate visualization charts o Data visualization report

vi)

Model Development

Activity
o Develop model on development sample o Validate model on hold-out sample

Process
o Build a scoring model using one or more statistical techniques.

Deliverables
o Initial model report

vii)

Evaluation & optimization

Activity
o o o o Evaluate models capability to meet business requirements Calculate business impact of deploying model Optimize model based on feedback from evaluation Present output to business users

Process
o Evaluate model performance through tests of statistical significance like KS statistic Gini coefficient Concordance Chi-square etc

Deliverables
o Final model o Model tracking report formats o Assessment of results vis--vis business success criteria

26 | P a g e

7. Credit Card
7.1. Entities Involved
Issuing Bank The bank which issues the credit card to the customers. Acquiring Bank The bank which sets up the POS (point of sale) machines at the retailers. Also known as the Merchant Bank Global payment gateways (Association) These provide the basic infrastructure or connectivity for authentication of the credit card Charge slips Tokens which carry the signature of the customer which promise to pay the retailer the specified amount

7.2.

Benefits of Credit Card

i)

Convenience

Benefits of Credit Card to Card Holders

o Convenience of not carrying cash to make purchases. o Hassle free online transactions. o Security of having money all the times.

Credit
o Free credit available for a period of about 50 days. o Facility of revolving credit. o Cash withdrawal facility available against card.

Others
o o o o o Status Symbol. Discounts, reward points Accident insurance, baggage insurance etc. Purchase protection Convenience of paying utility bill using a credit card

27 | P a g e

ii)

Increased Sales

Benefits of Credit Card to Merchants

o Makes easier for consumers to purchase goods hence boosting sales. o Consumers likely to make larger/more purchases when credit is used instead of cash. o Security of having money all the times.

Easy to Authorize
o Transactions are fairly easy to validate by the merchant o When a card is swiped, the information on the card is automatically authorized and the transaction amount is deposited in the merchants account.

Reduced Risk
o Merchant has none of the risks inherent in extending credit or accepting checks. o No need for the merchant to contact a purchaser directly in an attempt to collect money. o The sale is validated by proper procdure and marchant receives payment from its bank.

iii)

Benefits of Credit Card to Banks

Bank raises revenue in terms of joining fees, annual fees, interchange, interest on revolving credit Opportunity to attract customers who do not live nearby. This is because with credit cards, the customers location becomes irrelevant Many banks have seen their asset portfolio grow quickly as a result of credit card programs. Many banks heavily engage in attempts to cross sell additional banking products and services to card holders. In addition the merchants sales draft is also treated as a deposit item which again becomes source of funds for making loans.

28 | P a g e

7.3.

A typical credit card transaction

Figure 5: Typical Credit Card Transaction

7.4.

What does a customer have to pay?

o Joining fee. o Annual fee.

Ownership Fees. Usage Fees o Credit Usage

Penalty fees. Late payment charges. Over limit Account. Collection fee. Interest Fee. Cash withdrawal charges.

29 | P a g e

o Value added services

Charge slip Retrieval fee. Utility bill payment fee.

Some typical charges

Types of charges Joining fees Annual fees Annual Fees for Add on Card Interest fees (pm) Classic Nil Rs. 700 Rs. 350 2.95% p.m. 30% of outstanding subject to minimum outstanding of Rs. 350 and maximum of Rs. 600 Nil Gold Nil Rs. 2,000 Rs. 1,000 2.95% p.m. 30% of outstanding subject to minimum outstanding of Rs. 350 and maximum of Rs. 600 Nil

Late payment charges

Utility bills payment fees

Cash Advance Limit

40% of Credit Limit

40% of Credit Limit

Cash withdrawal fees

2.5%

2.5%

Over limit Fees

Rs. 300 p.m.

Rs. 300 p.m.

30 | P a g e

7.5.

Understanding a credit card statement

Following are the components of a typical credit card statement 1. Billing date: This is the date on which the credit card statement is generated. Different customers have different monthly billing dates. 2. Payment due date: This gives the maximum date till which a payment must be made. This is typically 20 days after the billing date. 3. Previous Balance: This is any previous balance which is due by the customer 4. Purchases: The total amount of purchases made using the card during current cycle. 5. Cash advance: A cash loan obtained by a cardholder through presentation of a card at a bank office, at an ATM, or by mail request 6. Total Payment due: The total amount that the customer owes to the issuing bank. 7. Minimum payment due: It is typically 5% of the total payment due. This is the minimum amount that the customer must pay before the payment due date.

7.6.

Various payment options

o No charges are applied o The customer pays without any interest.

Total Amount paid before the due date

Amount paid is greater than min but less than total.

o The outstanding balance gets charge for interest.

Amount paid is less than min or no payment at all.

o The customer has to pay late fees. o Interest is charged on the outstanding balance as well as any subsequent purchases made.

Amount paid is greater than the total amount due.

o The extra amount is carried forward as credit

31 | P a g e

7.7.

Revolving credit

A credit card statement is generated every month for a customer Among all the other things it also specifies the total amount due, min amount due and the payment due date The min amount due is typically 5% of the total amount due A customer has the option of not paying the total amount He can pay at least the min amount (5% of the total amount) on or before the due date and still not be considered as a potential defaulter. The remaining 95% which the customer still has to pay the bank is termed as the revolving credit The revolving credit along with any subsequent purchases made in the next month/cycle will be charged at the rate of 2.95-3% pm

7.8.

Revenue Streams

Interest Income: Product of Annual % rate, rollover rate and outstanding dollars. Interchange Income: % of the merchant sales that the acquirer is required to pay the issuer. Annual membership fee Other fees and income: Cash Advance fees, balance transfer fees, credit insurance and other nuisance fees such as late payments, exceeding ones credit line etc.

7.9.

Expense Streams

Cost of funds: The interest rate a card issuer pays to finance the outstanding balances in the bank credit card folio. Funds come from a variety of sources including customer deposits and borrowing from other institutions Losses: Bad debt, fraud losses and costs related to the investigation and prosecution of fraud. It is generally expressed as a % of outstanding balances. Servicing expenses: Front-end or customer service costs. Processing Expenses: Total back-end costs Marketing Expenses: All costs associated with acquiring new accounts, increasing usage on existing account and retaining profitable accounts.

32 | P a g e

7.10. How do the Banks make Profit

Credit Cards become more profitable business for the Banks when Customers: Spend more using Cards Revolve Stay with the bank for long time Go beyond due date but still pay Use Additional Services like Bill Payments Make more use of cash withdrawal facility Pay all dues and dont ask for waiver of charges

7.11.

Reasons for Profit Decline

Profits of the Banks on Credit Cards decline when Customers Default Are Inactive Ask for discounts and waivers Leave the service

33 | P a g e

8. Segmentation of Credit Card Customer for a Bank

8.1. The Workflow

Figure 6: The Workflow of Project The above picture depicts the total workflow of the total project. Here we can see the steps of our proceedings. The first step is to collect data from the bank on the total customer base and other details required for the analysis. The second step is to make the customer profile based on the data and business understanding and analyze it at portfolio level. Next step is to segment the customer profile into value based segments based on the profiling on demographics, card characteristics and behavioral dimensions. Now the business opportunities are needed to be identified which can be applied to targeted segments which may help to increase revenue. After doing all these it is time to suggest different marketing strategies like promotional campaign for the targeted segments. Lastly we have to prepare the segment report comprising all the result and analysis.

34 | P a g e

8.2.

Distribution of Raw Data

Figure 7: Distribution of Raw Data The data provided by the bank was of four categories. i) ii) iii) iv) Transaction Data (1st august 2007 to 31st July 2009). Billing Data (1st august 2007 to 31st July 2009) Card Characteristics (As on august 2008) Customer Demographics (As on August 2008)

These data elements are consolidated at the account level, and are used to derive the set of business variables which have been used for the analysis.

Data window considered for the analysis

Data Availability: 24 Months (Aug07 Jul09). Observation Window: 12 Months (Aug07 Jul08). Prediction Window: 12 Months (Aug08 Jul09)

Window Definition
35 | P a g e

Data Availability: 24 Months data i.e. Aug07 Jul09 Observation Window (Window during which the behavior of the customer is being analyzed) Aug07 Jul08 Prediction Window (Window during which customers response is being observed) Aug08 Jul09

8.3. Categorization of Raw data

There eight types of data were given to me with several variables and one million records. S.No. 1 2 3 4 5 6 7 8 Filename Mora 'Saldos 'Pagoingr 'Domic_compto 'Attrition 'Giros_2007 'Giros_2008 'Giros_2009 Details of File Delinquency Information Balances Billing Information Demographic Information Attrition score details Transaction Details Transaction Details Transaction Details Number of Number of variables Records 53 1000000 303 1000000 227 1000000 91 1000000 25 1000000 255 1000000 507 1000000 297 1000000

These are the different type of file given to me. The actual name of every file is also mentioned above along with the number of variables and records. Now the variables in these files have got its own definition. Below the definition of variables for a single month is being described as definition is same for all month. File Name: MORA (Delinquency Information): Variables 'IM_PAGO_MINIMO_09MAY PDTO2 NU_PAGOS_VENCIDOS1_09MAY Definition Minimum Payment Amount Type of Cards. Number payments due to the cutoff date

36 | P a g e

File Name: 'SALDOS (Balances): Variables IM_LIMITE_CREDITO_09MAY 'IM_COMP_09MAY 'IM_DISP_09MAY 'IM_FIN_MES_09MAY M_PROM_EDOCTA_09MAY 'IM_SDO_CORTE_09MAY 'IM_TOT_INT_09MAY 'NU_COMP_09MAY 'NU_COMP_SINT_09MAY 'NU_DISPOSICIONES_09MAY 'PC_TASA_POND_09MAY Definition Line of credit card Total amount of purchase to the court date 'Amount of cash to provisions of court date 'Amount at the end of month Average amount of the state of account Balance as at the date of court Estimated total amount of interest to the court date Number of purchase to the court date 'Number of purchases no interest to date of court Effective number of provisions of the court date Weighted interest rate monthly

File Name: ''PAGOINGR (' Billing Information): Variables 'IM_PAGOS_09MAY 'IM_CUOTAS_09MAY 'IM_COMISION_SOBREGIRO_09MAY 'IM_COMISION_MORA_09MAY 'IM_COMISION_PAGO_TARDIO_09MAY 'IM_TOT_COMP_SINT_09MAY 'NU_EFI_09MAY 'IM_TOT_EFI_09MAY 'NU_PAGOS_09MAY Definition Amount of payments to the court date Annual fee amount Amount of fees for overdraft Amount of fees for arrears in payment Amount of fees for late payment Amount of purchases months interest Number of times that apply immediate cash Immediate cash amount Number of payments to the court dat

File Name: ''DOMIC_COMPTO (' Demographic Information): Variables 'CD_BEHAVIOR_09MAY 'TO_TARJ_ADICIONAL_09MAY PDTO2 CD_CANAL1 CD_EDO_CIVIL CD_OCUPACION CD_SEXO CD_TEL_CEL_CORRECTO
37 | P a g e

Definition 'Behavior score 'Total number of additional cards Type of cards Product sales channel, the missing considered "branch" Marital status holder Profession of the holder Gender of the holder Valid flag cell phone

are

CD_TEL_CORRECTO CD_VAL_DOMI CD_VAL_EMAIL RNG_BEHAVIOR_09MAY RNG_EDAD CD_OFICINA_OR

Flag phone (not cell) valid Flag valid address Flag valid email Behavior score range Age range of account holder Origin of contract office

File Name: 'ATTRITION (' Attrition score details): Variables 'ATTRITIONSCORE_09MAY RNG_ATTRITION_09MAY Definition 'Score predictive of cancellation (attrition) Score range predictability of cancellation (attrition)

File Name: ''GIROS_2007 (' Transaction Details): Variables 'NU_OPERACIONES_07JUL IM_OPERACIONES_07JUL NU_ABARROTES_SUPER_07JUL Definition 'Total number of operations Total capital Number of purchases made in stores grocers / supermarkets Amount of purchases made in stores grocers / supermarkets Number of cash withdrawals Amount of cash withdrawals Number of operations performed in department stores Amount of operations in department stores Number of operations in restaurants Amount of operations in restaurants Number of purchases made in turn of clothing and shoes Amount of purchases made in turn of clothing and shoes

IM_ABARROTES_SUPER_07JUL

NU_RET_EFEC_07JUL IM_RET_EFEC_07JUL NU_TIENDAS_ALMAC_07JUL

IM_TIENDAS_ALMAC_07JUL NU_REST_COM_07JUL IM_REST_COM_07JUL NU_ROPA_CALZ_07JUL

IM_ROPA_CALZ_07JUL

38 | P a g e

NU_SUSCRIP_TEL_07JUL IM_SUSCRIP_TEL_07JUL NU_CLUB_MAYOR_07JUL IM_CLUB_MAYOR_07JUL NU_OTROS_07JUL IM_OTROS_07JUL NU_VIAJES_07JUL IM_VIAJES_07JUL NU_ENTRET_DEPOR_07JUL

Number of purchases made in turn subscription Amount of purchases made in turn subscription Number of purchases made in wholesale clubs Amount of purchases made in wholesale clubs Number of purchases made in other money Amount of purchases made in other money Number of operations in turn travel Amount of operations in turn travel Number of operations in turn of entertainment and sports Amount of operations in turn of entertainment and sports Number of purchases made in turn of consumer goods Amount of purchases made in turn of consumer goods Number of operations in the turn of health Amount of operations at the turn of health Number of operations in turn for home Amount of operations in turn of household Number of operations in other services Amount of operations in other services Number of purchases made in turn of telecommunications Amount of purchases made in turn of telecommunications

IM_ENTRET_DEPOR_07JUL

NU_BIENES_CONS_07JUL

IM_BIENES_CONS_07JUL

NU_SALUD_07JUL IM_SALUD_07JUL NU_HOGAR_07JUL IM_HOGAR_07JUL NU_SERVICIOS_07JUL IM_SERVICIOS_07JUL NU_TELECOM_ELECTR_TEC_07JUL

IM_TELECOM_ELECTR_TEC_07JUL

39 | P a g e

NU_AUTOMOTRIZ_07JUL IM_AUTOMOTRIZ_07JUL NU_EDUCACION_07JUL IM_EDUCACION_07JUL NU_SERV_FINAN_07JUL IM_SERV_FINAN_07JUL IM_NACIONAL_07JUL NU_INTERNACIONAL_07JUL IM_INTERNACIONAL_07JUL

Number of purchases made in turn automotive Amount of purchases made in turn automotive Number of operations in turn of education Amount of operations in turn of education Number of operations in financial services Amount of operations in financial services Amount of purchases made in the country Number of purchases made abroad Amount of purchases made abroad

8.4.

Preparation of DQ-report

After getting the data it is required to prepare the DQ-Report. In DQ- Report there will be some variables which are needed to be computed by SAS code. The variables those are needed to be calculated are N, Nmiss, Mean , Min, p1, p5, p25, median, p75, p90, p95, p99, Max. N = this describes the total records of the variable. Nmiss = this gives the total number of record missing. Mean = this is the average or mean of the value of all records of that variable. Min= the minimum value of that variable among all records. P1 = this is the 1st percentile of the variable. P5= 5th percentile. P25 = 25th percentile. P75 = 75th percentile. P90 = 90 percentile P95 = 95 percentile.

40 | P a g e

P99 = 99 percentile. Median = median of the variable. Max = Maximum value of the variable.

SAS Code for calculating above mentioned derived variables: proc means data=libname.dataset n nmiss mean min p1 p5 p25 median p75 p90 p95 p99 max ; var <variable name>; run; There are some other variables which have different value describing category of particular types such as PDTO2. This variable contains the value which describes different type of cards i.e. Classic, Oro, Infinite, Congelada, Platinum. Now the frequency of every type of card with percentage and cumulative frequency needed to be calculated. To get the Frequency , Percentage and Cumulative Frequency the SAS code is
Proc freq data= libname.dataset; Tables <Variable name>; Run;

Example : CD_EDO_CIVIL . This variable describes the marital status of customer i.e. Married, Unmarried, Divorced, Single union, Widow, Separated. So we need to calculate the frequency, percentage and cumulative frequency. The SAS code will be Proc freq data= libname.dataset; Tables CD_EDO_CIVIL; Run; In this way all the above mentioned derived variable will be calculated according to the variables for each and every dataset. Thus the full DQ-Report was prepared.

8.5.

Missing Value Treatment:

In the whole dataset provided by the bank there are lots of values which are missing. These are the bugs in the dataset. Analysis cannot be done including missing parameters which might lead to wrong interpretation. For example if age is missing then we take the average age and replace the missing age by average age. Sometimes we need to delete the record for missing value.

41 | P a g e

Suppose it is missing that the number of transaction in a particular month. In that case if we take average it might lead to wrong interpretation. So in those cases records are deleted.

8.6.

Rolling Up data in Customer level:

The data provided by bank was at card level. To analysis it from customer point of view and segment the customer of credit card we needed to rollup the dataset into customer levels. For Example all the dataset is having data according to customer id and card id. That means if one customer is having 10 cards then there are ten different card numbers with customer id. If customer made transaction with different card in different month we dont know how many times the customer has made transaction with a particular card. To know that we need to roll up the data into customer level. In customer level we get the total card one customer having, the total number of transaction one customer has done in a particular month with a particular card, total number of transaction in particular shop or restaurant or anywhere etc. All these information we obtained after rolling up data is in customer level, depending upon which we had done segmentation as we wanted to segment the customer not the cards. To roll up data for a particular dataset for a particular month the SAS code is given below: Example: Here the dataset Balance for the month august 07 has been rolled up. proc sql; create table indranil.cust_balances_ago07 as select NU_CTEBIS, count(ID_CUENTABIS) as cnt_acct_id, max (IM_LIMITE_CREDITO_07ago) as Max_IM_LIMITE_CREDITO_07ago , sum (IM_COMP_07ago ) as SUM_IM_COMP_07ago , max (IM_DISP_07ago) as Max_IM_DISP_07ago , sum (IM_FIN_MES_07ago ) as SUM_IM_FIN_MES_07ago , max (IM_PROM_EDOCTA_07ago ) as Max_IM_PROM_EDOCTA_07ago, min (IM_PROM_EDOCTA_07ago ) as MIN_IM_PROM_EDOCTA_07ago , max (IM_SDO_CORTE_07ago ) as Max_IM_SDO_CORTE_07ago , sum (IM_TOT_INT_07ago ) as SUM_IM_TOT_INT_07ago , max (NU_COMP_07ago ) as Max_NU_COMP_07ago , max (NU_COMP_SINT_07ago ) as Max_NU_COMP_SINT_07ago , max (NU_DISPOSICIONES_07ago ) as Max_NU_DISPOSICIONES_07ago, avg (PC_TASA_POND_07ago ) as avg_PC_TASA_POND_07ago from indranil.balances group by NU_CTEBIS; quit;
42 | P a g e

run; here in the above code we have calculated total amount spend and maximum number of transaction in a particular month.

8.7.

Outliers operation:

In outliers operation we eliminate the outliers i.e. the variable which have very high or low value or absurd value which is not possible. The outliers operation has been done on the customer level data. There are generally two types of outliers mainly high and low. To eliminate high value we need to find out the ratio of max to p99 ratio from DQ-report . Here we set up a benchmark ratio 4 as per business understanding. Now for those variables the ratio is greater than 4 we need to do outliers operation on those variables. Now we create a new dataset from the customer level dataset by replacing the value of those variables which had ratio greater than 4 by their p99 values if their actual value is greater than their p99 value. The same operation is done on value which is minimum and in that case the value of the variable is replaced by their p1 value. The sample code for outliers operation for a particular dataset given below. data indranil.outliers_balances_08; set indranil.outliers_balances_08; if SUM_IM_FIN_MES_08ene > if SUM_IM_FIN_MES_08feb > if SUM_IM_FIN_MES_08mar > if SUM_IM_FIN_MES_08abr > if Max_IM_SDO_CORTE_08abr 502.03 ; if SUM_IM_FIN_MES_08may > if Max_IM_SDO_CORTE_08may 230.16 ; if SUM_IM_FIN_MES_08jun > if SUM_IM_FIN_MES_08jul > if SUM_IM_FIN_MES_08ago > if SUM_IM_FIN_MES_08sep > if SUM_IM_FIN_MES_08oct > if Max_IM_SDO_CORTE_08oct 240.03 ; if SUM_IM_FIN_MES_08nov > if Max_IM_SDO_CORTE_08nov 230.82 ; if SUM_IM_FIN_MES_08dic > if Max_IM_SDO_CORTE_08dic 250.01 ; run;

-576.94 thenSUM_IM_FIN_MES_08ene = -576.94 -593.54 thenSUM_IM_FIN_MES_08feb = -593.54 -539.69 thenSUM_IM_FIN_MES_08mar = -539.69 -502.03 thenSUM_IM_FIN_MES_08abr = -502.03 > -502.03 thenMax_IM_SDO_CORTE_08abr =

; ; ; ; -

-533.02 thenSUM_IM_FIN_MES_08may = -533.02 ; > -230.16 thenMax_IM_SDO_CORTE_08may = -552.49 thenSUM_IM_FIN_MES_08jun = -552.49 -522.78 thenSUM_IM_FIN_MES_08jul = -522.78 -528.92 thenSUM_IM_FIN_MES_08ago = -528.92 -507.21 thenSUM_IM_FIN_MES_08sep = -507.21 -540.71 thenSUM_IM_FIN_MES_08oct = -540.71 > -240.03 thenMax_IM_SDO_CORTE_08oct = ; ; ; ; ; -

-298.16 thenSUM_IM_FIN_MES_08nov = -298.16 ; > -230.82 thenMax_IM_SDO_CORTE_08nov = -340 thenSUM_IM_FIN_MES_08dic = -340 ; > -250.01 thenMax_IM_SDO_CORTE_08dic = -

43 | P a g e

8.8.

Analysis considerations and assumptions

1) All analysis has been done at primary customer level. 2) The analysis window starts on 1st August 2007 and ends on 31st July 2008 3) Revolvers are cardholders who have paid >50 pesos finance charges at least once within the analysis window 4) Banks Risk Model has been used for the Default Probability to predict charge-offs 5) Since the card is blocked for further transactions after hitting 3 DPD, such customers were removed from the modeling base 6) Delinquency is 7 DPD (Definition provided by Bank. 7) Interchange revenue has been taken as 1.7%.

8.9.

Segmentation Framework

Figure 8: Segmentation Framework The above figure describes the framework of segmentation which is followed by Fractal Analytics. Here in this project a 3 X 3 matrix has been taken on the basis of two dimensions i.e. Risk Score and Expected Revenue. The two dimensions has been selected on the basis of business understanding which defines the problem i.e. what is the effect of risk score on expected revenue and on the basis of that segmentation has been done.

44 | P a g e

The cut offs for each and every bucket has been calculated on the cumulative basis. Here we have 3X3 matrix. So we should have 3 equal buckets and weight of every bucket will be 33%. So after dividing the two dimensions into three equal bucket i.e. 3 equal cut offs we have to execute cross tabulation macro in SAS so that we can get the all the windows where cumulative weight is 33% or near about 33%. Here we have created three main bucket i.e. A ,B,C and the cross tabulation will give three output for each and every bucket from where we will select the range having 33% weight i.e. A1+B1+C1=33% and A1+A2+A3=33%. The matrix below we got after executing cross tabulation macro for expected revenue, risk score and customer in SAS which directly gives the following output. In this output we can see all the distributions are not equal as mentioned earlier. This might happen due to replacement of data at the time of data preparation. So here we have taken into consideration whether it meets total 100% or not. After that we have created dataset by SAS code with the names A1,A2,A3,B1,B2,B3,C1,C2 and C3 according to the cut-offs we calculated before.
8.9.1. Population, Revenue and Risk Distribution

Figure 9: Cross Tabulation Result

45 | P a g e

Figure 10: Vintage Distribution In the above picture the bucket has been divided on the basis of Vintage. The Vintage describes for how many months the customer is active after the activation of his card. Through SAS code we have calculated the vintage and got the result displayed above. So here we executed vintage code for each and every bucket and got the result for each and every range of months. The range of months has been decided on the basis of business understanding. From this picture we can conclude that as the vintage increases revenue increases and risk decreases and the revenue is near about twice in HL than LL.

46 | P a g e

Figure 11: Distribution of Customers According to Credit Limit In the above picture we have calculated the percentage of total customer of that particular bucket for each and every range of credit limit. The range of credit limit has been decided on the basis of business understanding. The best bucket HL has approx. 70% population with Credit Limit greater than 648K pesos; highest across all segments. The worst bucket LH has the 70% population with Credit Limit less than 115K pesos.

47 | P a g e

Figure 12: Calculated Derived Variables for Every Segment In the above result we have calculated 5 variables by dividing Spends, Transaction, CA Spends, POS Spends, and International Spends by total number of customer of each bucket for each and every bucket. Customers in HL have 3 times higher international spend than the portfolio. Also, 12 times more frequent than customers in LH.

48 | P a g e

Figure 13: Distribution of Attrition Score

Here we have calculated percentage of attrition according to month of activation. High risky customers are showing great attrition trend.

49 | P a g e

Figure 14: Distribution of Behavioral Score Here we have calculated the distribution of customer for each and every bucket against the behavioral score. The range of scores has been decided on the basis of business understanding. Here we can see that more than 40% of the customers in the High Risk Segment have a score of less than 663.

50 | P a g e

Figure 15: Distribution of Revolvers for each Segment In the above figure we have calculated the revolvers percentage for each and every bucket. Here we found that LL Segment has 54% revolvers; lowest across all segments HH has twice more percent revolvers as compared to LL.

51 | P a g e

Figure 16: Distribution of Cash Advance Customer For Each Segment Here in this picture we have calculated the percentage of customer making cash advance transaction for each and every bucket. So we can see 57% high revenue generating customer making cash advance transaction.

52 | P a g e

Figure 17: Distribution of MCC Transaction Here we have calculated the percentage of total spends made on top 5 MCCs. We have found that approximate 55% of total spend has been made on top 5 MCCs and high revenue segments has spent higher percentage on travel.

53 | P a g e

Figure 18: Distribution of Age for Each Segment

Here we have calculated the percentage of different age group of the population of each and every bucket. We found that in low revenue segment the majority is younger, approximate 50% are less than 40 years.

54 | P a g e

Figure 19: Distribution of Different Types of Cards for Each Segment

Here we have calculated the percentage of different cards being used by customers of a particular bucket. High risk segment is using highest percentage of Classic card and high revenue segment is using 5 times more Gold cards than low revenue segment.

55 | P a g e

9. Segment Analysis & Recommendations

9.1.

Assumptions:

The risk probability is provided by bank for each customer Revenue component includes Interest on revolve, Interest on CA, Fees and Interchange revenue SWOT analysis differentiates the important segments from the not so important ones The 6 segments cover both dimensions including the risk and the revenue component Though these segments are given more importance, the other segments are considered while providing recommendation

9.2.

Segment selected for profiling

LL LH

ML

HL HM HH

Now we have selected 6 segments i.e. LL, ML, HL, HM, LH and HH among the 9 segments for profiling because these 6 segments will make substantial difference in the revenue. Other three segments are not so important in generating revenue. So we have done a SWOT analysis of these 6 segments and tried to find out which segments have got strength, which has the opportunity, which are weak and which are the threats.

9.3.

SWOT Analysis:

After doing the SWOT analysis we have categorized the segments as per Strength, Weakness, Opportunity and Threat.

i)

Strength:
High revenue and low risk (HL). High revenue and medium risk (HM).

The two segments which are the strength are

Properties of HL and HM: High Revenue generators. Least risky.

56 | P a g e

 Long term Loyal customers. Cross sell options. Very few customers delinquent. Defining Characteristics This Segment is an area of Strength. Total population - 95,604 The Segment contributes 52% of the total revenue, which is the highest in the portfolio These customers have higher credit limit (50% of the total portfolio) but low credit utilization levels (10%) 89% of customers make international transactions Customers in this segment pay 77% of the total fees as annual fees which is higher than the portfolio High vintage - 12yrs 94% of the total Infinite cards and 47% of the total Platinum cards are present in this segment The customers are high on CA transaction as this segment contributes 56% of the total CA Highest Merchant spends for HL and HM at: Travel, Super Market, Services, Departmental store and Telecom

Opportunity: 56% of the total cash advance is contributed by customers in this bucket. These are cash needy customers. Hence there is an opportunity for cross selling loans products. High percentages of customers make international transactions (89%). Hence travel linked promotions and product features may find more takers. Challenges The credit limit Utilization is low in the segment though the credit limit available is higher than the portfolio.

Recommendations:

These are your best customers, guard them. Run Loyalty programs to retain their spends Upgrade cards of Low risky classic card customers Increase the credit limit of the customers having high credit limit utilization, low credit limits & low risk Run premium club membership programs for these customers Offer Premium and Value Added services to these high revenue generating segments Widen the revenue profile from current interchange biased scenario Trigger based post purchase installment for high purchase transactions would help in exploiting the revenue potential.

57 | P a g e

ii)

Weakness:

There is only one section which can be weakness. Low revenue and high risk (LH). Properties of LH: Least Revenue generators. Highly risky. Expected future revenue very low. Customers with low vintage. High cash advance customers. Defining Characteristics This Segment is an area of Weakness. Total population - 54,274 The Segment contributes 3% of the total revenue, which is the lowest in the portfolio. These customers have the lowest credit limit in the portfolio and an average 6% credit limit utilization. Only 38% of customers make international transactions with an average spend of 1,940 pesos. Customers in this segment pay 46% of the total fees as late fees which is higher than the portfolio. These customers are the youngest on book with an average vintage of 2 years. 95% of the customers are classic card holders the remaining have gold card. Out of the total revenue 31% is contributed by Fees and only 7% by interchange revenue. In these segment only 22% customers does CA spend with an average of 1,011 pesos. Highest Merchant spends for LL and ML at: Super Market, Travel, Departmental store, Telecom and Clothing.

Opportunity: Reduce the promotions offered in this segment as this is the least profitable and highly risky segment

Challenges The risk is highest in this segment and there are high probability of customers becoming delinquent

Recommendations:

Reduce the credit limit of this bucket Promotions if offered should be for small ticket purchases Increase the % interest on revolve balance

58 | P a g e

iii)

Opportunity:
Low revenue and low risk (LL). Medium revenue and low risk (ML).

There are two segments which we have found the opportunity segments.

Properties of LL and ML: Low Revenue generators Least risky Low transacting customers Cross sell options Opportunity to induce spending Very few customers delinquent Defining Characteristics This Segment is an area of Opportunity. Total population - 83,973 The Segment contributes 8% of the total revenue 92% of customers make international transactions. Customers in this segment pay 77% of the total fees as annual fees which is higher than the portfolio These customers are on books on an average 9 yrs 19% of the total gold cards is present in this segment Out of the total revenue 30% is contributed by Interchange revenue This segment customers are low on CA spend as only 15% of the total CA is contributed from here Highest Merchant spends for LL and ML at: Super Market, Travel, Departmental store ,Services, and Entertainment

Opportunity These customers are low on transactions and have low ticket size (Avg. ticket size 949 pesos). Opportunity is to make them transact more by providing offers for their top MCC so that they migrate to segments like HL

Challenges The revenue equivalent is low i.e. they are generating much lower revenue as compared to portfolio. But these customers are low on risk These customers are low on revolve balance as only 4% of the total is contributed from this segment

59 | P a g e

Recommendations:

Introduce usage-linked promotional schemes so that customers use their card more often Provide incentives for international transactions since spend per international transaction is high Low on revenue, these customers need to be induced to revolve on balance Effective credit line management programs Revenue Maximization: Cash Advance Promotions for competition CA Users Maximize revenue generation by converting high ticket purchases to installments

iv)

Threat:

There is only one segment which can be threat to generating revenue. High revenue and high risk (HH). Properties of LH: High Revenue generators Highly risky Late fees a big component Balance at risk highest High cost on investment Defining Characteristics This Segment is an area of Threat. Total population - 38,289 This segment contributes 21% of the total revenue These customers have higher credit limit but low credit utilization levels (only 6%) 61% of customers make international transactions. Customers in this segment pay 26% of the total fees as late fees These customers are on books from past 7yrs which is lower than the portfolio average of 10 years 37% of the customers in this segment either hold a gold or a platinum card The customers in this segment have the highest POS ticket segment in the portfolio with 1,844 pesos Highest Merchant spends for HL and HM at: Travel, Super Market, Services, Telecom and Departmental store

Challenges Cash and Revolve cannot be further encouraged because of poor repayment track records High percentage of 90 DPD and above customers. Hence there seem to be recovery issues associated with this segment

60 | P a g e

Opportunity These customers are already spending on high ticket items. Opportunity is to make them use their card more often on small purchases

Recommendations:
Introduce usage-linked incentives to increase frequency of card usage. Introduce risk-based differential-APRs for customers in order to compensate for risk. Cross Sell loan products to the customers in this segment. Recovery score card for late stage recovery of 90 DPD and above delinquent customers.

61 | P a g e

10. Final Program:

i)

Recommendation

for

Different

Promotional

Card upgrade program:

Assumptions: HL and HM are considered for Card Upgrade Upgrade is run only on Classic cards to Gold Card Top 10% customer based on their spend are selected Customers are filtered out if they have reached 3 DPD status The segments have been upgraded from Classic card to Gold card and Platinum card to Infinite card It is assumed that at least 50% of the customers who are upgraded will use their cards similar to other gold card and Infinite card customers Card upgrade is done only on 25% of the Portfolio Classic and Platinum card are considered for upgraded to Gold and Infinite respectively The HL and HM segments are the better segment to run Card upgraded program. If we run card upgraded program for these two segments it may lead to incremental revenue of 11 million pesos.

Tag

HL HM Total

Top 25% customer based on spend 6506 8580

Average Average Average revenue from revenue from increment classic cards gold cards 8199 10956 9151 11783 952 827

Total increment in revenue 6192996 7093944 13286940 6643470

Total increase in revenue (Assuming 50% of the customers follow the gold card trend)

Tag

HM HL Total

2450743 5607232 8057975 Total increase in revenue (Assuming 50% of the customers follow the Infinite card 4028987
trend)

Top 25% customer based on spend 1018 1110

Average revenue from platinum cards 6720 8390

Average Average revenue from increment infinite cards 9127 13442 2407 5052

Total increment in revenue

Total Revenue

10672458

62 | P a g e

ii)

Credit Line Increase (CLI) Program

Assumptions: LL ,ML,HL,HM are considered for CLI Amongst these, only accounts with credit limit utilization greater than 25% qualify Customers are filtered out if they have reached 3 DPD status The segments have been given a 20% increase in credit limit It is assumed that at least 50% of the customers will have the same credit limit utilization (currently it is > 25%) It is assumed that the revolve balance is 45% (Revolve ratio of the portfolio) Interchange revenue is taken as 1.7% Credit Limit Increase is done only on 10% of the customers It is assumed that the customers will have the same credit limit utilization

The four segments LL, ML, HL, HM have been considered for Credit limit increase program. If bank run this credit limit increase program for the above mentioned segments could lead to a incremental revenue of 53 million pesos. Tag No.of custo mer Amount spend before credit increase 209,641 Amount spend after credit increase 251,569 Increa se in spend* per custo mer 41,928 Interest on revolve per custome r 4,245 Intercha nge revenue per custome r 356 Total increase in revenue per customer 4,602 Total increase in revenue

HL

4,619

21,254,849 9,208,918 15,724,601 7,018,653 53,207,022

HM 2,483 168,965 ML 6,874 104,216 LL 7,263 44,025 Total increase in revenue

202,758 33,793 3,422 287 3,709 125,060 20,843 2,110 177 2,288 52,830 8,805 892 75 966 (Assuming 50% of the customers follow the trend)

63 | P a g e

iii)

MCC promotions program:

Assumptions: LL,ML,LM,MM segments are considered for MCC Promotions The Top 3 MCC by Spends has been considered for offers Customers are filtered out if they have reached 3 DPD status It is assumed that the MCC specific promotions will induce a 50% increase in spend for 10% of the customers The revolve balance is taken as 45% (Revolve ratio of the portfolio) and Interchange revenue is taken as 1.7% MCC promotion is done only on 10% of the customers It is assumed that the revolve balance is 45% (Revolve ratio of the portfolio)

Here four segments LL, ML LM, MM have been considered for MCC promotions program. This MCC promotion program for these four segments may lead to revenue of near about 15 million pesos.

Total Revenue- 144,299,121

64 | P a g e

11.

References

www.fractalanalytics.com. www.sas.com. Materials Provided by Fractal Analytics Ltd.

65 | P a g e