London: Elsevier. The data consists of 8950 rows and 18 columns. However, the higher the credit card limit, the higher the annual fee you have to pay. QIWare (Quick Insights Ware) is an agile analytics solution that delivers a complete suite of capabilities to support an end-to-end data mining cycle. be adopted across a range of consumer finance products and markets. Credit cards arent banking theyre information Richard Fairbank, Founder of Capital One. This model has served the industry well for decades, enabling it to offer three main card typesrewards, low-rate and subprimeto cater to different users. To do this, we must first specify the number of clusters K. Here I used the elbow method to specify the best K. Elbow is a very simple method that gives us plots like elbow shape. The main aim of this algorithm is to minimize the sum of distances between the data point and their corresponding clusters. The challenge for issuers is aligning the right value proposition with the right consumers. Guaranteeing long-term satisfaction and loyalty by increasing relevance in communications and offers. It is, therefore, imperative to make use of any automation tools and techniques possible at this step in order to have enough time and resources dedicated to more value-added steps. The worst value is -1. machado Step 3. Visit my github to see the full notebook. CREDIT_LIMIT and MINIMUM_PAYMENT content some null values. The sample Dataset summarizes the usage behavior of about 9000 active credit cardholders during the last 6 months. The question is: how do issuers develop a suite of credit cards that fulfil customers needs more precisely without piling on features that add needless complexity or are not valued by users? The credit limit ranges from 10 million to 40 million IDR, depending on the credit card issuing bank. I applied Principal Component Analysis (PCA) to transform data into 2 dimensions for visualization because we wont be able to visualize the data in 17 dimensions. I handled these missing values by replacing them by means. We are building the next-gen data science ecosystem https://www.analyticsvidhya.com, Data Analyst at PT Gistex Garmen Indonesia, Determining Physicians Drug Combination Preferences for Regions in Africa Using the Apriori, Dimensionality Reduction: Zero to Hero (PartI). The user enters the inputs of all three models and produces forecasts and returns the same after routing those results. A score of 1 denotes the best meaning that the data point is very compact within the cluster to which it belongs and far away from the other clusters. From the credit card business perspective, this view provides an understanding of cannibalization effects, true customer potential, and opportunities for cross-sales of card products to non-cardholders. Keep up the good work! And the highest silhouette score is in k = 3. In the early days, most companies had only 1 segmentation model that was based on customer demographics, value, behavior, needs, or a mix of these. There are no more null values. Segmentation is critical because a company has limited resources, and must focus on how to best identify and serve its customers. I would like to thank my project lead Anuj Saini for constant support throughout the internship. Platinum credit cards are only owned by a few people because it is not easy to get the card due to strict procedures. Credit Cards ReadyWare addresses the most common customer value management challenges in cards business from segment management to value retention and growth. The algorithm takes the unlabeled dataset as input, divides the dataset into k-number of clusters, and repeats the process until it does not find the best clusters. (LogOut/ Normalization is a technique often applied as part of data preparation for machine learning. Just a fortnight ago, I along with my team completed a Customer Credit Card Segmentation project as a Data Science Intern with Packt. Today, leading organizations usually maintain and manage many segmentation models rather than only one up to 10-20 across the enterprise. As a result, apart from the traditional customer demographics and value segmentations, credit card data requires customer behavior to be analyzed from many perspectives. However, there are some behavioral micro-segmentation models that can address the requirements of most card providers. This gives businesses the ability to tailor marketing messages and timing to generate better response rates and provide improved consumer experiences. Market segmentation: how to do it, how to profit from it. Behaviours are often based on credit bureau reports on how a person spends and pays over time; customers are typically categorised as transactors, revolvers or subprime. Within credit card marketing, customer segmentation can be used across the board for customer value and lifecycle management, including but not limited to: Customer segmentation also acts as a building block for predictive analytics, as well as for campaign management and monitoring activities, providing a granular view of different customer profiles that have the potential to act and react differently. Segmentation in marketing is a technique used to divide customers or other entities into groups based on attributes such as behaviour or demographics. The machine learning model in the backend was deployed using the Flask API. comerford In addition to delivering finer details on specific customer behavior, these models also provide vast targeting opportunities when crossed with each other, e.g., targeting fashionistas who are also weekend shoppers for a weekend fashion show. For machine learning, every dataset does not require normalization. Card issuers can not only make higher-priced proposals, but can discover groups that have poorly serviced by present offers, using improved segmentation. Step 5. This step includes importing needed packages and dataset, checking data summary, handling missing values, checking data types, and selecting the features. Such automation also minimizes the risk of human error in this process, which is the primary root cause for failures in modeling. To understand better about each feature of the data means, heres the data dictionary. It allows us to cluster the data into different groups and is a convenient way to discover the categories of groups in the unlabeled dataset on its own without the need for any training. Following models are obtained by performing some of following operations: Model 1 is generated by certain operations on data as follows, Model 2 is generated by certain operations on data as follows. If you wish to work on similar industry-based data science project, contact Aditya Shah from Packt. Cluster 2 : This customer group indicates a small group of customers who have high balances and cash advances, low purchase frequency with high credit limit. Last, but not least, an understanding of opportunities and risks from all the analyses should be translated into business strategies and actions. The performance of the K-means clustering algorithm depends upon highly efficient clusters that it forms. Mine the Data: Using tools and algorithms available (such as K-Means, SOM, and Kohonen), alternative segmentation models should be developed. who is an active customer, a new customer, or even a customer) to set a proper scope and direction. Values near 0 denote overlapping clusters. Customer segmentation, which is similar to other customer analytics activities, is a cyclical process that requires continuous management and fine-tuning in order to adapt to changes in data sources, business models, and customer portfolios. Every market and customer portfolio is different due to similarities in the nature of the transaction and statement data as well as in common product characteristics. (LogOut/ It is used to manage HTTP requests and the API function for the data to be obtained and shown to the end-user. interactions. It is required only when features have different ranges. It is classified as a microframework since no specific tools or libraries are required. Customer segmentation is one of the most fundamental building blocks in getting to know customers. A client segmentation model allows consumers to efficiently differentiate themselves and provides issuers with insights into how their cards may be used more or attract new customers via innovative systems. Ever wondered, if you don't know about the customers you have and what purchase characteristics they follow using credit cards, how difficult it would be for your business to grow? In addition to analyzing credit card behavior, it is also vital to have a comprehensive behavioral view of customers at the bank level, analyzing asset, credit, and other product behavior. It is a centroid-based algorithm, where each cluster is associated with a centroid. The result is a race to attract new accounts. Demographics are derived from census reports and other non-financial databases and cover facts such as income, age Why clustering techniques are relevant in the world of data science. An example credit card micro-segmentation model could group customers based on their preferences for shopping time, resulting in segments such as weekend shoppers and late-nighters, whereas another model could look into lifestyle perspective and identify segments such as fashionistas and tech-savvy shoppers. There are many outliers (look at the max value), but I didnt drop them because they may contain important information, so I treated the outliers as extreme values. Nice article, Jay Shah! The most important but often-ignored step at this stage is incorporation of business inputs from end-users of segments into data mining activities; these can range from collecting initial hypotheses on possible segments to prioritizing specific variables and manually selecting some of the segments. Itis an Unsupervised Learning algorithm, which groups the unlabeled dataset into different clusters. Now I can successfully find a pattern in business models and statistics and try to improve the sales in an efficient way so as to bring profit to the business. Create a free website or blog at WordPress.com. and geography. Differentiating or segmenting these consumers based on their requirements, behaviors, and attitudes, as well as understanding the reason why customers use their credit cards by considering a limited number of financial behaviors, assists us in overcoming this problem. (2004). Issuers do have an alternative: they can maintain a profit margin and generate demand for their product by providing only the benefits that customers value. The answer lies in a more nuanced and powerful approach to customer segmentation one that can, by extension, For real this time. Customer segmentation is the technique of separating a customer base into particular groups of persons. However, it was hard to find the elbow point of the curve, so decided to use silhouette score. This is why the micro-segmentation approach is relevant for the credit card industry and frequently used across the world. For cluster 0, I recommended a silver credit card because its the most widely owned card. ReadyWare is based on Forte Consultancy Groups cards analytics practices across leading providers around the world and, yet, is 100% flexible for customizations.

Natural Language Queries. Cluster 1 : This customer group indicates a large group of customers who have medium balances, spenders (high purchase) with the highest credit limit. We can assume that this customer segment uses their credit cards as a loan. The output is a list of customers, each tagged with segment flags using a segmentation model. I used the K-means algorithm with the K value determined by silhouette score. In general, a new credit cardholder will receive a silver card and they can upgrade it later. WCSS stands for Within Cluster Sum of Squares, which defines the total variations within a cluster. The Elbow method is one of the most popular ways to find the optimal number of clusters. It is essential for industries where customer interaction is frequent and varied, as each interaction provides insight into opportunities and risks for every individual. This micro-segmentation approach looks at customers profiles from many different perspectives and groups them under different categories according to each perspective. We studied the patterns of the graph and the information provided in excel minutely to understand and implement so as to create a model which will not only depict you present state through statistics but also provide you points to improve your sales and profit. Some issuers have offered as much as $400 to customers signing up for a new card, and the top five US issuers are spending over $100m a year on advertising campaigns. document.getElementById( "ak_js_1" ).setAttribute( "value", ( new Date() ).getTime() ); Enter your email address to follow the Forte Consultancy blog and receive notifications of new posts by email. Set the Scope: Every segmentation study should start with a definition of business expectations from the outputs and use cases as well as basic customer definitions (i.e. For Cluster A, we found following pattern in out data, For Cluster B, we found following pattern in out data. Surprisingly, all around the world, there are many banks still not taking advantage of this opportunity, frequently due to a lack of tools or in-house resources to process and digest big data sources, sitting on top of a gold mine that erodes each day when left untouched. Change), You are commenting using your Facebook account. The complexity of these models can range from a simple set of business rules such as customers who havent used their cards last year to sophisticated data mining codes. [3] Bruce Cooil , Lerzan Aksoy & Timothy L. Keiningham (2008) Approaches to Customer Segmentation, Journal of Relationship Marketing, 6:34, 939, Analytics Vidhya is a community of Analytics and Data Science professionals. It is useful to identify segments of customers who may respond in a similar way to specific marketing techniques such as email subject lines or display advertisements. Eight of these frequently used models that utilize the basic transaction and statement data to full extent are listed below. Understand the Data: Before going into the data mining step, it is critical for the analysts to gain a deep understanding of the data, in terms of availability, quality, and distribution, through preliminary analysis. QIWare simplifies data preparation and modeling while increasing business productivity, minimizing human error, and significantly improving usability. Here the elbow method comes into light. The cardholder must have a monthly salary of at least 3 million IDR. Our dataset consists of a column tenure, which represents the customer using our service for some period. For banks across the world, credit cards are a significant source of revenue! Fill in your details below or click an icon to log in: You are commenting using your WordPress.com account. After getting the clusters, I interpreted them in the visualization using FacetGrid. When we plotted a customer's tenure against the frequency of cash advances, we discovered that all of our old clients had a lower cash advance frequency, whereas new customers had a higher cash advance frequency. Clustering is one of the most common exploratory data analysis techniques used to get an intuition about the structure of the data. While learning and implementing certain algorithms, we concluded and decided to use K-Means Clustering Algorithm to create clusters of customers based on certain characteristics. Credit cards arent banking theyre information, Click to share on LinkedIn (Opens in new window), Click to share on Twitter (Opens in new window), Click to share on Facebook (Opens in new window), Click to email a link to a friend (Opens in new window), Click to share on Reddit (Opens in new window), Five New Opportunities in Credit Card Analytics, Effective Channel Management Strategies Segmenting the Channels, The Under-Tapped Banking Consumer Segments of the World The Youth, Channel Migration Strategies Matching Customers to the Optimum Channels, Stopping Churn in Its Tracks Proactive Retention Strategies for Mobile Operators, Grading Performance The Automotive Industry BI Maturity Map. Step 4. But first, what is customer segmentation? segmentation. For each of the models that are integrated with the backend using Flask Framework, we incorporated three models with user inputs and predict buttons. Surprisingly, all around the world, there are many banks still not taking advantage of this opportunity, frequently due to a lack of tools or in-house resources to process and digest big data sources, sitting on top of a gold mine that erodes each day when left untouched. Step 2. The advantage of this type of card is the limit is large enough. PCA transforms a large set of variables into a smaller one that still contains most of the information in the large set. Their behavioral style can help study these groups and provide better alternatives and strategies to meet their demands. The value of k should be predetermined in this algorithm. While each of micro-segmentation models can be used solely for these purposes, pairing and crossing them with each other provides endless opportunities for targeting and getting to the bottom of what makes each customer tick. (You can download PDF version of this whitepaper here.). I created a model that estimates credit card customer segmentation to help the company to define its marketing strategy. QIWare helps companies in: With ReadyWare, which was built on top of QIWare, card providers almost instantly obtain a complete view of their customers portfolios, e.g., which of their customers are big-ticket weekend purchasers who have tech-savvy lifestyles and more than a 70% likelihood of becoming silent over the next 3 months so that the card providers can retain them with their exclusive discount on latest handset hitting the markets next month. Further on visualizing the data, we discovered that the majority of our clients had been with us for at least 12 years, implying that they are loyal to our service and have faith in our business, which benefits us indirectly. You can use it to repay big-budget items such as motorbikes or smartphones. Then, I assigned 3 to the KMeans model. It is essential for industries where customer interaction is frequent and varied, as each interaction provides insight into opportunities and risks for every individual. Herekdefines the number of predefined clusters that need to be created. Segment migration analysis should then become a continuous monitoring tool, observing change with each scoring cycle. https://github.com/TeamEpicProjects/Credit-Card-Customer-Segmentation-KJSIEIT, BALANCE : Balance amount left in customers account to make purchases, BALANCE_FREQUENCY : How frequently the Balance is updated, score between 0 and 1, PURCHASES : Amount of purchases made from account, PURCHASES_FREQUENCY : How frequently the Purchases are being made, score between 0 and 1, ONEOFF_PURCHASES : Maximum purchase amount done in one-go, ONEOFFPURCHASESFREQUENCY : How frequently Purchases are happening in one-go, INSTALLMENTS_PURCHASES : Amount of purchase done in installment, CASH_ADVANCE : Cash in advance given by the user, PURCHASESINSTALLMENTSFREQUENCY : How frequently purchases in installments are being done, CASHADVANCEFREQUENCY : How frequently the cash in advance being paid, CASHADVANCETRX : Number of Transactions made with "Cash in Advance", PURCHASES_TRX : Number of purchase transactions made, CREDIT_LIMIT : Limit of Credit Card for user, PAYMENTS : Amount of Payment done by user, MINIMUM_PAYMENTS : Minimum amount of payments made by user, PRCFULLPAYMENT : Percent of full payment paid by user, TENURE : Tenure of credit card service for user. On visualizing them, we found a linear relationship between purchase and one-off purchase. As a consequence, this is usually the most time-consuming step, taking up to 70-80% of the time. Here I used the K-means algorithm. Harvard Business Review, Sept.-Oct.: 113124. In the end, a backend call routed through the Flask framework helps to provide the end-user with valuable insights and recommendations after providing user inputs. The silhouette method can calculate the silhouette coefficient and easily find the exact number of K. The value of the silhouette coefcient is between -1 and 1. Reducing the number of variables of data. Data source : Credit Card Dataset for Clustering. However, with the dramatic decline in acquisitions over the past five years, issuers competing in similar segments with similar products are finding it hard to differentiate themselves: The total number of accounts at top issuers has declined by an average of 4% over the past five years according to The Nilson Report. The goal of normalization is to change the values of numeric columns in the dataset to a common scale, without distorting differences in the ranges of values. And we can easily guess the optimal number of K from the plot. K-means algorithm is an iterative algorithm that tries to partition the dataset into K pre-defined distinct non-overlapping subgroups (clusters) where each data point belongs to only one group. It means the best number of clusters, in this case, is 3. Segmentation is an integral part of the development of marketing objectives and strategies, where defining those objectives will generally include either[1,2] : (a) an analysis of how products should be sold or developed, based on an analysis of current customer segments. The file is at a customer level with 18 behavioral variables. (LogOut/ Credit card providers usually target users using their behavior and demographic information. Prepare the Data: Customer segmentation modeling, when done at the micro level, requires preparation of hundreds, and in some cases thousands, of variables describing each and every customer. out the cards in the coming months. Heres the summary of the data. This case requires to develop of a customer segmentation to define marketing strategy. The cardholder must have an income of at least 180 million IDR per year and have a good credit history. Profile and Utilize: Once segments are identified, they should be profiled for a more detailed understanding of the customer composition, loyalty, and risk within each segment, looking into various factors such as demographics, attrition likelihood, default risk, and cross-product utilization. An example of a credit card micro-segmentation model could group customers based on their preferences for shopping The cardholder must have a regular monthly income of around 5 million to 10 million IDR. Institutions have long sought closer connections with customers, but have struggled with limited data and arms-length A platinum credit card has a high limit from 40 million up to 1 billion IDR. Silver cards have the lowest credit limit, which is around 4 million to 7 million IDR. Change), You are commenting using your Twitter account. Issuers find it difficult to differentiate between their customers based on their behavioral patterns and payment or spending patterns. Now the question arises on how to calculate the value ofk. GitHub repository: https://github.com/TeamEpicProjects/Credit-Card-Customer-Segmentation-KJSIEIT, Live project: https://cccs-kjsieit.herokuapp.com/, Technical theory reference: K-Means Clustering Algorithm. While these models served early marketing needs well, as the sophistication of products and the number of customers is increasing, more comprehensive approaches are being sought after and are emerging such as micro-segmentation and segment of one. 300 billion credit card transactions are expected to take place each year by 2018, creating 300 billion opportunities to understand customers better. (1957). Credit card data are rich in terms of both volume and variety, providing insights into both customer shopping and payment preferences. But as switching incentives rise, profit margins inevitably fall. Change). We have used web development frameworks such as HTML or CSS to build an interactive and seamless UI for the final user to submit his/her inputs and obtain model insights.

Strategies for diversification. [2] McDonald, M. & Dunbar, I. Credit card customer segmentation is now essential for a profitable credit card customer portfolio. But choosing the optimal number of clusters is a big task. Credit card issuers have traditionally targeted consumers by using information about their behaviours and demographics. Using Elbow method, we determined optimal number of clusters for 2 models as follows: The above models, we can conclude that 3 & 4 will be the optimal value for k. Considering k value as 3 and 4, we get the following cluster labelled Cluster A and Cluster B respectively. The credit card industry is on par with telecommunications, e-commerce, and retail from this perspective, and the industry gains significant ROI from segmentation initiatives. I also used PCA for dimension reduction and better visualization. Author Identification, Verification And Profiling Using Stylometry. GlobalData research: the journey to a cashless world, Hyundai Card blazes innovative PLCC trail in Korea, Covid-19 has altered the way businesses handle payments, Amex survey, Country reports: Estonia, the Netherlands and UAE, South Korean prosecutors raid offices tied to Luna amidst ongoing crypto crash, Omnispace Launches Spark-2 Satellite into Orbit, SAVRpak Launches Moisture Control Technology to Increase Shelf-Life of Fresh Produce, Graforce Develops Plasma Electrolysis Technology for Green Hydrogen Production. Another model could look into lifestyle perspective and identify segments such as fashionistas and tech-savvy shoppers. Each of these cycles follows these key steps: Step 1. A new twist on needs-based segmentation attempting to develop a deeper and more rounded view of consumers is nothing new. Behaviors are often based on credit reports on how a person spends and pays over time. Unfortunately, many banks remain ignorant of this wealth of information at their disposal, and they opt for mass marketing and costly above-the-line communications. Many companies fall into the trap of skipping this step and jumping into data mining with both feet, and those companies ultimately end up duplicating their work or finding themselves with outputs that are irrelevant. Customer segmentation is one of the most fundamental building blocks in getting to know customers. As segmentation results depend on both the inputs and the algorithms used, it is important to test alternative combinations to determine a statistically solid and business-wise meaningful set of segments. This approach has already been used successfully for two recent credit card launches in the US, as well as one in Brazil, where a large issuer identified three distinct customer segments and shaped an integrated approach to the design of new products, messages and channels; the bank expects to see new accounts grow by 2 to 5% as it rolls The advantage of this card is the limit that is not too high. Finally, I visualized the clusters in a scatter plot. Last, for cluster 2, I recommended a platinum credit card with the highest level. Cluster 0 : This customer group indicates a small group of customers who have low balances, small spenders (low purchase) with the lowest credit limit. time, resulting in segments such as weekend shoppers and late-night spenders. This analysis not only provides insights into what to expect during modeling, but it also provides opportunities to discover anomalies, skewed distributions, and so on, which should be addressed for reliable and stable segments. Mohamed Dabo reports on a strategy that helps to uncover customer behaviour in credit card use.

However, the two-way communications opened up by online and mobile channels enable todays issuers to capture much more information about each customer. This method uses the concept of WCSS value. It can be defined as the task of identifying subgroups in the data such that data points in the same subgroup (cluster) are very similar while data points in different clusters are very different. To view or add a comment, sign in The credit card industry is on par with telecommunications, e-commerce, and retail from this perspective, and the industry gains significant ROI from segmentation initiatives. Increasing the share of wallet and becoming the primary card via identification of high-potential cardholders, Encouraging cross-sales of secondary cards and other banking products based on comprehensive customer understanding, Selectively upgrading card limits and tiers for maximum return on risk, Retaining the most valuable customers by accurately evaluating customer value and attrition risk, Maximizing profitability from payment operations, and migrating customers to more profitable payment products and interaction channels.