How to Use Data Mining with CRM in Banking to Segment Credit Card Customers

Credit cards are being used by consumers across age groups and for diverse purposes. They buy different products and services according to their purchasing power, habits, standard of living and life stage. The frequency of purchases and value of each transaction also varies, customers use credit cards for their utility bills, apparel, daily needs and occasionally for high value purchases. With e-commerce growing in popularity, having a credit card swiped instead of paying cash is increasing in popularity. This new trend is most evident for the 25-45 year age group, it is used for air fares, buying publications, booking movie tickets, etc.

Customers vary in terms of their payment behavior, there are some who tend to pay the full due amount, while others only pay the minimum amount and carry forward their previous balance.

It becomes very critical for banks to segment customers and target the profitable ones, while weeding out the low profit customers. A pattern can also be developed to identify customers with increasingly good credit worthiness and encourage them to increase their credit card usage.

Building the data set
The data required is available in the bank's CRM system and core system's transactional data marts. The recorded data covers various aspects of usage including number and amount of purchases, merchant details, demographics like age and gender and monthly balances. Banks can start by categorizing customers with respect to the merchants from whom they purchase and this will generate information related to the type of products/services. Again, each credit card customer can have more than one card including a primary account and add-on cards. This needs to be taken into account when generating a unified view of the customer and accordingly design the marketing strategies.

Data for the last six months can be an ideal data set for segmentation, it is large enough to take into account inconsistent usage patterns and recent enough to avoid outdated behaviours.

Formation of clusters
The following criteria can be used to develop a clustering model:

  • Monthly Frequency
  • Monthly average Value
  • Average value per transaction
  • Percentage of maximum limit used
  • Type of products/services
  • Age group
  • Marital status
  • Cash Advance
  • Payment to minimum payment ratio and statement balance ratio

The above would result in formation of clusters based on demographic profiles like age, marital status, buying behaviour, cluster size and contribution to overall purchase.

Cost-benefit analysis to prioritize actions
Depending on results from the above cluster formation exercise, a profitability analysis can be carried out for various clusters. This includes %defaulters, late payment fee revenue, patterns of migration from a less profitable to a more profitable cluster and patterns to identify how a person purchasing one category of product/services is likely to start purchasing other products and services.

Clusters where purchases can be increased easily without affecting credit erosion and thus profitability, should be targeted first. Patterns in behaviour of loss making customers should be identified and corrective actions should be taken.

Targeted campaigns for each segment
It can be ascertained that most good customers use credit cards for specific purposes. If this is the case, a campaign to increase their purchases for other products and services should be launched to motivate them. Alliances can be formed with various merchants to offer discounts for these products and services to specific groups of customers. Reward programs and incentives should be tailored as per their specific needs. Brochures sent with each bill cycle should be tailored to arouse interest. For example, a 'travelers segment' should be given discount for air tickets with related offers like hotel bookings, city tours, etc. A person spending on movies should be give offers for concerts, restaurants, DVDs, etc.

Segmentation can also be used to develop new products, partnerships and reward programs. These products/ services can be offered at point-of-sales where a customer uses cash for these products and encourages them to make their next purchase with a credit card. Special cards with pre loaded credit can be offered to professionals and employees to make them start using their cards. Effects of these campaigns should be analysed to improve and use the same strategy for new customers and help them migrate from a low profit to high profit segment. Candidates for product shifts and add-on cards can be identified that match purchasing habit patterns of existing clusters.

Data mining through a banking CRM system can be a critical source for improving profitability from the credit card business. It can also effectively identify trends for cross selling with results from the segmentation exercise of other products. All data related to various segments and their related reports can be hosted on a single centralized system for analysis by numerous associated teams around the globe.