Largest Scale : ICICI Bank : Pre Delinquency Management

Last year's economic slowdown had an adverse effect on everyone. Customers of
credit cards and loans from ICICI Bank were also impacted by the then prevailing
economic landscape. This resulted in a customer's inability to pay the dues to
the bank, which in turn directly influenced the performance of retail assets
business of the bank. Since the customer's income was stressed from various
payment commitments, the bank had to ensure that the dues towards the bank get
paid on priority. Especially, the unsecured product dues required a proactive
approach of contacting the customers. If a customer is not able to pay the dues
for a period, he ends up becoming delinquent and the bank's losses increase. The
economic scenario had led the Bank to focus towards two challenges: first, to
recover the maximum amount of outstanding from delinquent customers and second,
to minimize future delinquency. The bank already had a debt service management
function that takes care of the recovery of non-paid dues from delinquent
customers. To minimize delinquency it was required to identify the stressed
customers at an early stage. This in turn created the need of data driven
categorization of customers on basis of their probability to default. The bank
intended to initiate a comprehensive pre-delinquency management program (PDM).
The aim of this pre-delinquency program was to minimize losses arising out of
currently non-delinquent customers.

The process of identification of probable customers to default was done in
the data warehouse where customer's history was mapped against certain
parameters and behavioral patterns. The customers who are identified by the
system as probable to default are also evaluated on a risk scale score and this
whole process is automated. The score acted as an indicator of the customer's
propensity to default in near future. Once a customer has been identified, the
PDM system itself generates offers that can be presented to the customer and
also finds out the complete contact details of the customer from various
direct-indirect channels, IVR system, internet banking, POD from courier etc.
The system can map a customer across all relationships with the bank and produce
contact details. A credit card customer could also have a savings account with
the bank, the system can identify a customer and his relationships across
different services with the bank and then find out all contact points to get in
touch with him. Considering that ICICI Bank has over 3 million customers under
credit cards and loan products, and an overall customer base of 20 million for
their retail products, it can be envisaged how complex the system would be to
not only predict customers who would default based on their history but also to
map the customer across all relationships with the bank and know his complete
contact details.

The entire process of identification of the customer for PDM is automated.
The PDM generated offer is directly uploaded in the calling system database
which is used by the auto dialer to assign the calls. The technology platform
used for PDM includes Sybase IQ as Enterprise Data warehouse, SAS suite as
Enterprise Reporting Tool and SAS e-miner and Text Miner for analytics, Business
Objects for enhanced reporting, etc.

With PDM implementation the roll forward rate of customers reduced from 14%
to 6 % resulting in credit loss savings. For instance, if the roll rate for year
2008-09 was 100, then post deployment the roll rate for year 2009-10 came down
to 55. Due to process automation also the bank witnessed cost reduction by
manpower savings of more than 80 man hours per month. The Bank's customers also
benefited as they were given flexible offers of paying back in time of economic
stress.