Can collections ever
be a profit centre for
financial institutions?
Leveraging Analytics to build collection strategies to
gain competitive advantage
Page | 1
Post 2008 financial crisis, higher risk management has become the new normal at the
financial institutions in the US. Banks have started to employ new methods to
strengthen credit underwriting to screen faulty customers right at the origination
stage. Collections and recovery industry has also undergone significant changes.
Lenders have started contacting delinquent customers early in the collections life cycle
with the aim of recovering maximum in less time. Lenders are implementing
aggressive strategies to be the first recipient of the delinquent customers’ payment.
They have also increasingly started focusing on maximizing their ROI using smart
business analytics to create optimum strategies and strengthen underwriting.
Introduction
Banks primarily rely on a single risk variable such as credit bureau score or account balance
to prioritize delinquent customers. Delinquent customers with a lower credit score are
contacted early to recover more and customers with a higher credit score are given some
time to self-cure. Some banks also consider delinquency amount as a variable to prioritize
delinquent customers. Using this single risk variable approach, the banks are only considering
the probability of recovery. Probability of recovery from customers with high credit score is
more than from customers with low credit score. However, banks should ideally consider both
the probability of recovery and the expected dollar amount that can be recovered. Strategy
creation on two variables will help a bank identify the ideal collection strategy to implement
such that recovery time is low while the recovery amount is higher.
Lenders can use 2-Variable or 3-Variable matrix to prioritize delinquent accounts. Predictive Analytics
based scorecard will allow banks to identify accounts which needs to prioritized to target early while
giving time to low risk accounts to self-cure. In the above figure, model score is the internal score
computed by the bank.
Credit Score
ModelScore
Highest Priority
Priority
Lowest Priority
Priority
Delinquency Amount
ModelScore
H
L
H
L H
L
Page | 2
Creating Dynamic and Predictive Collections Model
Single risk variable based strategy used for prioritization of delinquent accounts or credit
approval/rejection decisioning can be biased as the lender’s risk can be different for
customers with same credit score. Consider two borrowers, borrower A and B, both with a
FICO score of 680. However, they have different risk profiles thus posing different risk levels
to the lender.
FICO score is computed on the basis of several factors relating to a borrower’s payment
history: Number of past and current credit accounts, utilization of credit card accounts, past
delinquencies, severity of past delinquencies, public records, etc. FICO’s research show that
a single 30-day delinquency can reduce a borrower’s FICO score by 90-100 points. Similarly,
continuous low utilization of credit card accounts could improve a borrower’s score
significantly over a long credit period. So how do Borrower A and B here have same FICO score
despite having different risk profiles?
Borrower A: Positive impact on borrower A’s FICO score has been created by a long credit
history along with continuous low utilization of credit card accounts. While long history of late
credit card payments and past delinquencies has impacted borrower A’s FICO score
negatively.
Borrower B: A seven year long credit history with zero late credit card payments has created
a positive impact on borrower B’s FICO score. While one 30-day delinquency reported on
borrower B’s profile along with continuous high utilization of credit card accounts has made
a negative impact on borrower B’s FICO score.
Such positive and negative impacts could bring many borrower’s FICO score in close vicinity
to each other (despite each borrower having a different risk profile). Can a bank then rely only
on credit bureau score to define risk profiles of borrowers? From collections perspective, can
Delinquency History
Borrower Profile
Borrower-Lender Relationship
Borrower A (FICO score: 680) Borrower B (FICO score: 680)
• Long history of late credit card payments
• Delinquent on two payments prior to
this assessment & self-cured both times
• Zero late credit card payment
• One 30-day delinquency prior to this
assessment
• Successful businessman in a highly
seasonal industry
• Stable job at a well-known multi-national
company
• Association with several other lenders in
past due to delinquent payments
• Associated with this bank for last 10
years for all deposit and credit accounts
Credit History
• A fifteen-year credit history
• Low utilization of credit card accounts
(her balances are 15-20% of her limits)
• A seven-year credit history
• High utilization of credit card accounts
(her balances are 40-50% of her limits)
Page | 3
a bank then rely only on credit bureau score to identify which delinquent customers to
prioritize in collection efforts?
Banks and Credit Unions have traditionally relied upon single risk variables such as account
balances and credit scores to determine the risk profile of a customer and decide collections
strategy for customer with different risk profile. But to accurately identify high risk accounts
and prioritize them with an effective collections strategy, lenders should use predictive
analytics where both probability of recovery and expected dollar amount of the recovery are
considered for decision making. This whitepaper is aimed at understanding how predictive
analytics can bring a paradigm change to effective collection strategy making at every stage
of the collections life cycle.
Having a dynamic analytics model for every stage of Collections life cycle
For any financial institution, any collection strategy is based on two parameters: Consumer
Risk to the lender and Customer satisfaction which leads to customer retention. A poor
collection strategy would be to target a low risk customer with hard collections tactics leading
to lower customer retention or to de-prioritize a high risk customer leading to lower recovery.
Using Analytics, financial institutions can create effective collection strategies unique to every
stage of collections life cycle to boost retention ratio and recovery rate.
Collection life cycle depicts the entire movement of a
delinquent customer from the point he/she enters into
delinquency to stage where he/she declares bankruptcy
leading to write-off of the loan by the bank. The entire
life cycle can be divided into three parts based on Days
Past Due (DPD) to actual payment: Early stage (0-30
DPD), Late stage (31-60 DPD) and Recovery stage (61-
90 DPD).
Behavioural Scoring for Early Stage Collections
At early stage into the collections life cycle (0-30 DPD), there are two main aim of the bank:
1. Identify low risk customers and give them time to self-cure to retain them.
2. Identify high risk customer early in the collections life cycle and recover more from
them.
As a first step, financial institutions can use a Behavioural Scoring Model to identify low and
high risk customers. The behaviour score can be computed using various parameters and
assigning weights to these parameters. Further risk ranges can be defined on these behaviour
scores and risk level can be identified for each customer.
Early
Stage
Late Stage
Recovery
Stage
Page | 4
Behaviour scoring to create risk levels
Parameter Weight
Payment History 30%
Current DPD 40%
Borrower Profile 30%
Total 100%
Based on the behavioural profiling of the customers, collection strategies can be created using
two risk variables. The second variable can be delinquency amount, delinquency amount to
loan amount ratio, current DPD, etc. A heat map showing high risk and low risk cases should
be prepared using these two risk variables. Banks can effectively segregate delinquent
customers by assigning them collection priorities. Based on the prioritization, a decision
matrix should be created to define strategies to target each priority type.
Communication strategies can be defined for each priority and automatic actions can be
assigned to different stakeholders/vendors.
Priority Action
Priority 1 Immediate Call to customer asking for reason of late payment and reminding
the customer to make the payment immediately
Priority 2 Customer Reminder using SMS reminding the customer of the late payment
Priority 3 No Action. Wait for the customer to self-cure
FIG: Heat Map using two risk
variables to prioritize delinquent
customers
Page | 5
Using Hard Collections Strategies for Late stage Collections
In Late Collections Stage (31-60 DPD), lender should identify those customers who are likely
to move to the next stage of the collections life cycle i.e., identify those customers who have
a higher probability of defaulting even after 60 days and moving to recovery stage. Remaining
customers will be identified as low risk customers who are likely to “normalize” and move
back to early collections stage (0-30 DPD).
Using Analytics, lenders can do a profiling of the customer to come up with a “Collection
score” of the customer. Collection score can be based on variety of parameters such as
Behaviour score, number of broken promises, number of kept promises, number of contacts
made to the customer, number of non-technical cheque bounces etc. This collection score
can be used with delinquency amount to create a 2-Variable heat map or delinquency amount
and credit score to create a 3-Variable heat map. (See figure below)
2 - Variable Heat Map 3 - Variable Heat Map
Page | 6
Communication strategies can be defined for each type of customer and automatic actions
can be assigned to different stakeholders/vendors:
Type 2-Variable map 3-Variable map Action
Type-1 High Collection Score –
High Delinquency
Amount
High Collection Score –
High Delinquency Amount
– Low Credit Score
High Likeliness to move to next
collections stage. Send Legal Notice.
Type-2 Moderate Collection
Score – Moderate
Delinquency Amount
Moderate Collection Score
– Moderate Delinquency
Amount – Moderate
Credit Score
Moderate Likeliness to move to next
collections stage. Send warning of Legal
action with a reminder to pay.
Type-3 Low Collection Score –
Moderate Delinquency
Amount
Low Collection Score –
Moderate Delinquency
Amount – Moderate
Credit Score
Low Likeliness to move to next stage
collections stage. Place an immediate call
reminding to make payment
immediately.
Type-4 Low Collection Score –
Low Delinquency
Amount
Low Collection Score –
Low Delinquency Amount
– High Credit Score
Likely to “normalize”. Send an SMS
reminding to make payment
immediately.
Targeting customers with higher recovery probability in Recovery Collections
Stage
In the Recovery collections stage (more than 90 days DPD), lenders need to identify defaulted
customers who have higher probability of recovery and target them through effective
channels. Lenders also need to identify those defaulted customers who have lower probability
of recovery and accept loss on such cases to save recovery collections effort.
Based on historical behaviour score data, lender can bucket the customers using their
behaviour score or credit score to assign them probability of recovery. For ex., all delinquent
customers can be divided into 6 buckets based on behaviour score assigning them
probabilities from 0% to 90% as follows:
Behaviour Score (out of 100) Probability of Recovery
0 - 15 0%
16 - 30 15%
31 - 45 30%
46 - 60 45%
61 - 75 60%
76 - 90 75%
> 90 90%
Further the lender needs to identify customers from whom recovery can be expected and
customer from whom no recovery can be expected. This can be done using several methods:
 Lender can identify certain data points in customers’ profile. Based on historical data
of customer profiling, the algorithm can determine which data points will show as
Page | 7
positive to indicate full or partial recovery. If some of these data points are positive in
a customer, then recovery is expected from that customer.
 Lender can also analyse historical Loss Given Default (LGD) ratio for a particular type
of customer profile. Similar customer profiles can be assumed to give low LGD and
hence recovery can be expected from these accounts.
Based on combination of the probability of recovery and expected recovery (Full/Partial or
NIL), accounts with higher expected recovery amount can be prioritized to maximize recovery.
Different communication strategies can defined for prioritized accounts:
 Assign to cases to collections agencies.
 Use in-house collections team for hard collection strategies (legal notices).
 Sell-off the debt.
Amount to
be recovered
Recovery
(Expected/NIL)
Probability
of Recovery
Expected
Recovery
Amount
Suggested Action
$10000 Recovery
Expected
90% $9000 High Priority. Assign case to collection
agency for tracing and follow-up.
$4000 Recovery
Expected
50% $2000 Medium Priority. Send Hard Collections
Letter with a reminder to make the
payment.
$20000 NIL Recovery 20% $0 Low Priority. Sell-off the debt.
Road Ahead: Leverage Analytics to build effective collections strategies
Predictive analytics does not just improve a lender’s collection efficiency, but increases
profitability as the lender is able to recover more dollar amount from delinquent customers
and retain self-cure customers who may bring new business to the bank. It is imperative here
that the bank considers a 360-degree view of the customer. A high risk HNI customer having
multiple accounts with the bank will have much lower LGD than a medium risk customer
having a single loan account with the bank. Automated loan management systems used along
with predictive analytics enabled collections solution can help the bank drive profitability and
growth while scoring more on customer satisfaction.
Page | 8
Arup Das
Lending Product Head (P&L Management), Nucleus Software
Arup is the Vice President and Lending Product Head (P&L Management)
at Nucleus Software where he is responsible to lead the flagship product
to the next level of global leadership. Before joining Nucleus, he has
played various roles in strategy and product management with leading
companies like CISCO, IPValue and Mphasis.
Author e-mail id: arup.das@nucleussoftware.com
Jayant Tondon
Senior Lending Product Manager, Nucleus Software
Jayant is the Senior Product Manager at Nucleus Software where he is
responsible for managing P&L for FinnOne Neo for North America and SEA
region. He has 17 years of experience in core banking domain and banking
product management such as Retail Banking Operations, P&L
Management, Audit, Risk and Regulatory Management, Business Analysis
and Product Implementation. Before joining Nucleus, he has worked with
leading banks like ICICI and IDBI.
Author e-mail id: jayant.tondon@nucleussoftware.com
Dev Surti
Lending Product Analyst, Nucleus Software
Dev is a Product Analyst at Nucleus Software where he is responsible for
managing P&L for FinnOne Neo for North America. He is an MBA from
Narsee Monjee Institute of Management Studies (NMIMS), Mumbai,
specializing in Finance. Previously, he has interned at DCM Shriram Ltd. in
areas of corporate finance and strategy.
Author e-mail id: dev.surti@nucleussoftware.com

More Related Content

PDF
Intro to Key Performance Indicators
PPT
Making Your Call Center the Heart of Your Strategy
PPTX
AI in Talent Acquisition
PDF
Sales Training PowerPoint Presentation Slides
PDF
Call Center Life 101 v3
PDF
Employee Hiring Process PowerPoint Presentation Slides
PPTX
AI on HR
PDF
Recruiting Metrics - Strategic and Tactical KPIs for Talent Acquisition
Intro to Key Performance Indicators
Making Your Call Center the Heart of Your Strategy
AI in Talent Acquisition
Sales Training PowerPoint Presentation Slides
Call Center Life 101 v3
Employee Hiring Process PowerPoint Presentation Slides
AI on HR
Recruiting Metrics - Strategic and Tactical KPIs for Talent Acquisition

What's hot (20)

PPTX
Recruitment Strategy.pptx
PPTX
key performance indicator
PDF
Deloitte - Supply Chain Strategy
PPTX
Digital customer experience
PPTX
People Analytics is the New HR
PPTX
Service Process Management and Quality Improvement
PDF
AI in HR by Farhan - 2025
PDF
Predictive Analytics in Retail - Visual Infographic Report
PDF
Hr and Talent Analytics
PPTX
Talent Acquisition Best Practices & Trends 2014
PDF
PPT
The Onboarding and Training Playbook for Big Customers
PPTX
Analytics and AI based Retention in e-commerce
PDF
Procurement and spend analytics
PDF
Digital HR - Trends in Digital HR and the Future of Work
PDF
RFM Analysis for Customer Segmentation
PDF
Key Performance Indicators-KPIs
PPTX
QBR Presentation Template
PDF
How to win on the customer experience battleground; where businesses are won ...
PDF
HR Business Partnership: Do our business partners really hear us?
Recruitment Strategy.pptx
key performance indicator
Deloitte - Supply Chain Strategy
Digital customer experience
People Analytics is the New HR
Service Process Management and Quality Improvement
AI in HR by Farhan - 2025
Predictive Analytics in Retail - Visual Infographic Report
Hr and Talent Analytics
Talent Acquisition Best Practices & Trends 2014
The Onboarding and Training Playbook for Big Customers
Analytics and AI based Retention in e-commerce
Procurement and spend analytics
Digital HR - Trends in Digital HR and the Future of Work
RFM Analysis for Customer Segmentation
Key Performance Indicators-KPIs
QBR Presentation Template
How to win on the customer experience battleground; where businesses are won ...
HR Business Partnership: Do our business partners really hear us?
Ad

Similar to Whitepaper - Leveraging Analytics to build collection strategies (20)

PPTX
Transform Your Credit and Collections with Predictive Analytics
 
PDF
Credit Risk Analytics
PPTX
Tomas Denemark
PPT
Canadian Banking Basics - Risk management
PDF
Financial Management : Receivables Management
PPT
Net30 Score Sample Validation Analysis
PPTX
Unit 4 Reveivables Management.pptx
PPTX
Apanps5210 - final presentation
PPTX
Receivables management
PPTX
Credit Risk Model Building Steps
PDF
Tech's role in debt collections
PDF
Automation and Analytics: Two Levers to Revitalize Retail Debt Recovery
PDF
Blog 2016 13 - impairment modeling in retail - many moving parts
PDF
Retail Banking 6 Steps to Improving the Collections Experience.pdf
PDF
The Science And Art Of Profitable Underserved Consumer Portfolios
PPTX
retailing-credit card
PDF
Retail Banking 6 Steps to Improving the Collections Experience.pdf
PDF
Forecasting peer-to-peer lending risk
PPT
Ac Sjzh92177
PDF
Credit Risk Management
Transform Your Credit and Collections with Predictive Analytics
 
Credit Risk Analytics
Tomas Denemark
Canadian Banking Basics - Risk management
Financial Management : Receivables Management
Net30 Score Sample Validation Analysis
Unit 4 Reveivables Management.pptx
Apanps5210 - final presentation
Receivables management
Credit Risk Model Building Steps
Tech's role in debt collections
Automation and Analytics: Two Levers to Revitalize Retail Debt Recovery
Blog 2016 13 - impairment modeling in retail - many moving parts
Retail Banking 6 Steps to Improving the Collections Experience.pdf
The Science And Art Of Profitable Underserved Consumer Portfolios
retailing-credit card
Retail Banking 6 Steps to Improving the Collections Experience.pdf
Forecasting peer-to-peer lending risk
Ac Sjzh92177
Credit Risk Management
Ad

More from Arup Das (15)

PDF
How CIOs should make Cloud investment - InfotechLead
PDF
Is Risk Management today only about Managing Risk_
PDF
predictive-analytics-the-silver-bullet-in-efficient-risk-management-for-banks
PDF
Why IT Needs More Analytics Talent - Tech News _ Latest Technology News
PDF
Whitepaper - Technological Innovation for Credit Unions and Microfinance Inst...
PDF
Whitepaper - Asset Based Financing - Gaining Popularity in the Middle East
PDF
Whitepaper_E_Customer centricity the survival strategy for Japanese lenders
PDF
Blog_E_Consumer lending in Japan
PDF
Whitepaper - Transformational journey of Regional Development Banks in South-...
PDF
Whitepaper_From Middle Line to Top Line
PDF
Blog-Technology – A Panacea To Mutuals’ Challenges
PDF
Blog-Embryonic Opportunities For Predictive Analytics In Australia
PDF
Blog-Cloud Computing
PDF
FICCE-White-Paper
PDF
Nucleus-Microsoft Cloud Alliance Announcement
How CIOs should make Cloud investment - InfotechLead
Is Risk Management today only about Managing Risk_
predictive-analytics-the-silver-bullet-in-efficient-risk-management-for-banks
Why IT Needs More Analytics Talent - Tech News _ Latest Technology News
Whitepaper - Technological Innovation for Credit Unions and Microfinance Inst...
Whitepaper - Asset Based Financing - Gaining Popularity in the Middle East
Whitepaper_E_Customer centricity the survival strategy for Japanese lenders
Blog_E_Consumer lending in Japan
Whitepaper - Transformational journey of Regional Development Banks in South-...
Whitepaper_From Middle Line to Top Line
Blog-Technology – A Panacea To Mutuals’ Challenges
Blog-Embryonic Opportunities For Predictive Analytics In Australia
Blog-Cloud Computing
FICCE-White-Paper
Nucleus-Microsoft Cloud Alliance Announcement

Whitepaper - Leveraging Analytics to build collection strategies

  • 1. Can collections ever be a profit centre for financial institutions? Leveraging Analytics to build collection strategies to gain competitive advantage
  • 2. Page | 1 Post 2008 financial crisis, higher risk management has become the new normal at the financial institutions in the US. Banks have started to employ new methods to strengthen credit underwriting to screen faulty customers right at the origination stage. Collections and recovery industry has also undergone significant changes. Lenders have started contacting delinquent customers early in the collections life cycle with the aim of recovering maximum in less time. Lenders are implementing aggressive strategies to be the first recipient of the delinquent customers’ payment. They have also increasingly started focusing on maximizing their ROI using smart business analytics to create optimum strategies and strengthen underwriting. Introduction Banks primarily rely on a single risk variable such as credit bureau score or account balance to prioritize delinquent customers. Delinquent customers with a lower credit score are contacted early to recover more and customers with a higher credit score are given some time to self-cure. Some banks also consider delinquency amount as a variable to prioritize delinquent customers. Using this single risk variable approach, the banks are only considering the probability of recovery. Probability of recovery from customers with high credit score is more than from customers with low credit score. However, banks should ideally consider both the probability of recovery and the expected dollar amount that can be recovered. Strategy creation on two variables will help a bank identify the ideal collection strategy to implement such that recovery time is low while the recovery amount is higher. Lenders can use 2-Variable or 3-Variable matrix to prioritize delinquent accounts. Predictive Analytics based scorecard will allow banks to identify accounts which needs to prioritized to target early while giving time to low risk accounts to self-cure. In the above figure, model score is the internal score computed by the bank. Credit Score ModelScore Highest Priority Priority Lowest Priority Priority Delinquency Amount ModelScore H L H L H L
  • 3. Page | 2 Creating Dynamic and Predictive Collections Model Single risk variable based strategy used for prioritization of delinquent accounts or credit approval/rejection decisioning can be biased as the lender’s risk can be different for customers with same credit score. Consider two borrowers, borrower A and B, both with a FICO score of 680. However, they have different risk profiles thus posing different risk levels to the lender. FICO score is computed on the basis of several factors relating to a borrower’s payment history: Number of past and current credit accounts, utilization of credit card accounts, past delinquencies, severity of past delinquencies, public records, etc. FICO’s research show that a single 30-day delinquency can reduce a borrower’s FICO score by 90-100 points. Similarly, continuous low utilization of credit card accounts could improve a borrower’s score significantly over a long credit period. So how do Borrower A and B here have same FICO score despite having different risk profiles? Borrower A: Positive impact on borrower A’s FICO score has been created by a long credit history along with continuous low utilization of credit card accounts. While long history of late credit card payments and past delinquencies has impacted borrower A’s FICO score negatively. Borrower B: A seven year long credit history with zero late credit card payments has created a positive impact on borrower B’s FICO score. While one 30-day delinquency reported on borrower B’s profile along with continuous high utilization of credit card accounts has made a negative impact on borrower B’s FICO score. Such positive and negative impacts could bring many borrower’s FICO score in close vicinity to each other (despite each borrower having a different risk profile). Can a bank then rely only on credit bureau score to define risk profiles of borrowers? From collections perspective, can Delinquency History Borrower Profile Borrower-Lender Relationship Borrower A (FICO score: 680) Borrower B (FICO score: 680) • Long history of late credit card payments • Delinquent on two payments prior to this assessment & self-cured both times • Zero late credit card payment • One 30-day delinquency prior to this assessment • Successful businessman in a highly seasonal industry • Stable job at a well-known multi-national company • Association with several other lenders in past due to delinquent payments • Associated with this bank for last 10 years for all deposit and credit accounts Credit History • A fifteen-year credit history • Low utilization of credit card accounts (her balances are 15-20% of her limits) • A seven-year credit history • High utilization of credit card accounts (her balances are 40-50% of her limits)
  • 4. Page | 3 a bank then rely only on credit bureau score to identify which delinquent customers to prioritize in collection efforts? Banks and Credit Unions have traditionally relied upon single risk variables such as account balances and credit scores to determine the risk profile of a customer and decide collections strategy for customer with different risk profile. But to accurately identify high risk accounts and prioritize them with an effective collections strategy, lenders should use predictive analytics where both probability of recovery and expected dollar amount of the recovery are considered for decision making. This whitepaper is aimed at understanding how predictive analytics can bring a paradigm change to effective collection strategy making at every stage of the collections life cycle. Having a dynamic analytics model for every stage of Collections life cycle For any financial institution, any collection strategy is based on two parameters: Consumer Risk to the lender and Customer satisfaction which leads to customer retention. A poor collection strategy would be to target a low risk customer with hard collections tactics leading to lower customer retention or to de-prioritize a high risk customer leading to lower recovery. Using Analytics, financial institutions can create effective collection strategies unique to every stage of collections life cycle to boost retention ratio and recovery rate. Collection life cycle depicts the entire movement of a delinquent customer from the point he/she enters into delinquency to stage where he/she declares bankruptcy leading to write-off of the loan by the bank. The entire life cycle can be divided into three parts based on Days Past Due (DPD) to actual payment: Early stage (0-30 DPD), Late stage (31-60 DPD) and Recovery stage (61- 90 DPD). Behavioural Scoring for Early Stage Collections At early stage into the collections life cycle (0-30 DPD), there are two main aim of the bank: 1. Identify low risk customers and give them time to self-cure to retain them. 2. Identify high risk customer early in the collections life cycle and recover more from them. As a first step, financial institutions can use a Behavioural Scoring Model to identify low and high risk customers. The behaviour score can be computed using various parameters and assigning weights to these parameters. Further risk ranges can be defined on these behaviour scores and risk level can be identified for each customer. Early Stage Late Stage Recovery Stage
  • 5. Page | 4 Behaviour scoring to create risk levels Parameter Weight Payment History 30% Current DPD 40% Borrower Profile 30% Total 100% Based on the behavioural profiling of the customers, collection strategies can be created using two risk variables. The second variable can be delinquency amount, delinquency amount to loan amount ratio, current DPD, etc. A heat map showing high risk and low risk cases should be prepared using these two risk variables. Banks can effectively segregate delinquent customers by assigning them collection priorities. Based on the prioritization, a decision matrix should be created to define strategies to target each priority type. Communication strategies can be defined for each priority and automatic actions can be assigned to different stakeholders/vendors. Priority Action Priority 1 Immediate Call to customer asking for reason of late payment and reminding the customer to make the payment immediately Priority 2 Customer Reminder using SMS reminding the customer of the late payment Priority 3 No Action. Wait for the customer to self-cure FIG: Heat Map using two risk variables to prioritize delinquent customers
  • 6. Page | 5 Using Hard Collections Strategies for Late stage Collections In Late Collections Stage (31-60 DPD), lender should identify those customers who are likely to move to the next stage of the collections life cycle i.e., identify those customers who have a higher probability of defaulting even after 60 days and moving to recovery stage. Remaining customers will be identified as low risk customers who are likely to “normalize” and move back to early collections stage (0-30 DPD). Using Analytics, lenders can do a profiling of the customer to come up with a “Collection score” of the customer. Collection score can be based on variety of parameters such as Behaviour score, number of broken promises, number of kept promises, number of contacts made to the customer, number of non-technical cheque bounces etc. This collection score can be used with delinquency amount to create a 2-Variable heat map or delinquency amount and credit score to create a 3-Variable heat map. (See figure below) 2 - Variable Heat Map 3 - Variable Heat Map
  • 7. Page | 6 Communication strategies can be defined for each type of customer and automatic actions can be assigned to different stakeholders/vendors: Type 2-Variable map 3-Variable map Action Type-1 High Collection Score – High Delinquency Amount High Collection Score – High Delinquency Amount – Low Credit Score High Likeliness to move to next collections stage. Send Legal Notice. Type-2 Moderate Collection Score – Moderate Delinquency Amount Moderate Collection Score – Moderate Delinquency Amount – Moderate Credit Score Moderate Likeliness to move to next collections stage. Send warning of Legal action with a reminder to pay. Type-3 Low Collection Score – Moderate Delinquency Amount Low Collection Score – Moderate Delinquency Amount – Moderate Credit Score Low Likeliness to move to next stage collections stage. Place an immediate call reminding to make payment immediately. Type-4 Low Collection Score – Low Delinquency Amount Low Collection Score – Low Delinquency Amount – High Credit Score Likely to “normalize”. Send an SMS reminding to make payment immediately. Targeting customers with higher recovery probability in Recovery Collections Stage In the Recovery collections stage (more than 90 days DPD), lenders need to identify defaulted customers who have higher probability of recovery and target them through effective channels. Lenders also need to identify those defaulted customers who have lower probability of recovery and accept loss on such cases to save recovery collections effort. Based on historical behaviour score data, lender can bucket the customers using their behaviour score or credit score to assign them probability of recovery. For ex., all delinquent customers can be divided into 6 buckets based on behaviour score assigning them probabilities from 0% to 90% as follows: Behaviour Score (out of 100) Probability of Recovery 0 - 15 0% 16 - 30 15% 31 - 45 30% 46 - 60 45% 61 - 75 60% 76 - 90 75% > 90 90% Further the lender needs to identify customers from whom recovery can be expected and customer from whom no recovery can be expected. This can be done using several methods:  Lender can identify certain data points in customers’ profile. Based on historical data of customer profiling, the algorithm can determine which data points will show as
  • 8. Page | 7 positive to indicate full or partial recovery. If some of these data points are positive in a customer, then recovery is expected from that customer.  Lender can also analyse historical Loss Given Default (LGD) ratio for a particular type of customer profile. Similar customer profiles can be assumed to give low LGD and hence recovery can be expected from these accounts. Based on combination of the probability of recovery and expected recovery (Full/Partial or NIL), accounts with higher expected recovery amount can be prioritized to maximize recovery. Different communication strategies can defined for prioritized accounts:  Assign to cases to collections agencies.  Use in-house collections team for hard collection strategies (legal notices).  Sell-off the debt. Amount to be recovered Recovery (Expected/NIL) Probability of Recovery Expected Recovery Amount Suggested Action $10000 Recovery Expected 90% $9000 High Priority. Assign case to collection agency for tracing and follow-up. $4000 Recovery Expected 50% $2000 Medium Priority. Send Hard Collections Letter with a reminder to make the payment. $20000 NIL Recovery 20% $0 Low Priority. Sell-off the debt. Road Ahead: Leverage Analytics to build effective collections strategies Predictive analytics does not just improve a lender’s collection efficiency, but increases profitability as the lender is able to recover more dollar amount from delinquent customers and retain self-cure customers who may bring new business to the bank. It is imperative here that the bank considers a 360-degree view of the customer. A high risk HNI customer having multiple accounts with the bank will have much lower LGD than a medium risk customer having a single loan account with the bank. Automated loan management systems used along with predictive analytics enabled collections solution can help the bank drive profitability and growth while scoring more on customer satisfaction.
  • 9. Page | 8 Arup Das Lending Product Head (P&L Management), Nucleus Software Arup is the Vice President and Lending Product Head (P&L Management) at Nucleus Software where he is responsible to lead the flagship product to the next level of global leadership. Before joining Nucleus, he has played various roles in strategy and product management with leading companies like CISCO, IPValue and Mphasis. Author e-mail id: arup.das@nucleussoftware.com Jayant Tondon Senior Lending Product Manager, Nucleus Software Jayant is the Senior Product Manager at Nucleus Software where he is responsible for managing P&L for FinnOne Neo for North America and SEA region. He has 17 years of experience in core banking domain and banking product management such as Retail Banking Operations, P&L Management, Audit, Risk and Regulatory Management, Business Analysis and Product Implementation. Before joining Nucleus, he has worked with leading banks like ICICI and IDBI. Author e-mail id: jayant.tondon@nucleussoftware.com Dev Surti Lending Product Analyst, Nucleus Software Dev is a Product Analyst at Nucleus Software where he is responsible for managing P&L for FinnOne Neo for North America. He is an MBA from Narsee Monjee Institute of Management Studies (NMIMS), Mumbai, specializing in Finance. Previously, he has interned at DCM Shriram Ltd. in areas of corporate finance and strategy. Author e-mail id: dev.surti@nucleussoftware.com