Collections Scoring

APDS Consulting UK Ltd

www.apds-analytics.com
 Problem Statement
• Credit Customers are obliged to make regular
(usually) monthly payments to repay their lending
facilities Increasing delinquency – Missed Payments
• Customers are categorised as sitting in a range of
buckets, representing the number of missed
payments
• If a payment is missed or not made customers will
‘roll’ to the next delinquency bucket
• If a payment is made the customer will ‘roll back’ the U
number of buckets the payments represents T X 60 90
30 Losses

• At 90 days past due the customer is deemed to have D

‘defaulted’, and a loss may be incurred
• NPLs occur at 90 days past due, Banks will aim to
avoid a 90-day status

Recovery – Making Payments

What is Collections Scoring?

• Collections Scoring targets accounts that are delinquent with the aim
to control the roll rate from bucket to bucket
• Outcomes are used to prioritise the order in which customers are
contacted when delinquent
• Calling Strategies
• Letter Strategies
• Recovery or Payment Projection Models aim to predict the amount of
collected post default
• Early Debt Sale
• Recoveries Strategies
• Input into Loss Given Default (LGD)
Benefits of Collections Scoring
Resource
Prioritisation Reduced
Focus on Roll
(best Default Rates Reduced
Rates to
collectors on and lower NPLs
Current
hardest ECLs
cases)

Collection
Champion
Case Advanced
Challenger
Increased Prioritisation Analytically
Strategies
Revenue (hard core Strategies
(constant
delq, most
evolution)
attention)

Lower Losses, Higher Revenue – Improved ECL

How to use Behavioural & Collections Scoring?
Exposure
Portfolio Data Management

Cross-Sell
Behavioural&
Collections B-Score
Scoring Chars Modelling
(Examples) Triage

• Max Delq last 6 Exposure

months Management
• Numbers times 30+
last 12 months
• Number of Overlimit
e
+ x
Strategy
Pr( x) =  + x
last 3 months
• Number of Payments Segmentation

1+ e 
greater than
minimum payment
last 6 months Collections
• Max Utilisation last 9 Priorisation
months
• Number of PTPs Roll-Rate
Taken Last 3 Months Management
• Number of PTPs Kept
last 12 months
C-Score
• Proportion of PTPs
Kept last 12 months
Payment
Projection

Recoveries
Management
Customer Analytics What is it?

• Behavioural & Collections Scoring is used to

predict which accounts or customers will go
into the late stages of delinquency (often 3
cycles plus, although more complex
High Risk performance definition can be used on a
product by product basis)
High Collections
Day 1 Intensive Contact • Often the score is grouped in to risk grades
Risk Strategy / Early and appropriate actions taken by grade
Debt Sale
Why

• Identify Future Delinquents, reducing losses

• Identifying cases that will self-cure will allow
Medium Standard
Tailored Use of resources to be concentrated where they are
Day 7 Collections
Resources needed
Risk Process • Identifying x-sell potential will increase revenue

How?

• A behavioural score is often an integral part of

Low Soft Touch the bank’s customer management
infrastructure and can be used to determine x-
Collections
Risk Day 21 Process
sell opportunities (from a revenue perspective)
New Facility offers • The score is also key to identifying customers
that are deteriorating so that preventative
measures can be taken to pre-cure before
delinquency
• The score would also be used in the early
stages of delinquency to prioritise strategies
Customer Management Analytics – Collections Management
GDS Link Asia
Account Status

1-29 Days 30 - 90 Days 90+ Days Past

Up To Date Past Due Due
Past Due

Behavioural Scoring Roll-Rate Model Payment Projection

Description Manage Customers Initiate Customer Continued Customer Decision made to end
Relationships Contact for Contact for the customer
through • Debt Control • Prioritised Debt relationship,
• Pro-active limit Control therefore switch to
• Triage
management • Arrears • Recovery
• Limit Management Strategies
• Competitive Pricing Management
• Pre-emptive Triage • Save the • Debt Sale

• X-Sell Programmes Relationship? • Litigation

Objective To maximise To minimise roll-rates To maximise recovery

To maximise profitability by
profitability by to non-performing cash-flows and
balancing expected bucket whilst minimise losses to the
balancing expected revenue with
revenue with determining whether bank
associated levels of to save the
associated levels of risk and exposure,
risk and exposure relationship or
whilst catching early minimise loss
delinquency
Modelling – Model Types

Fixed Performance Window

Static Model Final Delq Status Performance
Definition
Data Utilised
• Demographics
• Standard Behavioural
Score Chars
• Contact History
Rolling Window • Promises to Pay (Keep /
(for data volume) Broken)
Roll Rate Performance Definition based on • Collections Experience
Model Improved / Worsened Delq • Payment History /
Frequency
• Delinquency Status
and Movements

Segmented Rolling or Fixed Performance Window

by Delq Final Delq Status Performance
Bucket Definition
Performance Definitions

Model Goods Bads Indeterminates

Static Final Delq Status 30+ dpd X-Days
Current
Rolling Improvement over Worse than Start Same as Start
Start Position Position Position
Segmented X-Days Current 30+ dpd X-days dpd
Segmented 30-Days Current, X-Days 60+ dpd 30-days dpd
Segmented 60-Days Current, X-Days 60+ dpd 30-days dpd
Segmented 90+ Payment Projection Continuous
 Examples of Data that could be used when
Collections Scoring

Potential • Collections operations Behavioural Scoring

Data
Mobile

Collections
are highly data • Worst Status Last Month /
• Pre-Paid Y/N

• Number of device used to

dependent L3M / L6M / L12M
pay for goods L3M, L6M
• Min Balance Last Month,
• Data used will be • Time of first daily use

Modelling
L3M, L6M, L12M

generated from • Number of Payments L1M,

L3M, L6M, L12M
• Average Distance between
daytime and nighttime
account operation, • Average Payment to
location last week

from contacts with the

Datasources
Balance Ratio L3M

customer within the

collections /
delinquent Collections Contact
Data
Open Banking

environment • Number of right party

• Total Number of
Accounts in a Delinquent
contact last month state
• The use of external • Total Outstanding
• Number of Promises to
data, such as open Pay Taken L1m, L3M, L6M Balance

banking and other • Number of Promises to • Our Outstanding as a

proportion of Total
alternative sources Pay Kept last month, Last
6m etc. Outstanding Last Months
would be encouraged • Number of Partial • Total Late Fees Paid Last

as it provides a much Payments Made L3M,

L6M
3 months

• No. Months with

more holistic view • Ratio of Payment to delinquency last 3M,
Outstanding L3m L6M, L12m

Model Power increases as data availability expands

Customer Management Analytics - Sampling
GDS Link Asia
Independent Variables Performance Variables
Data
Elements Transactional
Data Bureau Data,
can be at e.g. No. of
e.g. Delq
account or L3M
accounts
customer
level OR be Internal Data,
Contact Performance
File
History, One Record per
external e.g.
e.g. £ of Good-Bad Flag
Collections Account / Customer
bureau type History
PTPs
based upon
taken
chars Delq Status

3 to 6 Months
Observation Period Outcome Period
(usually up to 12 months) (dependent upon model purpose)

Know Info @ Outcome Outcome is

Decision shorter for
Point Observation Colls
Models
Point
Linear Regression Vs Logistic Regression (Equations)
GDS Link Asia
General form of Linear Regression General form of Logistic Regression
Yˆj =  +   j x j
 +  x
j e Linear
Pr( x) =  +  x
Regression
Where: 1+ e
Y: dependent variable
 : general intercept
: co-efficient applied to the explanatory variable The output of the regression model is a probability from 0 to 1
x: explanatory variable

In Scoring:
Y: total score
Other Form of Logistic Regression
 : constant
: co-efficient applied to the characteristics
 Pr( x) 
x: explanatory variable (e.g. age, income, sex) g ( x) = ln   =  +  x
350 [Y] = 200 [ ] + 50 x Age30-40 + 40 x Income50k+ + 60 x 1 − Pr( x) 
SexFemale

www.apds-analytics.com
Champion-Challenger Strategies
Default Rates by Strategy
Strategy A across 12 months post
(Call Day 5, Letter Day 10) implementation

Comparison of Strategy A vs Strategy B

5.8
5.6
5.4
5.2
Models 5
4.8
4.6
4.4
4.2
4
0 2 4 6 8 10 12

Strat A Strat B

Strategy B
(Call Day 1, Day 5, Letter Day 20)
 Collections Scoring

APDS Consulting UK Ltd

www.apds-analytics.com