Data modelling and predictive analytics (Part one)
Tuesday September 03, 2019
Data modelling and predictive analytics – more
important than ever before? Part
one.
The global credit environment has adjusted to the financial crisis, to
compliance regulations and to expanding markets, including
the growing non-prime sector. However, regardless of how stressed the
market is or the current risk appetite, leading institutions can benefit from
the power of analytics and machine learning to transform collections operations
and generate value. This article will comprise two sections. The first covers the risk
climate and how scoring and modelling can mitigate risk. The second, how
analytics can improve client relationships with a multi-channel communications
strategy.
Client scoring for better risk assessment and
segmentation
Credit scoring has been the industry standard
for evaluating creditworthiness. However, advances in machine learning and
predictive analytics can take this process a step further – creating a 360-degree
profile of a customer that includes alternative data such as social media,
spending patterns, online transactions, payment history and more to
determine financial behavior and thus, provide credit based
on a wider pool of variables (instead of using standardized data and simplified models to make blanket
decisions).
Data-based insights (in line with the latest
regulations and compliance frameworks) allows clients to improve approval rates
to a wider array of consumers such as non-prime customers, such those needing
temporary finance who might not qualify for credit due to variable income,
foreign investors who don’t currently have a credit record in the country or recent
graduates. By taking into account a wider range of data, the accuracy of
prediction can continuously improve and be refined to allow a very
individualized profile. According to McKinsey, the idea behind this is to categorize every customer in a segment of one, allowing
better customization.
Transformation of collections models allow
lenders to move away from decision making based on the standard stages of
delinquency or simple risk scores. Better modelling and analytics-based segmentation
allows early identification of customers likely to cure or pay.
Propensity to cure or
pay
Each
client account can be evaluated for its likelihood to cure or pay. This helps lenders
categorize clients as self-curers or non-self-curers and present strategies for
collection based on the classification. The ability to predict the success of
collection operations and evaluate outcomes each month before the next billing
cycle begins, allows lenders to redirect their collections efforts from clients
likely to self-cure to clients not likely to meet their obligations.
Data inputs can include several variables such as demographics,
overdraft, transactions, contracts and collaterals. Similarly, if an account is
showing a continuous downward trend for client motivation or ability to pay, it
should be treated as a potential risk long before it becomes one. Data
modelling allows lenders to determine payment patterns and support preventative
mechanisms such as alerts that are triggered when any variation from payment
schedules occurs. The system itself can send out automated responses to ask the
client if they need support.
According
to McKinsey, the most sophisticated lenders are creating ‘synthetic
variables’ from the raw data to further enrich the data.
Machine learning helps identify markers for high-risk accounts from such
variables such as cash-flow status, ownership of banking products, collections
history and banking and investment balances. By using so many inputs from
different systems, lenders can improve the accuracy of their models, decrease
charge-off losses and increase recoveries.
To find out why 4 out of the top 5 Canadian FIs use SCORE models to manage their accounts receivables for improved
client relationships and a multi-channel integrated approach, read part two to find out if data modeling
really is more important than ever before.
Using advanced data modeling and predictive analytics,
SCORE helps businesses improve the recovery and collections process. Contact us
today on 647.309.1803 to get the conversation started.