Page 141 - Building Big Data Applications

Chapter 7 Banking industry applications and usage 139

always ensured that customer is the king and all services provided to customers have
been validated as being valuable. The bank evolved into being a part of the family for the
customer and their financial expectations from loans to investments and retirements are
all managed by the bank. The application requirements from the customer perspective
include the following:
Access to all accounts, balances, and transactions with up to date status
All transactions and associated details
All credit card accounts
All loan accounts
All investment accounts
All household information
The expectation is one application interface with access, security, device support,
interactive call center support, and more robotic assistance with real person support as
needed. Where do we lose sight of this customer and their expectations? How do we
proactively engage with this opportunity to turn it to the advantage of the bank while
benefiting the customer? This is where the application development for predicting churn
will take form and shape.
The basic requirements for this application include the following:
Dashboardda central predictive analytics dashboard that can be shared across call
centers, back office, and agents in the bank
Customer engagement windowdthis is the logging and tracking of every activity
the customer engages with the bank either through the application on a computer,
mobile device, or in-person with an agent. The logging happens as the applications
are used for conducting transactions and the average length of time each activity
occurred. The measure of time is critical here and cannot be missed.
Sentiment collectiondis a key activity to complete for each interaction, it is not a
long-haul list of questions or survey, it has to be a smiley face or a thumbs up emoji
that is needed to be collected. If the customer is dissatisfied they might not reply or
we need to watch for a social media outburst of poor service in the next few hours.
AI and machine learning algorithmsdthe data collected in the logging and the
sentiment response will be processed using AI and ML algorithms in a non-
connected backend process. The algorithm will identify the customer and their
transactions in a private container, and add identities of bank agents if they were
involved in the process. The algorithms will look at behavior, time, transaction
types, number of similar transactions executed, time in days between each trans-
action, online versus offline transaction, sentiment of customer, and associated
outcomes. This data will be then reviewed based on potential prediction outcomes
and impact analytics will be drawn for this customer. A grouping of similar cus-
tomers, their geographies, age demographics, and earning demographics will be
calculated and predicted for churn.