Using Your Data to Reduce Attrition in Banking
- 1. Using
Your
Data
to
Reduce
A3ri4on
in
Banking
Webinar
Wednesday,
April
29th,
2015
- 2. 2
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
A3ri4on
is
a
Top
Issue
for
Banks,
and
for
Good
Reason
The
opportunity
cost
of
falling
behind
the
compeCCon
is
extreme.
Over
half
of
customers
have
opened
or
closed
at
least
one
product
in
the
past
year
and
nearly
as
many,
40%,
plan
to
do
so
in
the
coming
year.
Each
of
these
customer
represents
a
new
business
opportunity
for
a
compeCng
bank
or
financial
service
provider.
Global
Consumer
Banking
Survey
2014,
Ernst
&
Young
- 3. 3
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Advantage
-‐
Banks
• Financial
Services
have
many
digital
touch
points
with
their
customers
where
they
can
drive
communicaCon
• Financial
Services
don’t
want
to
put
communicaCon
in
the
hands
of
third
parCes,
such
as
technology
companies
that
could
become
compeCCon
- 4. 4
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Focus
on
the
Most
Effec4ve
A3ri4on
Program
Involuntary
RotaConal
Voluntary
- 5. 5
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
TradiConal
Advanced
Most
Likely
A3ri4on
Predictors
Use
all
your
data
to
create
smart
Customer
DNA
Metrics
• Socio-demographic
• Slow-changing metrics
- product ownership
- subscriptions
- age
• Behavioral
• Rapidly-changing
metrics
- usage
- service interaction &
consumption
ADVANTAGE:
&
ADVANTAGE:
&
- 6. 6
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Addressing
the
A3ri4on
Challenge
Making
a
difference
throughout
the
process
1 Prepare the Data
Build a Better Attrition Model2
Become Actionable & Learn From Feedback3
- 7. 7
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Prepare
Data
1
TradiConal
Advanced
• Prepare
mulC-‐source,
omni-‐channel
Customer
DNA
which
can
serve
mulCple
use
cases
• Explore
relevant
predictors
• Export
dataframes
to
quickly
build
model
in
analyCcal
workbench
• IdenCfy
problem
&
gather
data
from
different
sources
specific
to
the
problem
• Understand
the
data
completely
• Sample
the
data
- 8. 8
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Build
Be3er
A3ri4on
Models
2
TradiConal
Advanced
• Easily
build
models
with
perfectly
prepared
dataframes
aligned
to
the
individual
churn
dates
• Combine
socio-‐demo
and
historical
predictors
with
behavioral
metrics
to
define
why
and
what
as
well
as
how
and
when
• Combine
and
easily
switch
between
short-‐
and
long-‐term
predictors
• Focus
on
predictors
indicaCng
behavioral
change,
such
as
trend
and
accelera4on
• Models
built
on
subset
of
data
at
one
point
in
Cme
• Segmented
data
-‐
not
related
to
individual
customer
acCons
or
intents
• Model
becomes
outdated
from
day
one:
maintenance
heavy
- 9. 9
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Become
Ac4onable
and
Learn
from
Feedback
3
TradiConal
Advanced
• ConCnuous,
real-‐Cme
aariCon
scoring
of
every
individual
customer
• Detect
&
alert
most
appropriate
moment
of
acCon
for
every
individual
customer
• Capture
feedback
to
learn
about
channel
&
offer
type
performance
and
preference
• Batch
scoring
of
customers
• AariCon
prevenCon
acCon
oben
too
late
• Offer
feedback
informaCon
oben
unavailable
• Slow
&
staCc
process
- 10. 10
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
The
Status
Quo
of
Insight
Where
is
the
Customer?
Gave
up
on
Customer
360
aber
large
investments
in
Datawarehouses
Use
hindsight
in
BI/
AnalyCcs
soluCons
building
complex
diagnosCc
models
for
customer
segmentaCon
Hire
an
army
of
data
scienCst
to
use
big
data
and
visualizaCon
tools
to
discover
insights
Rely
on
Rule
Engines
to
apply
segmentaCon
for
recommendaCons
and
targeCng
Most
Many
Several
Few
Rowan
Curran,
March
2015,
Forrester
Research:
“Digital
experience
delivery
vendors
have
generally
fallen
short
in
their
use
of
predic@ve
analy@cs
to
contextualize
digital
customer
experiences.
Many
of
these
vendors
offer
simple,
rules-‐based
recommenda@ons,
segmenta@on,
and
targe@ng
that
are
usually
limited
to
a
single
customer
touchpoint.”
- 11. 11
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
The
Lily
Revolu4on:
Customer
at
the
Center
DNA
metrics
can
be
sophisCcated
models,
coming
from
SAS,
R,
SPPS
of
other
staCscal
soluCons.
From
Data
to
DNA
–
1000s
of
metrics
determine
individual
customer
DNA
Alerts
in
real-‐Cme
on
all
metrics
to
drive
customer
interacCon.
Sets
on
DNA
metrics
to
drive
campaigns.
Trending
–
Keep
track
of
historical
values
and
trends
of
all
DNA
metrics
in
the
system
Manage
Big
Data
-‐
Breaking
down
data
silos
to
gain
insights
on
all
customer
interacCons
in
one
place
> > > >
From Manual Work Step by Step to Continuous Automation
- 12. 12
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Customer
Centricity
Creates
IMPACT
- 13. 13
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
customer removes multiple
products from portfolio
6
OCT
customer
churns
11
NOV
manual attrition score (bi-monthly)
portfolio size (weekly)
Figh4ng
A3ri4on
Before
it
is
Too
Late
- 14. 14
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
win-back period
customer removes multiple
products from portfolio
6
OCT
customer
churns
11
NOV
win-back sensitivity
manual attrition score (bi-monthly)
portfolio size (weekly)
Connect
at
the
Sensi4ve
Win-‐back
Period
for
Op4mum
Results
- 15. 15
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
win-back period
win-back sensitivity
Lily attrition score (continuous)
portfolio size (weekly)
customer
churns
11
NOV
customer
retention
actions
Lily alerts for in-
creased attrition risk
customer removes multiple
products from portfolio
6
OCT
Timely
Alerts
and
Ac4ons
for
the
Greatest
Impact
- 16. 16
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Decreasing
A3ri4on
-‐
Banking
• Created
thresholds
and
set
alerts
based
on
conCnuous
trending
scores
on
all
available
data
and
delivered
more
predicCve
acCons.
• Alerts
sent
to
bank’s
outbound
systems
to
take
acCons
reducing
aariCon
by
10%
Result
• CompeCCve
pressure
on
the
retail
business
• Need
to
substanCally
lower
aariCon
rate
(22%)
• Increase
customer
lifeCme
value
ObjecCves
• Aggregated
all
customer
data
(ATM,
branch,
call
center,
web,
mobile,
payment
system,
etc.)
• Built
individual
Customer
DNA
based
on
hundreds
of
metrics
• Focused
on
the
high
value
customers
(HVC)
based
on
CLTV
metric
• Informed
outbound
systems
of
HVCs
at
risk
based
on
conCnuous
aariCon
scoring
SoluCon
“
NGDATA
is
cri@cal
in
the
way
we
capture,
analyze
and
generate
ac@onable
intelligence
from
Big
Data.
With
Lily
in
place,
we
were
able
to
find
and
act
on
the
customers
most
at
risk
of
aMri@on
in
a
@mely
and
effec@ve
manner.”
—
CIO,
Large
interna4onal
bank
- 17. 17
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Lily
Enterprise
Connect
with
your
customer
by
being
relevant
Preferences
AffiniCes
Context
Behavior
Trends in Core Metrics and Preferences trigger
relevant communications in all digital channels3
Lily captures first and third party Customer Behavioral data, immediately translated into
Core Customer Metrics and Preferences in Real Time1
Core Metrics and Preferences actionable on individual customer level, continuously
available for personalized customer communications2
- 18. 18
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Deliver
What
Your
Customers
Want
OFFER
THE RIGHT
PERSON
THE RIGHT
TIME
THE RIGHT
CHANNEL
THE RIGHT
IMPROVED FREQUENCY
IMPROVED SEPARATION
- 20. 20
Copyright
2015
NGDATA
®,
Inc.
ConfidenCal
–
DistribuCon
prohibited
without
permission
Ready
to
take
the
next
step
to
reduce
customer
a3ri4on?
Learn
more
about
how
Lily
Enterprise
can
help
your
bank.
Schedule
an
appointment
with
an
NGDATA
representaCve
to
get
a
personalized
walkthrough.
Don’t
forget
to
follow
up
and
share
with
a
friend
stevenn@NGDATA.com
Editor's Notes
- Would tell the background story to this slide with an anecdote story of a customer need for what we were creating
- Involuntary: financial attrition
when a customer no longer has the finances to continue doing business with the bank or can no longer afford the product; customer dies; customer relocates and no longer continues the relationship with the bank
Voluntary Churner:
commercial churners- when a customer consciously decides to stop doing business with the bank
Rotational Churner - two scenarios:
Customer drops a product and purchases another product within the same bank to obtain better conditions (e.g. if you become a customer, you receive a promotional offer, take the company up on the offer and close the existing product in order to qualify for the new offer. This is a ‘fake’ new customer by opening the closed account – with the new deal that was offered)
Customer constantly (partially) switches between companies depending on where he can get the best conditions. (e.g. bank customer transfers money from savings account to competitor, having an interesting savings account offer.)
The most effective churn situations to address are for the customers who churn voluntarily, not involuntary churners nor rotational churners.
Need image (graphic designs not photos) illustrating the different churners.
Turn into a visual
Ability to act on lack of activity
People are used to acting on activity
Behavior – key differentiator can be activity and lack of activity
Customer walking out of the door
Don’t include involuntary churn
There are 3 types of churners and we’re focusing on voluntary churn for this
Identify the different types so you don’t waste $ on the wrong types
-
BOTH CAN BE MONITORED FOR (IN)ACTIVITY AND TREND, AND TO GENERATE ALERTS
Predictors –
Traditional scoring
Banks have sophisticated models from historic data not based on usage (could be account balance checks, web data etc)
Banks have been focused on historical models
Combining historical data and real time behavioral metrics
Product propensity
Behavioral and interaction data is important
Previously, models build on top of slow moving profiles
Only did churn scoring once a quarter bc values didn’t chance often. Now more valuable to measure more often
Traditional attrition predictors used by banks have been based on historical data, yielding dated insights from batch processing that represented a snapshot at a single point in time
Attrition prediction for voluntary customers:
Product level:
will the customer drop the product or not renew the contract?
Customer level:
fully: customer drops all products/services and stops doing business with the bank.
Banks address:
Who & Why
Lily Addresses:
Who, why, where and how by using usage and activity based data
Customers want a consistent omni channel experience
Using behavioral data, and tracking all drivers towards attrition, in real-time, in addition to socio-demographic data will yield the most actionable, real-time insights to proactively address these attrition predictors and enable you to meet customer expectations
- Prepare data to obtain a complete view on the customer
Build a better attrition model
Become actionable and learn from feedback
-
BOTH CAN BE MONITORED FOR (IN)ACTIVITY AND TREND, AND TO GENERATE ALERTS
Predictors –
Traditional scoring
Banks have sophisticated models from historic data not based on usage (could be account balance checks, web data etc)
Banks have been focused on historical models
Combining historical data and real time behavioral metrics
Product propensity
Behavioral and interaction data is important
Previously, models build on top of slow moving profiles
Only did churn scoring once a quarter bc values didn’t chance often. Now more valuable to measure more often
Traditional attrition predictors used by banks have been based on historical data, yielding dated insights from batch processing that represented a snapshot at a single point in time
Attrition prediction for voluntary customers:
Product level:
will the customer drop the product or not renew the contract?
Customer level:
fully: customer drops all products/services and stops doing business with the bank.
Banks address:
Who & Why
Lily Addresses:
Who, why, where and how by using usage and activity based data
Customers want a consistent omni channel experience
Using behavioral data, and tracking all drivers towards attrition, in real-time, in addition to socio-demographic data will yield the most actionable, real-time insights to proactively address these attrition predictors and enable you to meet customer expectations
-
BOTH CAN BE MONITORED FOR (IN)ACTIVITY AND TREND, AND TO GENERATE ALERTS
Predictors –
Traditional scoring
Banks have sophisticated models from historic data not based on usage (could be account balance checks, web data etc)
Banks have been focused on historical models
Combining historical data and real time behavioral metrics
Product propensity
Behavioral and interaction data is important
Previously, models build on top of slow moving profiles
Only did churn scoring once a quarter bc values didn’t chance often. Now more valuable to measure more often
Traditional attrition predictors used by banks have been based on historical data, yielding dated insights from batch processing that represented a snapshot at a single point in time
Attrition prediction for voluntary customers:
Product level:
will the customer drop the product or not renew the contract?
Customer level:
fully: customer drops all products/services and stops doing business with the bank.
Banks address:
Who & Why
Lily Addresses:
Who, why, where and how by using usage and activity based data
Customers want a consistent omni channel experience
Using behavioral data, and tracking all drivers towards attrition, in real-time, in addition to socio-demographic data will yield the most actionable, real-time insights to proactively address these attrition predictors and enable you to meet customer expectations
-
BOTH CAN BE MONITORED FOR (IN)ACTIVITY AND TREND, AND TO GENERATE ALERTS
Predictors –
Traditional scoring
Banks have sophisticated models from historic data not based on usage (could be account balance checks, web data etc)
Banks have been focused on historical models
Combining historical data and real time behavioral metrics
Product propensity
Behavioral and interaction data is important
Previously, models build on top of slow moving profiles
Only did churn scoring once a quarter bc values didn’t chance often. Now more valuable to measure more often
Traditional attrition predictors used by banks have been based on historical data, yielding dated insights from batch processing that represented a snapshot at a single point in time
Attrition prediction for voluntary customers:
Product level:
will the customer drop the product or not renew the contract?
Customer level:
fully: customer drops all products/services and stops doing business with the bank.
Banks address:
Who & Why
Lily Addresses:
Who, why, where and how by using usage and activity based data
Customers want a consistent omni channel experience
Using behavioral data, and tracking all drivers towards attrition, in real-time, in addition to socio-demographic data will yield the most actionable, real-time insights to proactively address these attrition predictors and enable you to meet customer expectations
- VOICEOVER TO BE REDONE
Moreover, it is really important to act on churn-related behavior as soon as possible, as the likeliness to win back the customer decreases over time.
- VOICEOVER TO BE REDONE
Moreover, it is really important to act on churn-related behavior as soon as possible, as the likeliness to win back the customer decreases over time.
- VOICEOVER TO BE REDONE
Lily will notify you at the first symptoms of churn, allowing you the appropriate time to take action and retain the customer. Thus allowing you to drive retention programs to maximum effectiveness.
- First identify high potential churners
Banks need to choose a scenario with the most value to the company – high value customers – whose attrition could be easily averted, with proper knowledge and clear implementable actions.
Include with the scale image
How can you best serve
Your
-Focus on high value
Spend market budget wisely (high value customers)
At the right time, channel, context (slide with guy sitting on a bench on banner)