Comptel social links_2.0_presentation v2

© Comptel Corporation 2012 JOINT CONFIDENTIAL
Product Presentation
Social Links 2.0
1

Agenda
2

“Current analytics and marketing practices are
falling short in combating churn and declining
revenue per user”Accenture
“Big data challenges mobile
operators: how to store and
analyse metrics of customers’
real-time activities to better
meet their needs and
preferences?” www.rcrwireless.com
CHALLENGECHALLENGE
OPPORTUNITYOPPORTUNITY
SOLUTIONSOLUTION
Shifting the focus of marketing
from masses to individual subscribers
and their context.
Contextual Intelligence:
Operational + Predictive + Contextual
COMPTEL SOCIAL LINKS 2.0

JOINT CONFIDENTIAL© Comptel Corporation 2012
Churn is usually the largest cost for mobile
operators in mature markets, at about two thirds of
EBITDA.
Churn costs 20% of revenue
Churn compensation &
retention cost (example)
EBITDA
Networks
(incl. personnel, IT)
Personnel (all)
CAPEX
Customer service
(incl. personnel, IT)
IT (all)
30%
20%
19%
15%
13%
6%
4%
Source: www.tefficient.com 2012 North American mobile carriers. Source: Hazlett 2011, George Mason University
2013: The Year Mobile
Broadband stops being profitable
Revenue per GB CAPEX per GB
Total Cost per GB OPEX per GB
$25.0
$30.0
$20.0
$15.0
$10.0
$5.0
$0.0
2010 2011 2012 2013 2014 2015
“Current analytics and marketing practices are failing to
combat churn and declining revenue per user”
4
Source: Accenture 2012

Ability to understand the
uniqueness of a person and/or
circumstance…
Contextual Intelligence =
CONTEXTUAL PREDICTIVE OPERATIONAL
…and convert that
understanding
into an opportunity

Advanced analytics with a business impact
6
25% 14x
Increase in
operator’s
service uptake
compared to
existing
methods
decrease in
churn through
improved
targeting and
more relevant
messaging
Over
400 000€
More revenue
from a single
campaign vs.
status quo
Over 90%
% of PoCs where
Social Links
outperformed
competing and
existing solutions

Agenda
7

Highly productized use cases for maximizing
customer lifetime value
8
Customer’s
Individual
Lifetime
Customer Lifetime
Value (CLV)
Customer
churn
Customer
acquisition
2. Increase customer
profitability
3. Prolong customer
lifetime
1. Customer insight
from day one
Churn prediction
Social churn prediction
Port-out destination pred.
QoS driven churn pred.Next best product prediction
Top-up optimization
Age & Gender Prediction
Value Segment Prediction
Social network Insight
Churn prevention
Contextual top-up optimization
Smart throttling for policy optimization
Operational prediction
Defining optimal actions

Social Links approach to analytics
9
ANALYTICS
OPTIMIZED
PREDICTIONS AND
ACTIONS
CONTEXTUAL
ANALYTICS
OPTIMIZED
ACTIONSADVANCED
ANALYTICS
SIMPLE
ANALYTICS
STATIC
DATA
DYNAMIC
DATA
RULE BASED
PREDICTIONS
AND ACTIONS
EVENT-
TRIGGERED
ACTIONS
Evaluate individual
contexts to decide
whether to take
action and define
optimal action.
Evaluate individual
contexts to decide
whether to take
action and define
optimal action.
Identify trigger (e.g.
balance), send offer
(same for all or
customized by
segments).
Identify trigger (e.g.
balance), send offer
(same for all or
customized by
segments).
Heuristics to define
whether an
individual needs to
top up.
Heuristics to define
whether an
individual needs to
top up.
Predict whether
individual customers
need and want to
be contacted with a
top-up offer.
Predict whether
individual customers
need and want to
be contacted with a
top-up offer.

Step-by-step improvements to business results
Right target, right offer and right context
10
Average
target
Average
offer
Superior
target
Average
offer
Superior
target
Superior
offer
Average or
no context
Superior
target
Superior
offer
Superior
context
Average or
no context
Average or
no
context

Understand uniqueness of individual
subscribes and circumstances:
11
BEHAVIORDEMOGRAPHICS
CONTEXTSOCIAL INTERACTION
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Early prediction of customer behavior enables
pre-emptive actions
12
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Zero-day Intelligence
Example: value segment prediction
13
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Social Intelligence
field-proven component of predictive modelling
14
Predicting the future by utilizing..
… behavioural changes in social networks
… similarities and peer connections
”My friends all
have smart
phones”
”Most of my
friends use a
different
operator.”
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Contextual analytics
Use case: Contextual top-up optimization
15
Morning time, Colin is
commuting, prepaid
balance at 10$
Noon, Colin is
working, prepaid
balance at 8$
Evening, Colin is
visiting mall, prepaid
balance at 2$
Average Colin’s day
Wrong context;
no offer
Urgent need,
offer redeemed
Future need identified,
offer sent!
Predicted need based on
contextual information
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Example: Balance does not always indicate
a need for a top-up offer
16
Medium
balance
Offer X should
be sent
Offer X should
be sent
Low
balance
NO OFFERNO OFFER
Calls a
friend
SMS to
a friend
At a mall
At a mall
Social Links Prediction:
•Will spend a lot during the weekend
•Would not top up without offer
•Will top-up in any case, do not lose reward costs to his offer
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Example: Context alone cannot define which
offer is optimal for an individual
17
Medium
balance
Data
usage
At home
Weekend
Medium
balance
Data
usage
At home
Weekend
Offer Y should
be sent
Offer Y should
be sent
NO OFFERNO OFFER
•Will spend a lot during the weekend
•Will accept an offer for higher top-up than usual
•Just checking some internetsite. Will not use up balance this
weekend  not interested in toppin up.
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Contexts Contexts
Offers Offers
Customer A Customer B
Defining the Contextual Universe
How does it work in practice?
Step 1
Social Links models calculate
a ”propensity map” per each
customer. The propensities
are calculated for all the
available offers and contexts
combinations.
Step 2
In each customer’s
propensity map, there is the
most optimal context-offer
pair that is used in the
event-based-marketing
campaign
Customer A
The most optimal offer and
context could be:
Context = ”Outside home
location, weekend, balance at 52%
of typical average”
Offer = ”Top up 20$, get 3$”
Customer B
The most optimal offer and
context could be:
Context = ”Home location,
working day, balance at 67% of
typical average”
Offer = ”Top up 35$, get 5$”
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Making data beautiful
Turning data into actionable marketing insights
ComptelComptel
Social LinksSocial Links
CDRs
• Local & International
• Voice, SMS, MMS, Video, data
Location
• Call locations
• Residence
• SIM purchase
Billing, top-up
• ARPU,
• Margins
• Top-ups
• Credit balance
Subscriber
• Subscription data
• Tenure, segment
• Survey results
• Device, tariff
plan
Services
• Data usage
• VAS /
Application
usage
Campaign results
• Take-ups, contacts
And any other available data,
e.g., device data, browsing
history, customer care
Quality of
Service
• QoS metrics
such as dropped
calls etc.
Predictive insightPredictive insight
Operational
predictions
Operational
predictions
Defining optimal
actions
Defining optimal
actions
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Agenda
20

• Superior prediction
accuracy
• Machine learning
• Social Intelligence
• Zero-day prediction
• Superior prediction
accuracy
• Machine learning
• Social Intelligence
• Zero-day prediction
• Turning insight into an
opportunity
• Scalability
• Productized use cases
• Process integration
• Turning insight into an
opportunity
• Scalability
• Productized use cases
• Process integration
• Big datastream processing
• Granular predictions over
subscriber and situation
• Holistic view on context
• Big datastream processing
• Granular predictions over
subscriber and situation
• Holistic view on context
Why Social Links?
20% world’s
mobile usage data
20% world’s
mobile usage data
Outperformed
existing solutions in
over 90%
of cases
Outperformed
existing solutions in
over 90%
of cases
”Social Links helps us
to move from fighting
churn to proactive
CRM.” AVEA
”Social Links helps us
to move from fighting
churn to proactive
CRM.” AVEA
21

Driving forces behind Social Links’ capabilities
22
Productized
expertise
Predicting right
things
Future proof
implementation
Engagement
model
Thought
leadership
10 year world-wide
experience in
analytics and 26
years experience
in Telco.
Knowing how to
apply predictive
analytics. Business
driven solution:
each productized
use case is
designed to solve
specific CSP
challenges
(maximizing top-
line, bottom-line
or take-up rate).
Rapid model
retraining
accounts for
changes in market
conditions and
seasonality.
Scalable across
use cases and
over subscribers.
New use cases
easy to deploy.
Supporting
operator teams in
the execution of
successful
campaigns through
analysis of the
Customer’s
current
capabilities and
practices and
sharing of best
practices and
learning.
Allow our
customers to stay
ahead of the game
by constantly
defining new,
better ways to
improve business
goals through
advanced
analytics.
Pioneering in areas
such as RDE and
CIQ4T.

Case study: Social Churn Prediction
and Expected Total Loss Summary
23
Background
Location: EE
ARPU: 9€
Pre-paid: 56%
Mobile subs: 12 million
Churn rate: 4% monthly
Background
Location: EE
ARPU: 9€
Pre-paid: 56%
€
Before Social Links
Solution
Results
Traditional models targeting only the
individual subscriber dimension of churn
Campaigning the subscriber only after
he / she demonstrates propensity to churn
25% decrease in churn
compared to control
group
15% increase in prediction accuracy
20% increase in reach rates for
retention campaigns
Step 1: Identify a customer who is
going to churn
Step 2: Find another customer who is
connected with and influenced by the
targeted customer within 30 days
Step 3: Target campaigns to
customer pairs that are most
likely to churn and / or have the
highest expected total loss

© Comptel Corporation 2012 JOINT CONFIDENTIAL24
Before Social Links
Background
Location: MEA
ARPU: 13€
Pre-paid: 80%
Churn rate: 1.5% monthly
Target group: 10 000
Background
Location: MEA
ARPU: 13€
Pre-paid: 80%
Target group: 10 000
Solution
Results
Voice-to-SMS take-up (below
0,5%) was considered low
Sub-optimal product stimulation
campaigns
Case study: Next Best Product Prediction
Phase 1. Defining two target groups
2.8%
0.2%
Take-up rate
Improvement:
over 14 times
higher take-up
rate
Social Links
selection (10k)
Operator selection
(10k)
1. Social Links combined
propensity, influence and
behavior predictions
2. Operator’s current behavior
based methods
Phase 2. Campaigning
Same MMS/SMS campaign for
both target groups
No incentives
Based on propensities, no
campaign optimization
”Try out
voice to
SMS”

Before Social Links
Background
Location: APAC
ARPU: 8€
Pre-paid: 47%
Mobile subs: 4.3 million
Churn rate: 3-4% monthly
Background
Location: APAC
ARPU: 8€
Pre-paid: 47%
Mobile subs: 4.3 million
Churn rate: 3-4% monthly
Solution
Results
Reactive and slow
campaign intelligence
No insight into new customer
potential value
Sub-optimal retention campaigns based
on inaccurate future predictions
the future value of customers
predicted correctly in 80-90%
of cases for the target group!
Case study: Zero Day Value Segment Prediction
Prediction of new customer future value
after 60 days with only 1 day input data
High value after 60 days
Mid range after 60 days
Low value after 60 days
New
customer
Project delivered in 6 weeks
after receiving the data
Superb results by utilizing
nearly 1000 variables

Case study: Inactivity Prevention
Results
Solution
21% reduction in churn rate
25% increase in revenue
Step 1. Prediction of the
likely churn pointActive stage of customer life-cycle
Background
Location: Europe
ARPU: 12€
Pre-paid: 85%
Campaigns: Monthly
Target group: Top 10% monthly
Background
Location: Europe
ARPU: 12€
Pre-paid: 85%
Campaigns: Monthly
Target group: Top 10% monthly
Before Social Links
Monthly campaigns were
unprofitable
High monthly
churn rate: 7%
Campaigns did not have a
retention impact
Step 3. Optimisation of personalized
retention offer for each high churn risk
customer and contacting the customer with
an SMS offer
More retained customers and better
secured revenue
Enabled by granular campaign
optimisation and predictive targeting
Step 2. Uplift modelling to
optimise revenue from
churn
26

Before Social Links
Identify those likely to respond positively
+
Tailor personalized required actions
+
Tailor personalized top-up rewards
Background
Location: MEA
ARPU: 13€
Pre-paid: 80%
Campaigns: Monthly
Target group: 1 million
Background
Location: MEA
ARPU: 13€
Pre-paid: 80%
Campaigns: Monthly
Target group: 1 million
Solution
Results
No analytics used for
optimizing top-up offers
Sub-optimal revenue from
prepaid top-up campaigns
Top-up recharge rewards always fixed,
e.g., 5% of the required action
“Top-up 15€
now, get 2€
extra!”
Net revenue from single campaign
Million€
400 000€ more revenue from
a campaign vs. status quo
threefold improvement on
CSP’s own method
Case study: Top-up Optimization
Individual offer to
all customers

Agenda
28

Social Links Processes for Campaign Optimizations
From Deployment to Operational usage
29
Monthly/quarterly review, which may
trigger new use case needs
~2w 1w-4w 2w 1w-4w 1w-4w Total: 7w-16w
Daily/
weekly/
bi-weekly/
monthly
Installation
+
automation
Deployment process
Operational usage process
Generic timeline:
PREDICTIVE
OPERATIONAL
CONTEXTUAL

Agenda
30

Churn Prediction
31

Use case: Churn Prediction
Business challenge: Who are the most likely customers about to churn (post-paid) or start
inactivity (pre-paid)?
Step 1. Customers
using CSP services
Intelligent Analysis
by Comptel Social Links
Step 2. Intelligent analysis predicts
churn/start of inactivity for the future
Mediating
usage &
customer data
Step 3. customers are
contacted before they
make decision to churn
or start inactivity
Call centre
SMS/MMS
Prediction in advance!
Tenure
Activity
Turning
analysis into
actionable
target list
Benefit: Detect churners before decision to churn is made

Port-out Destination Prediction
33

Use case: Port-out Destination Prediction
Business challenge: Which service provider are the customers churning to?
Step 1. Customers
using CSP services
Step 2. Intelligent analysis predicts which
service provider customers are churning to
Mediating
usage &
customer data
Turning
analysis into
actionable
target list
Step 3. Contacting and
offers are individualized
per customer
based on their most
likely port-out
destination
Call centre
SMS/MMS
Benefit: Allows CSP to tailor competitor specific prevention offers

Social Churn Prediction
35

Use case: Social Churn Prediction
Business challenge: Which customers in social network are most influencing others to churn?
Step 1. Customers
using CSP services and
communicating with
each other
Step 2. Intelligent analysis predicts the
social churn score per each customer
Mediating
usage &
customer data
Step 3. Customers with
highest churn influence
can be treated with
special offers
Try out
mobile
gaming
app with
your
friends for
free for 1
month!
Try out
mobile
gaming
app with
your
friends for
free for 1
month!
Step 1: Churner?
Step 2: Influenced
churner?
Turning
analysis into
actionable
target list
Benefit: Insight for selecting attractive offers for churn prevention

Value-segment Prediction
37

Use case: Value-segment Prediction
Business challenge: What is the future value of the new customers?
Step 2. Intelligent analysis predicts the
future value-segment starting only with 0-
day data
Mediating
usage &
customer data
Step 1. New customers
using CSP services
Step 3. Welcome and
loyalty programs guided
by customers’ individual
value-prediction
Targeted
Welcome
and Loyalty
programs
Profitable
offersNew
Customer
(0-1 days)
60 days in
the future
High
Mid
Low
Turning
analysis into
actionable
target list
Benefit: Profitable treatment of new customers based on their predicted value

Next Best Product Prediction
39

Use case: Next Best Product Prediction
Business challenge: What is the best up-sell product for each customer?
Step 2. Intelligent analysis predicts most
suitable next product per customer
Mediating
usage &
customer data
Step 3. Each customer is
offered with the most
suitable next product in
outbound or inbound
marketing
Call centre
SMS/MMS
Step 1. Customers
products
Suitability for
Gold data
package
Suitability for
Gold data
package
Suitability for
Silver data
package
Suitability for
Silver data
package
Turning
analysis into
actionable
target list
Suitability for
voicemail
Suitability for
voicemail
Benefit: Higher product take-up through more attractive offers

Age & Gender Prediction
41

Use case: Age & Gender Prediction
Business challenge: How to personalize marketing, when I don’t know anything about my
pre-paid customers?
Step 2. Intelligent analysis predicts age &
gender per customer
Mediating
usage &
customer data
Step 3. Personalized
marketing activities
guided by age&gender
information
Call centre
SMS/MMS
Step 1. Customers
using CSP services
Male
Age
26-35
Female
Age 46-
55
Turning
analysis into
actionable
target list
Benefit: Insight for selecting attractive offers for customer marketing

Social Network Insight
43

Use case: Social Network Insight
Business challenge: What are the social characteristics of the customers?
Step 2. Intelligent analysis calculates in
detailed level social and behavioral
characteristics of customers
Mediating
usage &
customer data
Step 3. Insight about who
customers communicate
with is used to personalize
messages and offers
Step 1. Customers
communicating with
each other
Call to
your off-
net friends
with
special
off-net
package!
Call to
your off-
net friends
with
special
off-net
package!Turning
analysis into
actionable
target list
Benefit: Insight for selecting attractive offers for customer marketing

Churn Prevention
45

Use case: Churn Prevention
Business challenge: Which offers would prevent predicted churners from churning?
Step 1. Customers
using CSP services
Step 2. Intelligent analysis predicts most
suitable churn-prevention-offer per
customer
Mediating
usage &
customer data
Step 3. Likely churners
are contacted with most
suitable offer to retain
them
Call centre
SMS/MMS
Prediction of
most suitable
offer
Tenure
Activity
Offer3Offer3
Offer1Offer1
Offer5Offer5
Turning
analysis into
actionable
target list
Benefit: Personalized prevention offers lead to superior business impact

Top-up Optimization
47

Use case: Top-up Optimization
Business challenge: Which top-up offers should be targeted to whom to maximize
profits/revenue?
Step 1. Customers
using CSP services
Step 2. Intelligent analysis predicts optimal
targets and top-up offers
Mediating
usage &
customer data
Low on
balance?
Top-up
10$ now,
get 3$
extra!
Low on
balance?
Top-up
10$ now,
get 3$
extra!
Step 3. Most suitable
target customers
receive most optimal
offer
“Top-up 10$
now, get 3$
extra!”
Individually
selected offer
Turning
analysis into
actionable
target list
Benefit: Maximized revenue from top-up campaigns

Contextual Top-up Optimization
49

Use case: Contextual Top-up Optimization
Business challenge: What is the best context to engage the customer with an offer?
Intelligent Contextual Analysis
Intelligent Contextual Analysis
Step 2. Contextual analysis predicts IF offer
should be sent at the current context
Mediating
usage &
customer data
“Top-up 10$
now, get 3$
extra!”
Context specific
individually selected
optimal offer
Low on
balance?
Top-up
10$ now,
get 3$
extra!
Low on
balance?
Top-up
10$ now,
get 3$
extra!
Step 3. Customer
receives targeted offer
IF the context for him
was relevant at this
time
Step 1. Customer’s
balance drops to 2$,
it’s a weekday and
he’s at his most typical
location
Turning
analysis into
action
Benefit: Context driven campaigning leads to supreme take-up rates

QoS Driven Churn Prediction
51

Use case: QoS Driven Churn Prediction
Business challenge: Considering Quality of Service from CDR data - Who are the most
likely customers about to churn (post-paid) or start inactivity (pre-paid)
Step 1. Customers
using CSP services
Step 2. Intelligent analysis predicts
churn/start of inactivity for the future
taking into account Quality of Service
Mediating
usage &
customer data
Step 3. customers are
contacted before they
make decision to churn
or start inactivity
Call centre
SMS/MMS
Tenure
Activity
Turning
analysis into
actionable
target list
Benefit: Enables CSP to take QoS driven churn prevention actions

Smart Throttling
53

Use case: Smart Throttling
Step 2. Intelligent analysis predicts, which
customers should get priority in network
Mediating
usage &
customer data
Step 1. Customers are
using data services Step 3. Customer priority
prediction used as input in
policy controlling
Optimised
bandwidth
distribution
Smart throttling
based on
predicted
customer priority
High priority
Mid priority
Low priority
Business challenge: Need to prioritize scarce network resources
Analysis is
used in
bandwidth
prioritization
Benefit: Optimal allocation of network resources based on customer priority
Customers have been prioritized
based on their predicted value

Thank you
55

Comptel social links_2.0_presentation v2

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Was ist angesagt?

Was ist angesagt? (20)

Ähnlich wie Comptel social links_2.0_presentation v2

Ähnlich wie Comptel social links_2.0_presentation v2 (20)

Mehr von Rafael Junquera

Mehr von Rafael Junquera (20)

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

Comptel social links_2.0_presentation v2

Hinweis der Redaktion