TELCO Big Data Churn Analytics

Identifying revenue growth opportunities and

strengthening CEM customer retention policy
BSS OSS COTS OTT CHURN DATA MODELING DATABASE CREATION
ACCURATE CHURN PREDICTION USING MARKOV PROCESS CHAINS
Prepared and presented by
Prof. Dr. Dipl. Wirtsch. Ing. Mehmet Erdas
MBA B.Sc. M.Sc. METU Ph.D. TU Braunschweig Germany
[email protected]
[email protected]
Mobile: +49 (0)1789035440
+43(0)6509111090
+90(0)5374154413

1
2
1. Objectives
2. Revenue growth and retention – our scope of work 2015
3. How to participate; Our reference architecture – guide
Identify Use Cases& Define the Business Use Cases, KPIs
3.1 Upsell
3.2 Cross Sell
3.3 Retain Customers: Churn Minimization
4. LOCATION
5. MOVEMENT
6. NEXT BEST COURSE OF ACTION TIMING&SPEC.
7. Movement Solution Scope
8. Role Based HR Project Resourcing and Budgeting
9. Use Case Identification SLA VIP etc..(by Presentation)

3
Why Big Data Analytics CEM Churn Project?
We will drive the development of appropriate technology and
steer technology & Service Quality delivery models for new
services and products based on deep profile customer
inspection/experience i.e big data subscriber profile involving
social networks and word of mouth after integrating structured
and unstructured data using in-memory HANA and Hadoop MR

1. Mobile operator agrees to participate in

use case focused workshops
 Mobile operator supplies customer data
samples
 Customer identity encrypted by operator
2. Provider builds and deploys operational
prototype for one or more of the use
case listed
 Operator can validate business value
3. Provider and operator agree solution,
service and technology roadmap
4
1. Customers benefit from the data they
Generate: Permission Based Marketing
 My mobile operator treats me as a person, not a KPI.
 They will try to understand what is important/relevant for me
- they will genuinely offer me the best deals.
 This includes not just their own services, but they will find and
provide me access to the best deals out there that improve my
life.
 Perhaps an easier/more economical route to work. Perhaps
access to a pay as you use car insurance scheme, or a life
insurance scheme that takes into account the amount of
sleep, exercise that I routinely partake of etc.

2. Operators generate more value for

customers by new APPS: CONTENT_
META_MASTER _TRANSACTIONAL DATA
 All commercial business needs to generate profit.
 There are two philosophies on this:

Inside-out:
We reduce costs, increase revenues, profit is the difference
between the two.

Outside-in
We generate value for our customers, profit is the natural 5
consequence of this.
 The challenge – right data!
1. Volume, Variety, Veracity, Velocity
χ It is neither possible nor beneficial to store all
data.
 It is important to store the right data: First
Achieve the Highest Data Quality Measures
Big Data
2. Value = Cost
 To identify the right data, experts are
Data
required.
structured $$$
Tsunami Right Data
= Profit
un-structured

Continuous Ingestion Continuous Queries /Analytics on

data in motion 6
 Our Big Data solution roadmap proposal

2015
Data Workloads
Scope Def.n

Analytics 2016
Platf. Spec.

HANA Hadoop Sys Int.

Automation
2017 • .

Processing
structured&unstructuredd 2018
ata combined

7
1. OBJECTIVES
1. We are strategically committed to help
our customers increase their
profitability

2. In support of this we will present an

overview of the work programs that we
are under taking in 2014 that focuses on
revenue growth and customer
retention

3. Reference Primer Architecture

We hope to solicit feedback from key
TELCO customers and identify
customers that are willing to participate
in a joint work program next year - 2015
8
2. Revenue growth and retention
- business focused use cases
1. Tradition/Off-line use cases
 Up-sell based on usage analysis – i.e. sell the
customer more of what they already consume
 Cross-sell based on usage analysis – i.e. sell the
customer additional products and features
 Targeted retention of existing customers based
on churn analysis

2. Next generation/On-line uses cases

 Targeted marketing of customer segments based
on location through event calendar correlation
 Targeted marketing of customer segments based
on movement along a transportation corridor
 Enhanced customer care handling through next
best action suggestion

9
10
 3. Our reference architecture - guide
Next generation
Network technology
Mgt - Stats Long lived data
Traditional
Network technology
Inventory

Device
On-demand
 Consolidation,
Inventory Periodic Collection Filtering and Data storage Data analytics
Filtering Correlation  Fast Retrieval Option  Unpredictable Queries
Immediate
Enrichment  Standard Retrieval Option  High Responsivesness
BSS
Event Correlation
Immediate Analytic
TT, Event Aggregation
Information Element Event On-demand Database
Workflow Periodic Repository

On-demand

Periodic
Immediate
Immediate

Immediate
CDR, Logs
 Real-time
Network Event
Streaming Data Data Presentation
Mgt - Event Writer
Sockets Automation  Dashboard
Event  Report
Driven
Raw Data Rules Logic Production
Reader Web based
Sockets
NE Notification API
Implementatio
Graphical
Event Factory n Context
Statistical and
Mathematical
Functions
UE
 Predicate based
Group/Set Logic
Probes Data production

IOT Sensors
Short lived data User Equipment NW Policy Control Service Subscription
BSS
Configuration Mgt Equipment Databases

11
 3.1 Up-sell
The objective of this use case is to proactively identify customers who have exceeded one
or more elements (e.g. mobile data) of their contracted tariff plan, and proactively offer
them additional capacity for an incremental fee.

The
business
case

For example:

Customer complains they have unexpectedly incurred additional charges for mobile data
usage. We verify through usage analysis that the charges are due to legitimate
downloading from Google market. We can offer a more suitable tariff plan based on the
actual usage profile.

12
 3.2 Cross-sell
The objective of this use case is to identify customers who are likely to purchase
additional products.

The
business
case

For example:

Through usage analysis we identify those customers who are routinely downloading
music from iTunes. We then offer them an alternative subscription to Spotify
highlighting how much they would have saved based on recent purchases.

13
 3.3 Retain customers The
business
case
The objective of this use case is to establish the propensity of our customers to
churn through the identification and analytic modeling of churn indicators
BSS (customer facing) – i.e. billing and CRM data
POSTPAY Calls to customer service PREPAY
Length Of Service
Initiation Credit Value
Contract Stage No. Of Products
Tariff Band Disconnect Reason Top Up Frequency
No. Of Upgrades
X-Net Ratio
Unpaid Balance freq. Avg. Credit Value
No. OB Calls [Delta
Complaints Flag
Sphere of influence Avg. Inactive Time
Age Band
Loyalty Discount
POSTPAY Competitor Promo Flag PREPAY
Geo [Urban Rural]
Device Type
For example: Top Up Method

We can produce a predictive model that encompasses both OSS and BSS data
sources that identifies customers most likely to churn. This data can be used to
inform retention policy within a mobile operator.

14
 4. Location
The objective of this use case is to correlate customer location, pre-provisioned events
and a customers profile, for specific promotions and communications.
Actual location Tailored promotion
User Preference Event Calendar

1. Frank has regular 2. System provisioned with 3. Correlate with location 4. Timely and tailored
access to Man event calendar (e.g. Man actual data promotion
United app United versus Barcelona @
Location, date, time)

For example:
Based on customer usage we establish Frank is a Man United football fan. Correlating
this information with his cell location (e.g. while attending a football game) and known
football fixture timetable can be used to route him towards an accessible but relevant
offering - e.g. sale on Man United club merchandise.

15
 5. Movement
The objective The objective
of this use of this use
case is to case is to
correlate correlate
Source device Source device customer customer
Track device Track device
location and Maps these location and Maps these movement movement
location and Publish the Retailers act location and Publish the Retailers act
movement segments to movement segments to with with
movement of local analytic upon movement of local analytic upon
data from commercially data from commercially customer customer
segmented data opportunities segmented data opportunities
available relevant areas available relevant areas profile for profile for
users users
sources sources specific specific
promotions promotions

Source device location and movement data from available sources and
communicatio
ns.
and
communicatio
ns.

Track device location and movement of segmented users

Maps these segments to commercially relevant areas
Publish the local analytic data
Retailers act upon opportunities
Source device location and movement data from available sources
Track device location and movement of segmented users
Maps these segments to commercially relevant areas
Publish the local analytic data
Retailers act upon opportunities

The objective of this use case is to correlate customer movement with customer profile for
specific promotions and communications.

The objective of this use case is to correlate customer movement with customer profile for
16
specific promotions and communications.
 6.Next best action
The objective of this use case is to enhance customer care handling through next best
action suggestion.

Communicate Improved First

Validate next best Call Resolve
problem action ratio

For example:

Mark rings first line customer care. He explains he is dissatisfied with his quality of his
mobile data service. Our system has validated that the download speed is below the
norm for Mark. It has correlated this with the application of a new software configuration
on his handset. Updating the configuration to the latest available version resolves the
issue for Mark.

17
7. Movement> Solution Scope
Mobile operator agrees to participate in use case focused workshops
Mobile operator supplies customer data samples
Customer identity encrypted by operator

Provider builds and deploys operational prototype for one or more of the
use case listed
Operator can validate business value

Provider and operator agree solution, service and technology roadmap

18
 Use case Blue
Role Based Project Resourcing &Budgeting Analytics Model Yellow
Service Model Red
BigData Platform Blue

Names Focus Profile/Role Onboard

NN Use Case Design(Campaign/Churn) 20 years experience. SQM/CEM product management Now
AB Use Case Design(Campaign/Churn) 20 years experience. OSS/SQM/CEM product architecture and design. Now
CD Use Case Design(Campaign/Churn) 20 years experience. Operator marketing operations management. June
EF Use Case Design(Campaign/Churn) 10 years experience. Marketing campaign design. Now
GH Use Case Design 20 years experience. Telcordia SQM/CEM market management and solution design. June
PK Use Case Design(Customer Care) 10 years experience. Huawei Core Network R&D Now
LM Use Case Design(Customer Care) 15 years experience. NSN SQM/CEM solution architect July
NO Analytics Model Design(Churn/ Campaign) 20 years experience. Data mining, familiar with Chun/Campaign Now
PQ Analytics Model Design(Churn/ Campaign) 20 years experience. Data mining, familiar with Chun/Campaign July
HF Service Modelling 10 years experience Huawei Core network R&D and SmartCare product management Now
PQ Service Modelling 10 years experience. Huawei Core network R&D and SmartCare Service modeling Now
FM Service Modelling/Transformation 20 years experience. IBM COTS service modeling design June
XY Service Modelling/Transformation 15 years experience. IBM COTS service modeling design July
UV reference architect 20 years experience. OSS/SQM/CEM product architecture and design. Now
Dr. Mehmet In-memory architect 30+ years experience of Data ware housing and SAP HANA in-memory database professor Now
NE systems architect 10 years experience. Business intelligence expert Now
NM streaming architect 10 years experience. Ericsson OSS/SQM/CEM research and application architecture June
NN DWH + ETL architect 15 years experience. Netezza Big Data system architect July
15 years experience. Online analytics and quantitative modeling of high-performance low-latency
NN data mining architect systems. June
Jingjin portfolio architect 14 years experience. Huawei R&D. Now

19
UC-1 Customer complaint handling
• Scenario 1
– Clearly demarcate Server (Video) issues beyond operator control

• Scenario 2
– Convert contact into additional revenue

• Scenario 3
– Clearly demarcate UE (APP) issues beyond operator control

• Scenario X1
– Improve TT handling efficiency (automatically insert technical
detail)

20
Complaint Handling #1– Prevent ticket creation with rapid
customer insights

Intermittent problems
with content server e.g.
Youtube in this case
21
Complaint Handling #2– Upsell premium QoS package

User doesn’t have a

profile suitable for
viewing HD video’s.
Upsell a premium QoS
package to provide
better QoS 22
 Complaint Handling #3 – Customer Overcharged ?

Looking at the detail we

see a number of
downloads from Google
Market are the cause of
the data usage.
23
 Complaint Handling #4 – Populate TT with accurate
customer data for problem resolution

Auto populate ticket to

ensure accurate data for
engineer to resolve
issue.

24
Complaint Handling #4 – Demarcate the problem

25
UC-2 Monitoring Top-up – Individual Retailer

Verifying that this

retailer has been
experiencing a number
of delays with their top
up service

26
 UC-2 Monitoring Top-up – Individual Retailer

Drilling down identifies

the specific transactions
that have been
impacted
27
UC-2 Monitoring Top-up – Individual Retailer

Drilling down on the

specific transaction
identifies delays on the
billing interface. Doing
this for multiple
transactions shows this is
a common problem with
all of the delays

28
UC-2 Monitoring Top-up – Are there other retailers
impacted by this same issue?

Drilling down provides

visibility to which
customers are impacted
delay

29
UC-2 Monitoring Top-up – Analysis for all retailers

Multiple retailers are

impacted by the same
issue. With 4 retailers in
Xian (incl Retailer0561)
impacted.

30
 UC-2 Monitoring Top-up – Analysis for all retailers

Drilling into the

impacted customers
shows the different
retailers in this area
impacted.

31
Individual Retailer
 UC-3: Enterprise SLA Monitoring Use Case
Provider
E-commerce applications require a high quality and reliable real-time mobile services
Business
that perform up to an operators SLA commitments.
Agreement

SLA C bank has implemented an online payment service for their customers. To guarantee
individual account security it is required to enter a verification code (sent via an SMS
SLS by C bank) before confirming the online payment. It is necessary for the user to input
KQI
this code within 5 seconds or the payment transactions will timeout. C bank wants
the operator to guarantee a SLA (e.g. delay, success rate) for all SMS originating from
the C bank’s set of pre-defined number. This is especially critical during holidays and
KPI KPI
special events

Customer
A BANK

Banking Transaction SMS Origination Success Rate

32
 UC-3 Enterprise SLA Monitoring

SMS Success Rate and

Delay have gone into a
warning state. Looking at
the recent history shows
that the declining over a
period of time

Lets drill into the most

recent period to
understand the root cause
behind the decline

33
UC-3 Enterprise SLA Monitoring – Understanding SLA Breaches

Drill down on the specific

regions having the lower
success rate

Failure Analysis shows large

numbers of failures due to
capacity problems specifically
- ‘Submit Message Queue
Full’ and ‘Bandwidth Limit
Exceeded’
34
Drill down on the
specific regions
having the
increased delays

Drill down into the

detailed transactions
shows the specific
transactions timing out
35
 UC-4 Enabling Network Operations to evaluate impact of
Marketing Promotions

Select the criteria to

identify target Target customer
customers segments based on the
36
selection criteia
 UC-4 Enabling Network Operations to evaluate impact of
Marketing Promotions

Adjustable
parameters –
which will be
directly
reflected in the
maps/graphs
on the bottom
of the screen,
to determine
the impact on
the network.

37
 UC-5 PSPU Service Quality Improvement

Page Response and

Page Browsing Success
Rate have breached 38
their thresholds
 UC-5 PSPU Service Quality Improvement

Server problems
constitute the majority
of the failures. Drill
down the specific
sessions impacted

39
 UC-5 PSPU Service Quality Improvement

Analyzing the failures

by web site and per
user identifies a
specific web site i.e.
CNPC

40
UC-6 Real Time VIP Care

41
UC-6 Real Time VIP Care

42
UC-6 Real Time VIP Care

43
44
Thank you.
Prof. Dr. Dipl. Wirtsch. Ing. Mehmet Erdas
MBA B.Sc. M.Sc. METU Ph.D. TU Braunschweig Germany
[email protected]
[email protected]
[email protected]
Mobile: +49 (0)1789035440
+43(0)6509111090
+90(0)5374154413 45