garanti bank´s journey to big data - oracle · big data discovery projects customer network...
TRANSCRIPT
Private document of GT
Garanti Bank´s Journey to Big Data
Ayşen Büyükakın
Business Intelligence & Analytics Unit Manager
Private document of GT
...through a planned change journey in BI & Analytics
2004200320022001200019991998
Bankwide Reporting Concept Infrastructural DevelopmentExpanding BI & Analytics
Platform Investment in Fast Computing and
New Analytical Tools
Consultancyformanagementreporting
Segmentation
Management ReportingPlatform
Dashboards
Campaign & CommunicationManagement
Strategic change projects…….
…. Enabled by technology
Single version of Truth
Aggregationmethodology
Analytical platform for Data Mining
ETLs Financial Data Warehouse
Campaign Management System
Finalist in Information Awards
AnalyticallySupportedRetentionPrograms
Profiling on Issuing & Acquring
Propensity models Attrition models EXADATA SAS product suite
First Management Dashboard Financial Analysis Report
EXADATA SAS product
suite
2
Private document of GT
...through a planned change journey in BI & Analytics
3
20162015201320082007
Scoring engine forlending processes
New modelingtechniques
Streaming Social data & useof optimization
Big Data ArchitectureMachine Learning &
Event processing
Strategic change projects…….
Price predciton Models Pricing Simulations AUTONOMY
Sentiment & Categorization
Data Pool Introduction of Big Data
technologies Daily populated
Analytical DW
AnalyticallySupported Sales
Efficiency in ATM replenishment
Analyticallysupported Risk Decisions
PersonalizedPricing
Social CRM Unsctructured data
Utilization of Big Data
Federation of Data
Rısk Decision DW Application scorecards Behaviour scorecards Risk capacity Models Fraud Models
Marketing Decision DW Propensity V2 Income prediction ATM replenishment Models
RTD OEP SNA DNA Volume Model
…. Enabled by technology
Fraud decisionmanagementaward
2014
Private document of GT
Identify / FormulateProblem Gather /
PrepareAnalytics
Data
ExploreData
TransformData
BuildModels
ValidateModels
Deploy / ExecuteModels
MonitorModels
PredictiveAnalytics LifeCycle
Oracle BI
DVD
BDA
SAS EM v14.1 & EG v7.1
Oracle BDD
Oracle DVD
SAS EM v14.1 & EG v7.1
SAS Visual Analytics
R w Oracle Advaced Analytics
SAS EM v14.1 & EG v7.1
R w Oracle Advanced Analytics
SAS EM v14.1 & EG v7.1
SAS Visual Statistics
SAS EM v14.1 & EG v7.1
FICO Blaze Advisor
Model Repository - Inhouse
SAS Decision Manager v14.1
Predictive Analytics LifeCycle - Tools
Data Engineer
Analytical DW & Big Data
Management
Model Deployment
Model Execution
Model Monitoring
Report Design
R&D Activities
Data Scientist
Data Dictionary Management
Exploratory Data Analysis
Predictive Modelling
Model Validation
Model Monitoring
4
Private document of GT
Predictive Analytics Lifecycle – Platforms
5
Data Store
Hive
BDA
Data
Warehouse
Exadata
On Demand
Ma
ste
r D
ata
/ O
pe
ratio
na
lD
ata
On Demand
Data
Enriched
Data
SAS
Discovery & Exploration
SAS VA
R
SQL / HSQL
Modelling
R StudioOracle R Advanced
Analytics for Hadoop
SAS Visual
Statistics
Oracle R Enterprise
on Exadata
R
R SAS
Studio
SAS In-Memory
Statistics for Hadoop
Enterprise Guide
& Miner
SAS
Analytics Laboratory
Oracle Big Data
Discovery
Scoring
R
SQLite SQL
FICO Blaze
Advisor
Java
R
Decision Tree
Text Miner
SAS
Production
Mo
de
l M
on
ito
rin
g
Changed Data
Model
Repository
SAS Decision
Manager
Deployment
Data
Qu
alit
y
Private document of GT
Analytical Inventory
Marketing
Propensity ChurnIncome
PredictionVolume
Estimation
TermDeposit
Pricing & Simulation
Risk
Application Scorecard
BehaviorScorecard
FraudDetection
Collection Optimization
Process
ATM Replenish.
6
Private document of GT
7 Pillars
7
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomerNetwork
Urban Analytics
CustomerDNA
Private document of GT
7 Pillars
8
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomerNetwork
Urban Analytics
CustomerDNA
Private document of GT
Data Pool
9
Customer Complaints
Social Media data
Speech – to – text
System and Application
Logs
Channel logs (ATM,
Internet, IVR, CC )
Click Stream Data
Credit Card Behaviour
Customer Financials
Customer Behaviour
Credit Bureau
Smart Application location
Structured Data Semi Structured Unstructured
Private document of GT
10
Social Media Monitoring
Keeps socialmedia data in a
distributedqueue
(Staging Storage)
Apache Kafka Cluster
Big Data Appliance
CTG
Exalytics
Categorize socialmedia streamingdata and applybusiness rules
Apache SparkCluster
ConsumeContent
Apache HBASE Cluster
InsertContent
Index socialmedia data andprovide search
dashboard
Apache SOLRCluster
PropagateData
Manage socialmedia reports
with usingHBASE table
data
OBI 12C
HTTPS Rest Services
TCP
UNIX Server
Managebusiness rules
for social media
Drools (BRMS)SAS
Categorizationof tweets
SAS Text Miner
SOAP WS
Dat
a D
isp
atch
er
Private document of GT
7 Pillars
11
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomerNetwork
Urban Analytics
CustomerDNA
Private document of GT
Sandboxing
• Citizen data scientists
• Self - service data preperation
• Self - service reporting
• Data visualization
• Storytelling
12
Private document of GT
7 Pillars
13
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomer
SNA
Urban Analytics
CustomerDNA
Private document of GT
Real Time Decisions - RTD
A solution that addresses a business
issue faced by all organizations :
…how to make personalized and accurate
decisions, using the most up to date
information, in real time…
…consistently and in large volumes.
Real-Time Closed Loop
Self-Learning
Business Process ExecutionBusiness Process Execution
INFORMANTProcess data
ADVISORRecommendations
DATA SUPPLIERS
Enterprise Information Model
OLTP DW Grid
14
Private document of GT
15
Next Best Offer
In Garanti Bank’s public page (http://www.garanti.com.tr/en/ ), there is a link for Internet Banking. Garanti customers login by clicking and supplying necessary credentials.
RTD implemented
Training OOT Test %25 sales increase
Private document of GT
7 Pillars
16
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomerNetwork
Urban Analytics
CustomerDNA
Private document of GT
17
Alert Systems
Start Date End Date Company
# of compl
Private document of GT
18
ATM Journal - Oracle Event Processing (OEP)
Private document of GT
7 Pillars
19
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomerNetwork
Urban Analytics
CustomerDNA
Private document of GT
20
Big Data Discovery Projects
CustomerNetwork
CustomerDNA
Social
Warning
Systems
Demographics
Financial
Lifestyle
Life Stage
Private document of GT
Financial Network
Network Analysis of Delinquent
and LabeledCustomers
Sector andCustomer
Type BasedNetwork Analysis
Network Analysis for
DifferentUsage Areas
Demographics & SocialNetwork
Data Quality
improved
New Relatives
Found
SocialRelationships
21
Customer Network
Private document of GT
Goal
Identification of customers’ relational networks using customer
demographics, marital status, spouse info
22
Demographics Networks
Totally 17 million customers are analyzed
Data Description
Customer demographics and National-Id database are utilized
Active retail customers with verified demographical information are used
Private document of GT
23
Current spouse information of customers are gathered via National-Id database
40.23% of 1.5M newly identified spouse pairs can be added to relational database since both customers are found to be active GB customers
42.4K of single, widowed or divorced customers are found to be married in National-Id database
Demographic Networks – Customer Relations vs. National-Id Database
New relations between GB customers are identified
Private document of GT
24
81.8% of current parent-children relations are verified
39.9% of current sibling relations are verified
Customer Relationship vs. National-Id Database
4M new sibling relations are identified in addition to current 2.9K
3.4M new parent-children relations are identified in addition to current 198K
New Relationship Identification
Matching Affinity Relations with Locations
Among customers with affinity relations (excluding spouse relations);3.9% of them live in the same house / apartment,8% of them live in same neighborhood,23.4% of them live in same district
Demographic Networks – Other Relations
Private document of GT
Goal
To define financial network among customers using money transfer
data
25
Financial Network
SME &Commercial Customers’ transaction data is chosen to build thenetwork.
Data
The network is built for 3 monthperiods to examine seasonalchanges.
Cheque and Money transfer transactions among customers aretaken into account.
Private document of GT
26
Financial Network- SME’s in Tourism Sector – Labels
2.2.2016 3.2.2016 9.2.2016 25.2.2016 15.3.2016 22.3.2016
Customer 1 DelinquentLegal
Customer 2 Customer 3 Customer 5 Customer 6Customer 4
Money received from Cust1 /
Total money received%100 %100 %80 %75 %60
Money sent to Cust by Cus1 /
Total money sent by Cust1%29 %2 %2 %2 %2
%25
Etki %si = %25
%25
%10
%10
%5
Private document of GT
27
Customer DNA- Life Style
Kart Profilleme
Luxuries
Essentials
Fashion
Tech-geek
Weekend Lovers
Traveller
Stable
SummerHouse
Credit Card
BehaviorCampaign Preference
CampaignLovers
Curious
Installment
Bonus Seekers
Private document of GT
28
Customer DNA Analyses with Oracle Big Data Discovery
Private document of GT
29
Goal
• Identify life stage of customer by analyzing credit card and ATM usage data
Findings
Moving
o Decrease in ATM & credit card usage in current district
o Increase and stability in ATM & credit card usage in
another district
Getting Married
o Decrease in Food & Drink and Education categories
o Increase in Home & Decoration category
Expecting a Baby
o Decrease in Nightclub & Bar category
o Increase in Infant & Child Clothing and Medical Testing
categories
Customer DNA – Life Stage
0
10
20
30
40
50
Moving Customer
Sabitlik DeğişiklikStability Change
Private document of GT
7 Pillars
30
Next Best Offer
AlertSystem
Data Pool
SandboxingCustomer
SNA
Urban Analytics
CustomerDNA
Private document of GT
Urban Analytics
1. Analyze spatial data to create a better market intelligence for the branches
2. Increase sales team’sproductivity
3. Intellegent marketing campaignmanagement
31
Private document of GT
Urban Analytics - Implementation
32
Private document of GT
Urban Analytics - Implementation
33