STRICTLY PRIVATE AND CONFIDENTIAL
Making Advanced Analytics Better With Graph
An Introduction to TigerGraph
Pawan K Mall
Senior Solution Engineer
Founded in 2012 , HQ in
Redwood City, California
Global Presence - 2022
40 Countries covered
Enterprise Scale,
40-300x faster than
competition
Foundational for AI
and ML solutions
Global Presence
Series C Funded -($105Mn-Tiger
Global Management)
Total Funding : 170Mn+
APJ Presence
Operations in India,
Singapore, Malaysia, S Korea,
China, Japan, Philippines,
Indonesia & ANZ
CONFIDENTIAL
Workload drives database choice
Key value
Document
Relational
Graph
High speed storage/retrieval apps
Front-end apps with
unstructured data
Transactional systems of record
with structured data
AI/ML powered apps and
wider, deeper analytics
Order1
Product
Order2
Product
Customer
Customer
●●●●
●●●●
●●●●●●
●●●●
Customer
●●●●
●●●●●●
●●●●
Location 1
Payment
●●●●
Order
Order
●●●●
Order3
Merchant
Order4
●●●●
●●●●●●
●●●●
●●●●
●●●●●●
●●●●
Merchant
Supplier
Supplier
•
Supplier
Merchant
Product
●●●●
Shopping carts, real-time bidding,
messaging and chats
•
Gaming, content management,
mobile apps, IoT
●●●●
•
Queries
Simple
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
Location 2
Customer billing, payroll,
inventory monitoring, bank accounts
•
Fraud detection, compliance, supply chain, customer 360,
predictive maintenance
Complex
3
1. Connect
CONFIDENTIAL
Relational databases don’t tell the whole story
Relational databases store facts in tables
Graph databases view the world as it is
David
Name
Location
Product
Bank
⚫
John
Palm Springs
iPhone
JPM Chase
⚫
David
Los Angeles
MacBook
Capital One
⚫
Lisa
San Francisco
Watch
JPM Chase
⚫
Jim
Palm Springs
iPad
Capital One
⚫
Sally
San Francisco
Watch
HSBC
⚫
Steve
Los Angeles
iPhone
JPM Chase
⚫
Lisa
Siblings
iPhone
Capital One
San
Francisco
Resides
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
Resides
Jim
JPM Chase
HSBC
Watch
Coworkers
Steve
Limited scalability for analytics
Los
Angeles
Palm Springs
John
MacBook
• Cannot easily model indirect relationships
• Cannot run queries across data sets without slow joins
• Cannot add new relationships without schema changes
iPad
Sally
• Flips the perspective from facts to relationships
• Scales massively without sampling
• Faster queries at greater depth
Rich, intuitive analysis
5
Deep Link Analysis Example
Fraud Detection in Financial Services(Payments)
Used_with
Phone_number 1
Device 1 (phone)
Sends_payment
Sets_Up
Hop 1
User 1
Account 1
Payment 1
Hop 2
Email 1
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Credit Card
Learn more by reading this article: How the World’s Largest Banks Use Graph Analytics to Fight Fraud
Technology Difference
Regular Analytics (Shallow)
vs
Advanced Analytics(Deep) with TigerGraph
Phone_number 2
Hop 5
Device 101
Used_with
Phone_number 1
Hop 6
Phone_number 101
Has
Hop 4
Device 1 (phone)
Account 101
Stolen Credit Card
Sends_payment
Sets_Up
User 1
Account 1
Sets_Up
User 2
Hop 1
Hop 3
Payment 1
Payment 101
Hop 2
User 101
New accounts 1 & 2 - linked back to device
101 used for prior fraudulent payment 101
& account 101 - Payment 1 rejected!
User 1 & User 2 flagged for investigation.
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Email 1
Credit Card
Account 2
Bank
7
2. Graph Algorithms
7 Key Data Science Capabilities Powered By A Native Parallel Graph
Multi-dimensional Entity &
Pattern Matching
Deep Link Analysis
A
1
2
Query Pattern P
Relational Commonality
Discovery and Computation
3
A
Hub & Community Detection
4
Community 1
Match
B
C
D
From a set of entities (e.g. customers,
accounts, doctors), show all links or
connections
B
Given a pattern (e.g. a type of
suspicious activity), find similar
patterns in the graph
Given 2 entities (e.g. customers,
merchants, devices), follow their
relationships to find commonality
Community 2
Find most influential members
(customers, doctors, citizens) & detect
community around them
5
Geospatial Graph Analysis
Analyze changes in entities & relationships with location data
6
Temporal (Time-Series) Graph Analysis
Analyze changes in entities & relationships over time
7
Machine Learning Feature Generation & Explainable AI
Extract graph-based features to feed as training data for machine
learning; Power Explainable AI
© 2022 TigerGraph. All Rights Reserved
Centrality
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
10
Community
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
11
Similarity
Same as
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Algorithm Types
❏ Centrality
Assign numbers or rankings to each vertex corresponding to their network position
❏ Classification
Classify the vertices into sets according to some external rule
❏ Community
Group the vertices so that each group is densely connected
❏ GraphML/Embeddings
Convert the neighborhood topology of each vertex into a fixed size vector of decimal values
❏ Path
Find the best paths from one vertex to another (shortest, lowest weight, or other criteria)
❏ Similarity
Compute similarity between pairs of items
❏ Topological Link Prediction
Predict the existence of a link between two entities in a network
❏ Frequent Pattern Mining
Find subgraph patterns that occur the most frequently
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
13
Algorithms
❏
❏
Centrality
■ PageRank
■ Article Rank
■ Betweenness
■ Closeness
■ Degree
■ Eigenvector
■ Harmonic
■ Influence Maximization
Community
■ Connected Components
■ K Core
■ K Means
■ Label Propagation
■ Local Cluster Coefficient
■ Louvain
■ Speaker-Listener Label
Propagation
■ Triangle Counting
❏
❏
❏
GraphML/Embeddings
■ FastRP
■ Node2Vec
Path
■ Astar_shortest_path
■ BFS
■ Cycle_detection
■ Estimated_diameter
■ Maxflow
■ Minimum_spanning_forest
■ Minimum_spanning_tree
■ Shortest Path
Classification
■ Greedy Graph Coloring
■ Maximal_independent_set
❏
Similarity
■ Cosine
■ Jaccard
■ K Nearest Neighbors
■ Approximate Nearest
Neighbors
❏
Topological Link Prediction
■ Adamic Adar
■ Common Neighbors
■ Preferential Attachment
■ Resource Allocation
■ Same Community
■ Total Neighbors
Frequent Pattern Mining
■ Apriori
❏
Some of these, such as Shortest Path and PageRank, have
several variations, so the total number is over 50.
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
14
3. Improve ML with Graph
features
Machine Learning has reached a limit in accuracy
● You can only squeeze so
much out of what you
have.
● Tuning your ML algorithms
can only yield diminishing
returns.
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
16
Top banks have used graph features to get a
quantum leap in detection rates
Core
Systems
4 tier 1 US Banks are using TigerGraph in production to
generate Graph Features for their existing AI fraud
detection systems.
●
●
●
One bank reported a 20% increase in synthetic
identity fraud detection using TigerGraph.
Another told us that TigerGraph was by far the best
technology ROI from their technology investments
that year, and it is making them $100 million pa.
JPMorgan Chase gave TigerGraph the coveted Hall of
Innovation Award 2021.
Data
Lake
Graph Database
ML Feature
Generation
Graph Features
Fraud Detection
Application
Investigation
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
17
ML Generalisation
Data
Features
Model
Output
Conventional feature
engineering pipeline
Combine
Features
Structure as a graph
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
Graph
feature
engineering
ML
Test, Train,
Predict
20%+
Uplift in
Accuracy
18
CONFIDENTIAL
Graph databases make AI/ML models faster and more accurate
Richer, smarter data
Deeper, complex
questions
Accelerated
performance
Explainable results
• Relationships-as-data
• Connect different datasets, break down silos
% of data and analytics
innovations using graph
technologies –
80
%
• Look for semantic patterns (implicit relationships)
• Easily and quickly search far and wide
• High-speed queries
• Relationship powered algorithms and machine
learning
• Intuitive models, queries and answers
• Visual exploration and results
10%
2021
2025
Source: Gartner
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
19
Graph database use cases
Powered by Machine Learning and AI
Supply Chain Management
Fraud Detection
and Prevention
Knowledge Graphs
Investment
Opportunity Analysis
Law Enforcement
Network & IT Resource
Utilization
Recommendation
Systems
Product and Service
Marketing
Drug Reaction Analysis
Entity Resolution
Cybersecurity
Social Network
Analysis
Customer 360
Influencer & Community
Identification
Data Lineage
CONFIDENTIAL
Best of breed graph platform powered by a native distributed GraphDB
Tools
User Interface
AI / ML
ML
Workbench
Performance
Deployment
Graph Data
Science Library
Visualization Toolkits
Customer
360
Financial
Crimes
Connectors
Supply
Chain
Deep analysis
Extreme scalability
Real-time
In-database analytics
Search for patterns
across 5-10+
relationship hops
Automatic storage partitioning
and massive parallel
processing
Query speeds 40-300x faster
than any other graph database
Continuous graph-based
feature generation and training
Self-managed
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
Cloud DBaaS
21
CONFIDENTIAL
Seamless integration with modern cloud infrastructure
BUSINESS APPLICATIONS
DATA SOURCES
Restful APIs
Visual design
DATA USES
GSQL queries
Data Clouds
Machine Learning
ML
Workbench
Enterprise Apps
Graph Storage
Engine
Graph Processing
Engine
Artificial Intelligence
Analytics
Relational DBs
Visualization
Business Intelligence
Streaming
Self-managed
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
Cloud DBaaS
22
Financial Services
Impact Analysis:
Communities or
Clusters impacted
by the fraud rings
8 of the Top
10 Global
Banks Use
TigerGraph
Credit Card Fraud:
Is applicant
connected to
potential fraudsters?
Merchant Analytics:
Transaction
sequencing to detect
geolocation proximity.
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Trade
Surveillance:
Are employees
following the
rules?
Credit Scoring:
Real-time credit
scoring to help
recommend
offers best suited
to customer
profiles?
Wealth
Management:
What Accounts,
HNI to target for
stocks or life
change events.
24
Use Cases
Key Drivers
Strategic Objectives
Example Use Cases
Trade Optimization | Portfolio
Core Business Platform
Increase
Revenue
(make money)
$↑
Customer Experience
(CX)
Digital
Transformation
KYC - Customer Journey
Execs & Mgt
Cross Sell | Upsell
Fraud Analysts
Churn Avoidance
ML & Data Science
Audit | Compliance
$
Business
Value
Roles /
Persona Types
Decrease
Costs
Increase Operational
Efficiency
Fraud Prevention - AML, Cyber, etc
(save
money)
$↓
Risk & Segmentation
Migrate to Cloud
Trade & Sanctions
Disputes & Claims
Mitigate Risk
Material C360
Detection
Audit & Governance - Data Providence
(protect money)
$↔
Audit, Regulatory
Compliance Process & Regulatory
Technical Teams
Graph usecases for financial services
Corporate Bank
Customer 360 - generate
revenue with personalised
and targeted experiences
AML - detect networks of
people and transactions
Fraud Prevention - detect
unlikely payments
KYC - determine the
legitimacy of the customer
identity
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Private Bank
Retail Bank
Financial Crimes Solution Goals
●
Increase Accuracy - Human and Algorithms
(e.g., False Positives)
●
Decrease False Positives - Capture and incorporate
human instincts and expertise (esp for expert users)
●
Increase Speed & Efficiency - More accurate and trusted
outcomes (both automated & human)
●
Adapt to changes - Alert and ‘Quarantine’ transactions on
lesser seen cases, etc.
●
Continuous improvement - A system that can learn,
adapt & get better over time
●
Utilize My Current Tech - May have 3rd Party (Actimize,
Cybersource, etc) - Bolt on and make better
27
Deep Link Analysis Example
Fraud Detection in Financial Services(Payments)
Used_with
Phone_number 1
Device 1 (phone)
Sends_payment
Sets_Up
Hop 1
User 1
Account 1
Payment 1
Hop 2
Email 1
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Credit Card
Learn more by reading this article: How the World’s Largest Banks Use Graph Analytics to Fight Fraud
Technology Difference
Regular Analytics (Shallow)
vs
Advanced Analytics(Deep) with TigerGraph
Phone_number 2
Hop 5
Device 101
Used_with
Phone_number 1
Hop 6
Phone_number 101
Has
Hop 4
Device 1 (phone)
Account 101
Stolen Credit Card
Sends_payment
Sets_Up
User 1
Account 1
Sets_Up
User 2
Hop 1
Hop 3
Payment 1
Payment 101
Hop 2
User 101
New accounts 1 & 2 - linked back to device
101 used for prior fraudulent payment 101
& account 101 - Payment 1 rejected!
User 1 & User 2 flagged for investigation.
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Email 1
Credit Card
Account 2
Bank
29
Real-Time Fraud Detection & Prevention
Customer
Activity
T1
T2
T3
T4 (Fraud)
1
T4
TigerGraph can leverages rules:
●
Basic relational logic
●
Graph based logic
3
T1
T2
Data Ingest,
Matching,
And In-DB
Pattern Matching
1
Location comparison:
● High-risk location
● Compare to previous
location
2
Link analysis: 4 hops to fraud
community
C1
T3
Matched rule
2
FC1
3
High Risk Alert
Alert
Yes
Investigator
Investigate
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
No
Link analysis:
● community detection included in fraud
community
Link analysis: direct relation
with fraud community
30
TigerGraph Fraud Platform vs Other Fraud Applications
Incremental business value
Fraud Capabilities
Entity Resolution
Customer 360
Risk Scoring
Threat/Pattern
Detection
Enable cross-system
matching even when
matching IDs are not
available
Increase fraud
detection accuracy by
using all relevant
customer context
Take a weighted
approach to avoid
high numbers of false
positives
Auto-detect fraud
risks so that they can
be dealt with
appropriately
In-Database Fraud
Detection
Deep-link
In-Database Fraud
Detection
Real-time
Detection at any
Data Volume
Process fraud logic in
the same platform
with data – no data
copy & redundancy
Catch all convoluted
and sophisticated
fraud, not just fraud
that’s easy to spot
Prevent fraud in realtime regardless of the
scale of data required
to catch it
Cannot handle 4+
hops/connections
analysis
Cannot scale to
terabytes of data
Other Graph
Platforms
Legacy Fraud
Detection Application
ACI PRM
Common
issue:
Huge bottleneck to join all information
Traditional Fraud
Detection Application
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
Extensive data preparation process
before using the software
Data duplication
for fraud and other
analysis activities
TigerShield Anti-Fraud Solution Focus
Account
Monitoring
health scanning, risk
Transaction
Monitoring
Fastest Graph Analytic with
Distributed MPP Architecture
for Unlimited Data
(<100ms for 10TB of data)
Easy to Integrate and
consolidate all data sources
(structured + unstructured
data)
Deep Linked Analytics
(10+ hops to find the
hidden Fraudster)
Fastest ROI
Shortest Time to
Implement and Use
Ready to Use Toolkit
Built for and used by
Fraud Investigators
money movement, payer-payee
collusion, account takeover
Fraud Detection
Identify Fraud at account
opening, credit abuse,
synthetic ID
Open Source Algorithms
Highly customizable
(no blackbox and vendor lock-in)
TigerGraph Graph Studio
●
●
●
●
Design schema
Load data
Explore Graph
Run Graph Queries
© 2022 TigerGraph. All Rights Reserved
33
Financial Crimes Workbench
●
●
●
●
Rules / Alerts
Dashboards
Analytics
Case & Alert Mgt
© 2022 TigerGraph. All Rights Reserved
34
AML Explorer Workbench
VALUE PROP
User Specific UI & Workflow
ROI
●
●
●
●
Increase Accuracy & outcomes
Better returns/otherwise missed
opportunities
Deep link patterns for regulatory audits
Cost of manually tracking and complexity
ML + GRAPH MODEL
Temporal deep learning & Connections
●
Deep Links, No-Code Investigation
●
Pattern & Community Algorithms
●
Deep learning/Model interpretability
© 2022 TigerGraph. All Rights Reserved
35
Anti Money Laundering
Cutting Through Layers of Synthetic Identities & Accounts To
Find Money Laundering
Visit the solution page - https://www.tigergraph.com/solutions/anti-money-laundering-aml/
© 2019 TigerGraph. All Rights Reserved
37
Why does Graph create new KYC opportunity in
addition to existing KYC automation?
Step 1: Build the identity graph
Personal ID Data
Robert Evans
r.evans@gmail.com
07777444333
Brand Marketer
PEP List
Robert Evans
07888999888
Corruption List
Rob Evans
r.evans@gmail.com
Law Enforcement Data
Rob Evans
07777444333
bob.e@gmail.com
Social Media Data
Bob Evans
07888999888
bob.e@gmail.com
Age 24
Male
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
1. Data relating to a customer is not always captured
uniformly - which makes it difficult to use all possible
data sources that would enhance KYC.
2. In this example, it is only through capturing indirect
connections through Corruption, Law Enforcement and
PEP lists that the right social media profile can be
identified and related to the Personal ID record.
3. The more indirect connections to the social media profile
from the personal ID data we can find, the more
confident we are that it is the same person.
Without a graph for this exercise, you must logically
anticipate all the potential connections and then develop
computer rules that search for those connections in one
swoop. This quickly becomes a time consuming and clunky
process to run - and can often miss connections or even
timeout before finding them.
Fraud Prevention analytics is enhanced by KYC data - and
graph can uncover fraud currently going undetected
Step 2: Overlay Transaction data onto the identity graph
Consumer Type
$
A
$
Account
1. Once your customer identity is clear from KYC
activities, the customer account begins to transact.
Transaction
$
Merchant Type
A
Geography
$
$
Account
Business Type
$
$
$
Transaction
A
$
A
$
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Payee Business
Type
Geography
2. The context provided by the KYC identity will help to
determine anticipated types of transactions - and
ringfence any unexpected behaviour.
3. For retail bank customers, layering transaction data
such as merchant types and geo-location of spend can
help isolate any unexpected transactions for that KYC
customer type - e.g. this may indicate credit card
fraud.
4. For business customers, layering transaction data
such as payee business category and geography can
help isolate any unexpected transactions for that KYC
customer type - e.g. this may uncover invoice fraud.
This kind of fraud prevention without graph requires
joining many tables - meaning longer processing times,
and a risk that the query will timeout before identifying the
crime.
AML analytics is enhanced by KYC and Fraud data - and graph can
uncover laundering rings currently going undetected
Step 3: Model the network of transactions and identities
1. Your graph can model the journey of all
money movements between accounts.
Co-Directors
A
Source
Account
Destination
Account 1
A
A
A
A
A
A
A
A
Destination
Account 2
Destination
Account 3
Payee
Business
Type
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
$
Geography
2. Combining this journey data with fraud and
KYC data helps to determine whether
money is moving in the right directions for
reasons known to be common to normal
business and consumer activities.
3. Where movement is unexpected, it is
possible that it is a result of financial crime
e.g. money laundering.
Without graph, this kind of financial crime
detection requires joining a very high volume of
data points together - meaning much longer
processing times, and a high risk that the query
will timeout before identifying the crime.
Deploying TigerGraph in Your AML Network
1
2
3
Extract Features
● Suspicious
activity patterns
Heterogeneous data
sources including
NICE Actimize
5
Train ML Model
● In-Graph
or
● Out of Graph
● Neighborhood
graphlets
Connect
datasets and
pipelines
4
Enrich DB
with features
on nodes
Use
model
to predict
money
laundering
User
● Exec
● Analyst | ML
● IT - DBA
TigerGraph
Toolkit
7
6
Test Incoming Transactions
● Insert new transaction
● Extract its features
● Apply model
If money
Issue Alert
laundering
detected
41
Credit Risk
Solution Approach: Knowledge Graph
Core Banking:
Companies Data:
● Customer data
● Transactions
● Company information
● Ownership & Directorship
HR Information System:
● Demographic profile
● Position & transfers
● Performance review history
Geospatial Data:
CREDIT ASSESSMENT
SYSTEMS
Loan System:
● Loan submission
● Collateral information
● Loan approval &
rejection
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
● Region Hierarchy
● Geohash
● Longitude & latitude
Blacklist Information:
● Customer
● Blacklist type
● Source
43
Connected Information for Credit Assessment
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
44
Connected Information for Credit Assessment
Geospatial Data
Companies Data
Loan System
Core Banking
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
Blacklist Information
HR
45
Information
System
Collateral Fraud
1
Credit
Disbursement
Account 1
CS
2
CD
Loan
Application
CO
Process Workflow:
Using graph algorithm Path Finding - Circular Path Detection
1. Find all Credit Disbursement (CD) from specific date range
2. Do traversal analysis from CD to all Loan Applications (LA)
3. Do circular path detection from LA1 -- Branch (B1) -- Area (A) -- B2 -- LA2 -- CO - LA1
Collateral
LA
3
B
Account 2
CS
LA
Loan
Application
Branch-A
A
Area-1
B
Graph traversal pattern above will find all of the loan applications using the same
collateral in same area
Branch-B
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
46
Broker Fraud
4
Transaction
TRX
IE
CS
Account
CS
2
work_in
LA
IE
1
Credit
Disbursement
LA
CD
LA
CD
3
Loan
Application
Internal
Employee
B
ST
Collectibility
Status = BAD
Branch
Process Workflow:
Using graph traversal analysis
1. Find all credit disbursement (CD) from specific date range and traverse to the
account (CD -- CS1)
2. Find all Loan Applications (LA) from each CS that handled by the same sales
(internal employee - IE1)
Traversal CS1 -- LA -- IE1, Group by CS1, IE1 having count(LA) > 3
3. For all LA from #2, check if x% of it has bad collectibility status
LA -- (ST == “Bad”)
4. Graph traversal analysis to check if there is a transfer to internal account related
to the process
CS1 → TRX → CS2 -- IE2 -- B -- IE1, with additional logic:
○ Total transaction > 30% of credit disbursement value
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
47
Customer 360
CONFIDENTIAL
Building A Real-time Customer 360 Data Hub
Business Challenge
Combine all available data for the customer with transactions (orders,
payments, calls, rides..) in real-time to improve business outcomes
Solution
Build on top of current investments in master data management, data
warehouse/Hadoop data lake and NoSQL repositories
Find new relationships among data to drive product & service
marketing, targeted cross-sell and up-sell recommendation, superior
fraud and money laundering detection and improved credit risk scoring
and monitoring,
Analyze temporal (Time Series) and spatial data to find new patterns
and insights
Expand schema (attributes/fields, relationships) to accommodate new
data sources & use cases
Business Benefits
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
Improve customer experience, increase revenue and lower costs
49
Customer Journey - Mapping out the customer engagement and
attribution over time
CONFIDENTIAL
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM
50
TigerGraph for Customer 360
Discover hidden trends and connections with extensive, configurable
data visualization and activity time series.
Acquire
For a deeper look, we offer segmentation for
customer targeting by demographic, along with
the ability to view individual journey maps.
Retain
Avoid churn with customer scoring, augmented
by machine learning and graph to pinpoint
silent complaints and take action immediately.
Grow
With graph and machine learning, you can
utilize customer analytics to bring new patterns
to light that you’ve never seen before.
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
51
All Data Culminates to Customer 360
Where Due Diligence and Compliance Regulations have driven customer information integrity and
richness, there is additional opportunity for revenue generation, experience improvement and cost
reduction from the same data.
Example Customer 360
Use Cases
Corporate Bank
Retail Bank
Upsell/ CrossSell
Private Bank
● Credit Rating
● Company Name
● Company Address
● Directors & Shareholders
● Transactions
● Accounts
● Geographies
● SARs
● Company Ownership
Structures
● Nature and location of
business, products &
services
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
● Personal ID
● Tax #
● Nature and location of
customer income &
spending habits
● PEP Lists
● Criminal Activity/
Corruption lists
● Merchant Geography
● Merchant Credit
Rating
Customer
Experience
Customer
Segmentation
Content
Personalisation
Support
Personalisation
Adoption
Personalisation
Customer 360 for Retail Customer
●
Relation identification
●
●
●
Commonly, not all of the customer
demographic data is updated,
such as the employment status,
employer, etc
TigerGraph customer 360 helps
the identification of the relationship
between corporate customers and
retail customer.
With the rise of digital payment,
TigerGraph can help with the
digital account identification
between the customers to
merchant
Both of the point above will enrich
the retail customers profile for
better analytics and campaign
purposes
Digital account identification
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM | CONFIDENTIAL INFORMATION |
53
Using TigerGraph to Provide Next Best Offer
Combine demographic, personal
preferences, interests, reward, etc to get
full subscriber view. Use Graph algorithm
to score importance of preference and life
value
Graph Algorithms can provide
products bests suited for upsell
and/or cross-sell based on Entity
Resolution and customer
preference
Graph NLP for Sentiment Analysis
and build Knowledge Graph and
Integrate with Chatbot
© 2020. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Market
Vestibulum
congue
Segmentation
Prebuilt Graph Queries with Trained
models can drive real-time
personalized next best action and
recommendation. Ie, customized
offer on mobile app based on recent
purchase
Understand LTV and profitability of
each customer (high value, etc…)
TigerGraph Enriches Consumer 360 Insights
Entity
Resolution
Data Driven
Marketing Campaigns
Data
Democratization
Actionable
Insights
• Take the “Artwork” out of
the equation
• Consistent Decisions
• Augment Decisions with
ML/AI
• Proactive Real-time and
personalized Campaigns
• Front line sales (regional
sales to store managers)
• Marketeers
• Product managers
• Data Analysts & Data
Scientists
© 2020. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
Data
Monetization
• Single entity for
each customer who
views content on
different platforms /
devices
• promote targeted
content and ads to
existing users to
maintain their
customer loyalty
• Powerful and
flexible API to
provide unique data
insights
• Single Pane UI
empowering non
technical users to
drive insights
Expanding the C360 dataset further creates net new
revenue opportunities
Omnichannel Retail Banking
Value Chain Banking
(Consumer-to-HNW, Cons/HNW-toSMB)
-
-
Capitalise on captive market
- increase revenue
Target bundles / develop new
propositions
Life stage upsell/cross-sell
Compete with Fintechs on experience
Marketing ROI/conversion uplift
AML - detect networks of
people and transactions
Fraud Prevention - detect
unlikely payments-out
KYC - determine the
legitimacy of the customer
identity
B2B2C Business Banking
-
Usage-based pricing - enabling new
propositions and higher margins
Better customer service
Division Portfolio Banking
Including additional system data in the Customer
360 will uncover further revenue, cost and
experience opportunities
© 2021. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
-
Upsell/Cross-sell across Portfolio
Portfolio credit/risk scoring
Profitability gains (targeted pricing)
TG STUDIO
TIGERGRAPH EXPLORER
End to End Solution for Customer 360 Use Cases
EXECUTIVES
LINE OF BUSINESS
CUSTOMER FACING
TEAM INVESTIGATION
ANALYTICS
IT | PLATFORM
Dashboards & Reports
Macro Trends
Drill In Elements
Product Specific Dashboards & Reports
Upsell | Cross Sell
Journey
Segmentation
Churn
Analytics
Loyalty
C360 Interaction and Communication
Visual & Non Visual
Alerts & Updates
Investigation Teams - Workbench
Investigations & Teams
Findings & Connectivity
Exploration
Relationship Diagrams
ML & Algorithms
Complex analytics and ‘scenarios’
Graph & Management
Data Models
Graph Ops & Attributes
Core Capabilities
TigerGraph: 3rd Generation Graph Database
Real-time Performance
Deep Link Multi-Hop Analytics
Sub-second response for queries touching
tens of millions of entities/relationships
Queries traverse 10+ hops deep into the graph
performing complex calculations
Ease of Development & Deployment
Transactional (Mutable) Graph
●
●
Hundreds of thousands of updates
per second, Billions of transactions
per day
●
GraphStudio - visual SDK
GSQL - Intuitive, Turing complete graph
query language for developing
complex analytics in days
User extensible graph algorithms library
Enterprise Grade Security
●
●
Scalability for Massive Datasets
100 B+ entities, 1 Trillion+ relationships
© 2021 TigerGraph. All Rights Reserved
59
Encryption Support
Control access to sensitive data based on
user role, dept or organization with
MultiGraph
TigerGraph Distributed Database Architecture
Simple setup, Performant design
●
●
●
Setup: Just tell TigerGraph how many servers.
TigerGraph seamlessly distributes data.
Users see a single database, not shards.
Real-time active
replication
for High Availability (HA)
● write to all
● read from any
● strong consistency
MPP OLAP
OLTP concurrent
ACID transaction
Unlimited
scale-out
simple to expand
$
Simple to setup
and manage
Economical
60
MPP - Distributed Cluster in Single or Distributed mode
Distributed Mode
Server 1
Server 2
Server 3
● The server that receives the
query becomes the master.
● Computations execute on
all servers in parallel.
Distributed Query
Distributed Query
(Master Node)
Distributed
Query
● Global accumulators are
transferred across the
cluster.
● If your query starts from all
or most vertices, use this
mode.
© 2020. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |
61
Enterprise Security: MultiGraph for RBAC and Data Sharing
●
Share & Collaborate
○ Multiple groups share one master
database
⇒ data integration, insights, productivity
●
Real-time, Updatable
Shared updates, no copying
⇒ cleaner, faster, cheaper, safer
●
Fine-Grained Security
○ Each group is granted its own view
○ Each group has its own admin user,
who manages local users' privileges.
62
GSQL Design Features
Schema-Based
Optimizes storage
efficiency and query
speed. Supports dataindependent app/query
development.
Conventional
Control Flow (FOR,
WHILE, IF/ELSE)
Makes it easy to
implement conventional
algorithms
© 2019 TigerGraph. All Rights Reserved
Built-in HighPerformance Parallelism
Achieves fast results
while being easy to code
Procedural Queries
Parameterized queries
are flexible and can be
used to build more
complex queries
63
SQL-Like
Familiar to 1 million
users
Transactional Graph
Updates
HTAP - Hybrid
Transactional / Analytical
Processing with real-time
data updates
The TigerGraph Difference
Feature
Real-Time Deep-Link Querying
5 to 10+ hops
Handling Massive Scale
In-Database Analytics & Machine
Learning
© 2022 TigerGraph. All Rights Reserved
Design Difference
Benefit
● Native Graph design
● C++ engine for high performance
● Storage Architecture
● Uncovers hard-to-find patterns
● Operational, real-time
● HTAP: Transactions+Analytics
● Distributed DB architecture
● Massively parallel processing
● Compressed storage reduces
footprint and messaging
● Integrates all your data
● Automatic partitioning
● Elastic scaling of resource usage
● GSQL: High-level yet Turingcomplete language
● User-extensible graph algorithm
library, runs in-DB
● ACID (OLTP) & Accumulators
(OLAP)
● Avoids transferring data
● Richer graph context
● Graph-based feature extraction for
supervised machine learning
● In-DB machine learning training
● No-code migration from RDBMS
● No-code Visual Query Builder
●
Democratize self-service analytics
to derive new-insights from
legacy/external data stores
Uncompromising Performance & Technology
Latest 36TB LDBC Benchmark available on
https://www.tigergraph.com/benchmark/
Reference Architecture
66
TigerGraph Banking Reference Architecture
Data Sources
Real Time Queries
EDW &
Data Lake
Mobile App / Website
Raw & Normalized data
TigerGraph
Insights Engine
Activity & Usage
Batch or
Real-Time
Streaming
API
Batch or Real-Time
Streaming
Customer
Service/Chatbot
Customer 360 Machine Learning
Transaction
Visualization
Batch or Real-Time Streaming
Model
Recommendation
Management
Engine
Interactions
Campaign Application
Archive Graph Data
Campaign results and logs
Products & Services
© 2022. ALL RIGHTS RESERVED. | TIGERGRAPH.COM |