During this webinar you will learn:
How new advanced fraud detection models, including clustering, data/text mining, machine learning and network analysis can detect more suspicious transactions and behaviours
How workflow decision learning will make your system smarter by learning based on previous decisions and interactions
How batch file attachments can be used to attach invoices, receipts and other documentation to alerts for proper record keeping during investigations
Our new search feature that allows organizations to search alerts, work items, cases, regulatory reports, comments and attachments, as well as data from outside sources, to look for potential risks (for example, searching Export Control Lists to screen for export controlled goods)
How Concur users can now open original images of receipts directly in CaseWare Monitor, making investigations easier
9. • Reduce rework and repetitive decision making
• Configurable to re-perform decisions
• Options to break decisions
Workflow Decision Learning
!
Alert
Learned?
Decision Decision Learned
Decision taken by System
No
Yes
10. Agenda
• What is CaseWare Monitor
• Remediation Enhancements
• Navigation and UX
• Batch Attachments
• Workflow Decision Learning
• Enterprise Search for Investigations
• Real-time Onboarding Due Diligence
• Advanced Analytics
14. Agenda
• What is CaseWare Monitor
• Remediation Enhancements
• Navigation and UX
• Batch Attachments
• Workflow Decision Learning
• Enterprise Search for Investigations
• Real-time Onboarding Due Diligence
• Advanced Analytics
15. Real-Time Onboarding API
• Screen customers/vendors in real-time to gauge risk levels
• Reduce number of post acquisition investigations
List Screening
Create Risk Score
ID Verification
Onboarding System
Data
CaseWare Monitor
16. Periodic Screening
• Keep risk levels of existing customers/vendors current
List Screening Updated Risk Score
Customer/Vendor
Database
Data
CaseWare Monitor
17. Agenda
• What is CaseWare Monitor
• Remediation Enhancements
• Navigation and UX
• Batch Attachments
• Workflow Decision Learning
• Enterprise Search for Investigations
• Real-time Onboarding Due Diligence
• Advanced Analytics
18. Rules-Based is not Enough
There is no way to know all the unknowns
Data you Know
Criminal records
Offender records
Detective experience
Operational systems
Data you Don’t Know
Social Media
Machine Data
Geospatial
Anomaly Detection
AML Predictive Models
Insight and Intelligence
Optimize AML Decisions
Prevent Fraud
Statistically Accurate
Operationally efficient
Cost effective
Compliant
+ =
Structured and active Unstructured and unknown
19. Analytics Roadmap – Part 1
Descriptive (Business Intelligence)
• Focuses on visualization, summary and
delivery of information
• Dependent on clean and organized data
• Based upon structured historical data
• Trends and forecasts provide forward
looking insight
• Dashboards, scorecards, real-time
monitoring are part of solution
Descriptive Analytics
What happened?
How many, often and where?
What exactly is the problem?
What happens if trend continue? Forecasting
AnalyticsSophistication
20. Analytics Roadmap – Part 2
Predictive Analytics (Predictive Modeling)
• Current and historical information via
models predict an outcome based upon
a set of inputs or conditions.
• Leverages multiple types of data
including large unstructured data sets
• Less dependent on high quality or
organized data
• Uses mathematical/statistical algorithms
• Provides answers like “Based upon these
factors, this is going to happen..”
Descriptive Analytics
What happened?
How many, often and where?
What exactly is the problem?
What happens if trend continue? Forecasting
Predictive Analytics
How do I prioritize this?
What will happen next if…..?
What is the probability this will happen?
AnalyticsSophistication
21. Analytics Roadmap – Part 3
Prescriptive Analytics (Optimization)
• Models and data simulate or optimize a
course of action based upon a set of
rules, models and business process
• Built upon rules engines and algorithms
to simulate probable outcomes
• Leverages multiple types of data
• Usually fed by specific data and models
• Ex: Simulate a scenario multiple times
with multiple factors to identify the
fastest, cheapest, safest option
Descriptive Analytics
What happened?
How many, often and where?
What exactly is the problem?
What happens if trend continue? Forecasting
Predictive Analytics
How do I prioritize this?
What will happen next if…..?
What is the probability this will happen?
Prescriptive Analytics
How can we achieve the best outcome?
What order should I address this?
What should I do if I am constrained by this?
AnalyticsSophistication
22. Advanced Analytics Capabilities
Capabilities
• Deploy advanced analytic
platform for AML, fraud that
can scale and provide
capabilities for:
• Predictive Analytics
• Segmentation / Cluster
Analytics
• Anomaly Detection
• AML Predictive
• Text Analytics
• Social Media Analytics
• Content Analytics
• Increase maturity and ability
to utilize more of the
available data (e.g.
unstructured data)
Benefits
• Shorter model development
and refresh times
• Predictive models and insights
into future outcomes. E.g.:
predict fraud, risk, up-take,
arrears, AML predictive
models, optimize AML
decision etc…
• Augmentation of current
systems with new models and
data
• Segmentation / cluster models
focused on risk, behavior,
demographics, etc…
• Insight into client sentiment
Solution Characteristics
• Platform supports Advanced
models
• Hadoop extension
• Real-time support
• Case management feeds
machine learning
• Fuse alerts, risk scoring to set
priorities
23. Anomaly Detection
• Detects changes in behavior based
on history
• Clusters to detect behavior that is
drastically different from similar
peers
24. Predictive Analytics
• Using known fraudulent cases to identify the patterns that normally
results in fraud
• Using machine learning to train models, predict outcomes and establish
risk levels
Gender
Male
Credit Score
> 800
Low Risk
Credit Score
< 800
Age < 30 High Risk
Age > 30 Low Risk
Female
Credit Score
< 600
Age < 25 High Risk
Age > 25 Low Risk
Credit Score
> 600
Low Risk
25. Network Linking
• Many frauds involve collusion
• Identifying associations by link
analysis creates significant
insights
• Using static and/or
transactional data
• Risk inheritance based on
strength and distance of
associations
27. Procurement Use Cases
• Vendor submitting non-PO invoices when they and others in
their cluster always submit PO invoices.
• Identifying strong correlation between requestor and vendor
that is different from other vendors in the same cluster.
• Using network link analysis to detect unknown relationship
between vendor and employees.
28. Procurement Use Cases
• Identifying invoices for product classes that are outliers based
on vendor type, existence of PO or price ranges.
• Identifying unusually high contract wins for a vendor when
compared to others in the cluster.
• Data enrichment with third party sources: e.g. World-Check,
D&B, Country Corruption Index, Tax Haven Locations, etc.
29. Key Takeaways
• CaseWare Monitor offers
– Smart detection of issues and anomalies to detect complex issues and
reduce false positives
– Uncover schemes that could otherwise go undetected using rules
based only
– Business engagement for what is important and to reduce effort
– Investigate issues easily and thoroughly for comprehensive resolutions
– Machine learning capabilities within the model to continuously
improve processes