Whatever Happened to Rosey Jetson? How Banks/ Retailers/ Processors/ Networks Could Use Artificial Intelligence in Day-to- Day Operations

Transcription

1

2 Whatever Happened to Rosey Jetson? How Banks/ Retailers/ Processors/ Networks Could Use Artificial Intelligence in Day-to- Day Operations Kevin Johnson Christopher Souser Thursday March 1 st, :30 5pm MT Session Start:

3 The Jetsons: Rosey with Attitude (Start at 16:17)

4

5 Artificial Intelligence: Robots => Devices

6 Why it Matters to IT / Businesses Any day without a fire is a good day What useless task is going to consume too much of my time today? I ve been meaning to look into that I shut off the alerts I couldn t keep up Stores call in and tell me when they are down I m pretty sure everything is working nobody has called to complain Well, we have SLA s but we can t track them I ve heard rumors about poor interchange performance but can t prove anything The person that implemented that is no longer here

7 Data Turned to Value 24x7 Accessibility of System Performance Transaction Volume (Region / Store) KPI / SLA Enforcement Insight into Interchange Performance Approvals / Declines OS Health (CPU, Disk, Memory, etc) Minimize Expensive Outages Improve User Experience & Acceptance Successful Deployments, Upgrades & Replacements Optimize IT Operations & Resources 24x7 Accessibility of Processing Fraudulent Transactions Per Store / Terminal Transaction Performance Complete Payments Ecosystem Visibility Per Merchant Visibility Card Program Performance

8 Top 10 Alerts 1. Interchange Availability and Performance 2. Individual Issuer Performance 3. Approval Rates 4. Store and Regional Terminal / Transaction Availability 5. Types / Department Revenue Breakdown 6. Fraud 7. Availability of Token Management Service (NFC Transactions) 8. Stored Value / Gift card 9. Supporting Services / Apps 10.OS / Platform (CPU, Disk, Memory) Before 2.35 After Through 2018, I&O organizations with an average ITSIO over Level 3 will realize greater than 20% cost savings in their I&O operations.

9 Roles Tier 1 Support Track open/closed alerts Comprehensive easy to consume OS / payments processing health User configurable UI enabling optimal tech productivity Business Insight Real-time visibility of transaction amounts by type, region, location, brand and more Loyalty and Gift card insight NOC Wallboard 24x7 global view of retail payments Loyalty and Gift card insight Easy to consume key indicators (configurable) Manager Track open/closed alerts (Graphical) Key performance indicators of customer payments Mobile availability Alert / MoM Integration Content dependent alert dispatching to quickly deliver information for resolution Ability to deep dive for additional root cause analysis KPI / SLA Example Payment Applications: - XPNET/BASE24 - XPNET/BASE24- eps - Postilion - CONNEX - Custom Data Store Dashboards: Visualize your real-time transactions even at 2am (example) Thresholds: Recognize when a specific terminal isn t processing transactions (example) Alerting: Notify individual(s) to alert that action is required Automation: Initiate a command and / or reset a process / server when needed Replay: Store and replay data as if you were watching the event happen in real-time JASI SNMP SOAP AXL HTTP Traps In Web Services WMI SQL Socket CLI Log Parse ISO8583 On- Board Retail POS channels Cards Interchanges Per Store Trans Platform Windows UNIX LINUX NonStop (including X, virtual, etc) Fraud Rules Reviewers False Positive Rates Fraud Patterns ATM Availability Faults Cash Levels Service Providers Processors / Networks ATM/POS channels Cards (in Apple,NFC) Multiple Merchants FI/Bank Credit / Debt Trans Interchanges / Host Transaction Time Exec Dashboard

10

11

12 Evolution of Monitoring Thresholds: Using industry best practices, the system is configured such that when values occur outside of static ranges, notifications are sent to various individuals. Value ranges, escalations, # of occurrences, multivariable analysis, etc. Dynamic Thresholding: Insight & Trending that understands what is expected and alerts when data points occur outside of what is expected. Limited to single value correlation. Artificial Intelligence: Machine learning algorithms that profile normal behavior that correlates and alerts on activity that deviates from the profiled activity. Yesterday Today Tomorrow

13 80 Daily Transactional Rate (TPS) 90 Daily Transactional Rate (TPS) Business Hours Non-Business Hours Non-Business Hours 4am 5am 6am 7am 8am 9am 10am 11am12pm 1pm 2pm 3pm 4pm 5pm 6pm 7pm 8pm 9pm 10pm 0 Top Range Actual Low Range Actual Static Threshold Limitations Statically configured typically aligned with business hours, static thresholds are time consuming to initially align with the business and don t evolve as the business grows. Learning your Environment Learns the transaction patterns based on historical data and automatically adjusts the thresholds accordingly. Notification of potential issues at the first sign of trouble, giving you time to repair the problem before it affects your customers.

14 Innovation: Machine learning AI Engine Machine learning engine that analyzes and correlates variables per instance to detect normal operating behavior and alerts when symptoms occur outside of the norm Benefits Significantly reduces the detection time of complex problematic behavior (intentional or unintentional) Minimal setup time that maximizes desired solution performance Significantly reduces false positives / alerts Never worry about knowledge lost as employees change roles

15 AI Prediction: WHEN IT IS LIKELY TO HAPPEN? Prediction Use historical data to predict when potential events might occur. Inform Inform the technical and business users when the ecosystem is processing outside of expected operation parameters. Complexity Looks at multiple input variables to determine when something is likely to happen to facilitate preventative maintenance and optimizations of servicing resources.

16 Artificial Intelligence: Going Beyond Automated Analysis Real time Analysis of complex Traditional and Non-Traditional Data to go beyond rule analysis. Inform Inform the technical and business users when the ecosystem is processing outside of expected operation parameters. Complexity Looks across defined criteria and period to provide multiple input variables to calculate auto-adjusting criteria based on business thresholds.

17 The Evolution of Thresholding and Alerting Simple Threshold Thresholding with Recurrence Multiple Data Source Correlation Dynamic Thresholds Artificial Intelligence

18 Summary 1. Define key successful criteria Transaction rate < 10 tps Decline rate > 5% 2. Get a handle on where the data currently is 3. Get visibility and alerting on the data that aligns with the key success criteria (reference top 10 alerts) 4. Leverage dynamic thresholding advancements Kevin Johnson IR - Prognosis Kevin.Johnson@ir.com Christopher Souser IR - Prognosis Solutions Consultant Leverage existing knowledge / solutions / resources and ASK QUESTIONS

19