How Does Real-Time Graph Technology Empower Smart Enterprise into New Digital Realm?

Monica Liu (Co-Founder and COO, Ultipa)
Monica Liu (Co-Founder and COO, Ultipa)

Monica Liu (Co-Founder and COO, Ultipa)

As IT evolution and Covid-19 pandemic fostering our world move into the new digital realm at the unprecedented speed, both multi-national companies and SMEs have collected and accumulated enormous amounts of data from transactions, user behaviors, IoTs, etc. How to better utilize data to facilitate business intelligence becomes a crucial pain point that smart enterprises have to face today.

Smart Enterprises Seek Innovative Solutions to Fulfill Their Digital Transformation Needs

Recently, massive innovative technologies like AI engineering, privacy-enhancing computation have been introduced into businesses. Data is viewed as the number one asset for most enterprises. Customers have increasing demands of data processing to facilitate business decision making:

● Higher computing performance to process bigger trunk of data that they collect.

● Agility of data management, given data are stored in all sorts of diverse, distributed, and complex environments. Smart enterprises need to do ultra-deep traversal to find out and analyze connections among data in a compelling number of tables, databases, and systems.

● Easy-to-use experience, since everyone has continuous engagement in digital transformations.

● Innovative solutions that bring new business opportunities or big improvement in efficiency.

● IT upgrades that are more environmentally friendly and cost effective.

Innovative Use Cases That Ultipa Has Successfully Achieved for Smart Enterprises

1,000 Times Performance Improvement of Online Fraud Detection.

For one of Ultipa’s customers, a large retail and commercial bank, which deals with 30 million transactions on average day, real-time fraud identification and handling can be very challenging and resource consuming, because modern anti-fraud process involves countless demands for networked behavior analysis among huge amounts of historical data. Th bank explored solutions like In-memory RDBMS, Apache Spark, and other graph systems, but none of their performance turned out satisfactory. By adopting Ultipa’s RTD (Real-Time Decision) solution, the bank is now processing 90-day transactional data in real-time, largely extended their 7- day scope previously. The processing time for each fraud detection was reduced from 300+ milliseconds to less than 20 milliseconds, and the number of graph-computing models against each transaction was increased from 5 to over 20. In addition, Ultipa achieved this unprecedented performance using a much smaller PC-server cluster (4-instance), which greatly reduced TCO (Total Cost of Ownership), and carbon-emission (compared to 40-instance previously).

36,000 Times Faster in Ultimate Beneficial Owners Identification

Regulators, banks and stock exchanges around the world are penetrating companies holding structure to identify their UBOs (ultimate beneficial owners) or fraudulent activities. A stock exchange customer wanted to examine all links between their 4,000 listed companies and 10,000 key people. The computational complexity is unrealistically high, because one query between a public company and a key person may have 1,000 links (or shortest paths), same kind of queries have to be conducted 40,000,000 (4,000x10,000) times, and the resulting paths could be in the range of 40,000,000,000. The customer understood that this scenario could only be addressed by graph database, but even using Neo4j, for some complex queries, it still took hours to retrieve UBOs or ultimate investment paths. Ultipa is designed to be thousands of times faster than Neo4j. By utilizing Ultipa, the customer retrieves the same complex queries within 0.2 seconds to identify a company’s UBO that sits 30-hop away while Neo4j takes 2-hour to finish. The performance gain is 36,000 times.

World’s First Graph-based Real-time Liquidity Risk Management

Banking customers complained about traditional liquidity management systems that were slow, black-box, incapable of attribution analysis or stress-testing. These liquidity management systems were built using RDBMS (i.e., Oracle) that were not designed to handle big number of transactions in real-time. For a large bank, having 350 million daily transactions, Oracle-based system can only deliver Liquidity Coverage Ratio (LCR) in 3.5 hours, while Ultipa’s lightning-fast graph solution calculates LCR in real-time (1 seconds), which is 12,600 times faster. In addition, the customer is now free to do real-time multi-day liquidity comparison, real-time LCR contribution and impact analysis, real-time back-testing, scenario-simulation, and stress-testing in a white-box explainable fashion, and highly interactive user interface. All analysis can be conducted to the finest granularity based on any single transaction and any penny of any bank account.

Smart Prediction: Bank-wide Credit Card Turnover Prediction Accuracy Increased by 50%

A major retail bank uses ML (machine learning) based prediction method for monthly credit card turnover with a mismatch rate over 2.2%, which equals ~$1 billion. This inaccuracy affects the bank’s liquidity arrangement and profitability. The process is lagging - the ML system takes weeks to do data sampling and training, making the entire process absurd for prediction. The bank’s IT decided to leverage Ultipa for accelerated and intelligent sea-volume data processing ability. They achieved 100x improvement in data sampling and data training speed, the accuracy was improved by 50% (mismatch ratio from 2.2% down to 1%), and hardware investment was reduced from 20-instance big-data cluster to 3-instance HTAP graph cluster.

 ​How to better utilize data to facilitate business intelligence becomes a crucial pain point that smart enterprises have to face today 

Smart Email Analytics Help Insurance Company Save Millions of Dollars

A renowned reinsurance company employed Ultipa Graph database’s lightning-fast ultra-deep traverse and full-text search features to search from tons of email communications among its employees, agents, and insurers to identify critical clauses, supporting evidence and other contextual information. With the help of Ultipa SEA (Smart Email Analytics) toolkits, the customer was able to effectively identify thousands of cases of unqualified insurance payouts, fraud claims, and saved millions of dollars in a relatively short time.

Smart Health Knowledge Graph in Individualized CDM (Chronic Disease Management)

Without personalized CDM, patients will never receive the desired level of disease management they deserve. Running individual patient’s data through the medical logics represented in knowledge graph closes the loop, making it possible to generate individualized and real-time CDM advice (treatment plans, recommendation and more). By integrating Ultipa’s high performance graph computing power, healthcare providers could implement graph augmented healthcare intelligence to provide better services to millions of diabetic patients with personalized behavior management.