Accelerating data utilization for financial crime prevention with Alteryx at MUFG

In the field of financial crime prevention, it is necessary to respond swiftly and accurately to regulations that are added almost yearly by different countries worldwide.

Mr. Yoshiro Fujimori of the Global Financial Crimes Division (GFCD) at MUFG Bank explains, “In data utilization, four perspectives are required: be better, be faster, work without mistakes and provide explanations, or assume accountability.”

To be better, although knowledge regarding financial crime and data utilization is evolving, one challenge is that this knowledge cannot always be fully utilized due to disparate personnel being responsible.

To be faster, agile development is desirable, allowing users to independently develop based on accessible data, rather than following a waterfall development model based on annual work plans.

To work without mistakes, there are many situations where sensitive data is used, and testing with production data is limited. To this end, a shift-left approach is required to conduct tests with real data in early development stages.

Providing explanations and assuming accountability requires that source code be made more digestible. Programming language is difficult to read verbatim and specifications are often voluminous and difficult to understand.

To address these challenges, it’s effective to conduct analysis based on the data mining framework “CRISP-DM.”

Specifically, the business department handles the business understanding phase, whil the IT department is responsible for everything from defining requirements to data preparation. Data scientists are involved in the subsequent machine learning and analysis stages. Since different departments are responsible for each phase, it’s crucial to ensure smooth handover and communication.

“Each department is responsible for its own specialized area, making it difficult to substitute them when coordinating tasks. Not to mention, there is always a risk of communication lapse when sharing information between processes,” explains Mr. Fujimori.

Of course, Subject Matter Experts (SMEs) in specific fields are extremely busy and do not always have the luxury of time to study. To this end, by gradually advancing the use of data in operations, business departments are slowly becoming proficient in the IT domain, aiming to further build skills related to IT and data.

Alteryx enables natural, business-driven skill retention

The Global Financial Crimes Division (GFCD) at MUFG Bank has been implementing an in-house training program for the past three years.

Mr. Fujimori details, “The key point in personnel development is ‘low learning cost.’ We must solve problems using tools during work intervals and then expand into other applications. In this regard, Alteryx is very approachable, even for beginners.”

The program operates on an annual schedule starting in May and ending in March of the following year, initially focusing on improving operating environments and acquiring basic skills. Participants learn everything from basics to applications in the subsequent theory and practice sessions, and in September, a workshop-centric “Innovation Day” is held at the Alteryx facility.

After approximately four months of development and implementation, results are presented in February, where outstanding cases are recognized.

Last year, nine business cases were selected, and each team received 30 minutes of support weekly. A specific example using Alteryx is the automation of false positive checks in sanctioned person screening systems.

Mr. Fujimori explains, “Previously, tasks that involved visual inspection using Excel were implemented through a process to empirically extract patterns prone to false positives and check them. Each logic was designed as an independent macro, and by integrating these macros, reports can now be generated upon input.”

In the data assessment case study, MUFG Bank solved the problem of automating the manual verification process of access log storage, in which a large unmatched percentage appeared after automation.

“It’s now possible to visualize the number of cases processed before and after each flow in real time, including how many cases were processed and how much unmatched data occurred. Aggregated results and number of cases during processing can also now be effectively used as indicators in the data assessment phase,” adds Mr. Fujimori.

In the example of automating Excel tasks, the integration of multiple files is managed within a single workflow, preventing human errors like accidentally opening the wrong file. During output, sheets can be divided by task, reducing mistakes and enhancing data accuracy.

Future challenges and opportunities

“There are cases where the tools stop being used immediately after training ends. Follow up post-training will be necessary to ensure they continue to be used,” states Mr. Fujimori.

The department is advancing efforts to quantify usage status and is considering the introduction of Customer Managed Telemetry, which would aggregate Alteryx Designer execution logs recorded individually on PCs to a network drive, enabling analysis.