By Emma Yu, Product Marketing Director, Oracle
For all of the enthusiasm and intrigue surrounding the field of data science in the last several years, data cleansing usually doesn’t make the news. Yet, did you know that 80% of the time and effort in data science involves cleaning, collecting, and organizing the data?
If this is news to you, you’re not alone. Last month at Modern Finance Experience, audience members (primarily from the finance line of business) were surprised to hear this statistic.
During their session, “Data Science & Machine Learning Techniques to Automate Financial Controls,” Oracle vice president of product strategy Sid Sinha and vice president of development Reza B’far educated the audience on the basics of data science and its subset, machine learning—examining the processes and challenges involved in using data science to reduce risk. Before companies can use data science to make decisions, they must invest significant effort, time, and expense in data collection and preparation. Oracle Risk Management Cloud has automated these laborious tasks so that organizations can start using advanced analytics right away.
Finance is a data-intensive discipline where many tasks can be repetitive and predictable. Data science can automate many of those routine tasks—especially those involving audit procedures, security, and fraud. However, most finance users rely on desktop tools to access, manipulate and analyze data.
Sinha drew attention to the fact that 86% of companies still use spreadsheets to manage financial and operational risks such as fraud, errors, business policy violations, and compliance. The prevalence of manual, labor-intensive processes exposes organizations to risks (and potential losses) that human beings can easily miss—such as suspicious transactions, inappropriate access, and non-compliance—leading to damaged reputations, data losses, penalties, and cash leakage.
Risk management is an ideal fit for data science’s cognitive computing capabilities. There are countless use cases for risk management—including identifying duplicate invoices, unusual credit terms, high-risk user access, and ghost employees, to name a few. Sinha and B’far focused on one use case that helps identify invoices that are created by a user who also created the supplier site—clearly a red flag. Using data science models that can proactively monitor hundreds of transactions and activities, finance users can be alerted to these highly suspicious transactions in a user-friendly dashboard.
B’far also spoke about specific data science techniques such as “clustering”—a method in which data records are distributed into clusters with the nearest mean—with the intention of finding patterns that you didn’t know existed. With this method, you could identify a high number of expensed meals by a business unit. Another interesting pattern recognition technique is Benford’s Law, which can help identify anomalies in payments. Benford’s Law states that the frequency distribution of leading digits is predictable; for example, approximately 30 percent of values begin with the digit 1. This helps accurately identify fake expense reports, invoices, and so on, as fraudsters are unable to replicate the natural distribution of leading digits in numbers.
The fantastic news is that Oracle Risk Management Cloud already incorporates the data science and advanced analytic techniques described above, as well as many others. Finance teams can be alerted to suspicious transactions and activities as sophisticated techniques continuously monitor for them, all from within their ERP cloud.
Oracle Risk Management Cloud allows CFOs and the office of finance to take control of finance and operational risks, requiring fewer resources to manage these risks and putting more resources to work where it counts. Ultimately, the time and money spent on managing and reacting to risks will decrease, and the focus can shift to autonomously protecting the organization against non-compliance and losses—as well as to strategic risks.