Procurement fraud is, to put it simply, huge. An estimate by PwC a few years ago suggested that it was one of the most common economic crimes around the world. That estimate has not changed. What’s more, few companies have taken steps to address the problem.
“Procurement fraud is notoriously difficult to detect and investigate, because it takes so many forms and can be driven by any number of actors, internal or external, at any point in the procurement life cycle,” says Mickey North Rizza, Program Vice President of the Enterprise Applications and Digital Commerce Research Practice at IDC. “Manual detection is futile. Only the right combination of advanced analytic techniques can arm large organisations to battle the fraudsters.”
In the public sector, including local government and health care, awareness of procurement fraud is rising. Organisations are slowly starting to put systems in place for its prevention and detection, including the use of continuous monitoring.
Understanding continuous monitoring solutions
Continuous monitoring solutions pull millions of records from back-office systems, including purchase orders, purchase requests, invoices, payments and other accounts payable data, and supplier data. After extracting the data, they apply AI to cleanse it. This increases detection quality without asking legacy systems for impossible data cleanups.
The system can enhance the data by drawing on external sources, such as company registers, information about politically exposed people and transparency indexes. Then it processes the data using alert generation processes based on mathematical algorithms and models, such as clustering and link analysis.
Depending on the industry and the company needs, organizations can use analytics to check for bid rigging, duplicate invoices, travel expenses, returns, fidelity cards, subscriptions and more.
Examples of continuous monitoring
Companies have to follow rules when finding new vendors and exercise due diligence. These rules are often based on answers to questionnaires and can be difficult to manage because of the number of vendors or the difficulty of verifying information. Automated solutions can sift through supplier master files looking for high-risk managers or links with risky individuals that warrant additional due diligence.
A recent SAS survey suggested that travel and expenses are the second-most important area of procurement fraud. In total, 35% of 850 companies interviewed had faced this in 2018. To reduce bribery and corruption, continuous monitoring can be used to check gifts, hotels, venues, events and hosting expenses and detect anomalous behaviour.
Points of sale in retail and subscriptions in telco are also rife with fake returns. This outright fraud leads to massive losses but can be controlled by continuous monitoring to reduce exposure.
Organisations also struggle with data quality issues. These may be the result of errors, but can also be indicative of fraud. Rapid changes of address for suppliers, for example, could show a recurring error, a startup that often changes offices or a ghost vendor avoiding controls.
Living the AI revolution
According to the chief analytics adviser at Eskom, “Eskom has a fiduciary responsibility to ensure that all of its payments are legitimate. Using advanced analytics to identify suspect transactions and relationships in our massive data, in near-real time, will enable us to be better stewards of public funds. The savings will also allow us to electrify more homes, build more generating capacity to meet growing demand, and avoid unplanned outages by better maintaining our aging fleet.”
Reducing procurement fraud is essential to avoid reputational damage and hefty fines levied by regulators. Data-driven solutions automatically attribute risk scores to suppliers. They identify relationships between politically exposed people and third parties and highlight possible collusion to give compliance departments near-real-time insights into new suppliers and suspicious transactions. They are, therefore, a world away from the traditional approach of sending teams of investigators to sift through spreadsheets and documents to identify illegal payments and behaviour.
The average auditor now reviews over 1 billion dollars of spending each year. Automated data-driven anomaly identification enables them to take a proactive and systematic approach that was almost impossible before the use of fraud detection analytics. The combination of cloud computing power, accessibility and volumes of data, and powerful algorithms are turning compliance and audit into an entirely new job. These functions are living the AI revolution every day.