A recent report from the Association of Certified Fraud Examiners estimates that occupational fraud costs almost $200,000 per incident. Unmonitored supply chains, underlying inventory assets, a host of third parties, and frequent transactions make the manufacturing and retail industries an especially sweet spot for procurement and inventory fraud schemes.

Procurement fraud involves unethical behavior and deceit by an employee, vendor or partner working to obtain a financial or unfair advantage. Common vulnerabilities and fraud risks in the manufacturing sector include bid rigging, conflict of interest, warranty claims, intellectual property infringement, theft or misuse of inventory, product counterfeiting, supplier collusion, purpose orders and employee spend.

While fraud is impossible to eliminate entirely, there are common ways to reduce its probability and more quickly identify red flags:

• Evaluate and actively monitor internal controls. Existing controls, thresholds and procedures should be regularly reviewed and assessed for relevance, adequacy and effectiveness.
• Develop a comprehensive, well-communicated fraud-response plan. Providing regular training to teams on how to effectively spot, report and respond to fraud is crucial to limiting its impact.
• Conduct due diligence on your supplier. Background checks and integrity due diligence can ensure that manufacturers or suppliers are reputable, by highlighting their interests, associations, related parties and possible conflicts of interest.
• Perform regular quality checks. Conduct regular checks such as routine assessments for non-deliveries, repeat deliveries for the same order, and discrepancies between purchase orders and delivery.
• Put the power of your data to work. Big data not provide key insights, but allows organizations to identify unusual or suspicious behavior.

How Data Can Fight Procurement Fraud

Artificial intelligence and machine learning automate the detection and triage processes, allowing for faster resolution. With the interconnectivity of data growing like never before, simple anomaly-detection techniques that involve one variant frankly do not cut it anymore.

The most advanced companies are relying on sophisticated anomaly-detection techniques to pinpoint abnormalities in their data, instead of trying to manually find them buried in dashboards and reports. By using AI to automate fraud detection, a manufacturer can instantly pick up on such red flags as excessive shrinkage in inventory, an abnormal rise in invoice volumes, split purchase orders, multiple payments to vendors without any corresponding services rendered, unusually low or high bid prices, and a sudden and unexplainable rise in customer complaints.

Companies need algorithms that look across multiple data sources, metrics and segments to uncover trends and assess where anomalies lie. Organizations achieve the most significant performance improvements when humans and AI tools work together. Through such collaborative intelligence, they enhance one another’s strengths. 

Where Do You Start?

Following are the five most common ways to use AI and ML to protect your organization against procurement fraud.

• Anomaly detection is an unsupervised learning technique that’s especially value when utilizing large data sets.
• Logistic regression is a supervised learning technique that’s used when the decision is categorical. It can be used to flag suspicious transactions.
• Decision tree algorithms in fraud detection are used where there’s a need for the classification of unusual activities in a transaction from an authorized user. These algorithms consist of constraints that are trained on the dataset for classifying instances of fraud.
• Random forest uses a combination of decision trees, each designed to check for different conditions. Trained on random datasets, each tree gives the probability of the transaction being “fraud” or “non-fraud.” Then the model predicts the result accordingly.
• Neural networks is a concept inspired by the working of the human brain. It uses cognitive computing to build machines able to use self-learning algorithms involving the use of data mining, pattern recognition and natural language processing.

Detecting Unseen Fraud

One global electronics manufacturer has been able to proactively put its data to work to identify and investigate fraud in its procurement process.

The organization was experiencing an increasing number of procurement fraud cases, with very specific concerns in high-risk regions. Its corporate audit committee identified the need for improved data and analytics to support fraud detection and investigation at scale. Unfortunately, existing compliance tools were limited to ad hoc reports and data requests, meaning they were highly manual, time intensive and took months to refresh.

To help improve its fraud-detection and investigation capabilities, the manufacturer partnered with a provider of data, analytics and technology services with experience in the manufacturing and retail sectors. The provider began by conducting a four-week assessment to review available data, evaluate existing tools and prioritize compliance issues and opportunities. From there, the provider was able to deliver a split-purchase order anomaly-detection prototype in five weeks, with an automated, on-demand dashboard implemented within the next two weeks. The solution was then expanded to include supplier-favoritism detection through risk scoring, covering such elements as collusion, conflict of interest and kickbacks.

With sophisticated anomaly-detection technology in place, the manufacturer has realized 42% incremental procurement cost recovery through detection of “unseen” fraud. The company has also seen fully qualified fraud cases increase by 22% and achieved a 20% improvement in investigation efficiency. Additionally, project payback was achieved in one month with identification of just one split-PO fraud case.

With the growth in both data and economic uncertainty, companies are devoting more attention to predicting procurement fraud instead of just reacting to it. Manufacturers around the world can learn to harness the power of AI and ML to take control and minimize profit-cutting procurement fraud.

Bobby Falconer is associate partner consulting services with Kaizen Analytix.