Machine learning: The new way to combat expenses fraud?

Aug 11, 2016Uncategorized0 comments

Receipts-PixabayExpense reimbursement fraud makes up about 15 percent of business fraud with a median annual loss of $26,000, according to a study by the Association of Certified Fraud Examiners. And, according to this study, it takes about 24 months before expense reporting fraud is detected.

Consider the following expenses claims: registration fees for a cancelled seminar, two separate claims for mileage when the employees travelled together, and a sandwich-and-coffee dinner claimed as the full per diem.

While it’s easy to believe that a few dishonest claims won’t hurt, for individual victims, expenses fraud can be costly. Imagine if 20 per cent of your staff added 10 per cent to each mileage claim; the cumulative loss for the company would quickly become significant.

Existing fraud detection systems flag dubious-looking expenses according to a set of rules, such as challenging claims in excess of a fixed cash amount or those that are 5 percent higher than claims submitted by peers in similar positions. While these methods can help to stop a number of fraudulent expense claims, they are only as good as the rules you code.

But there’s one thing that could take fraud detection to a new level: It’s called machine learning — and it may soon be coming to an enterprise system near you.

A new kind of fraud detection

Machine learning is the process of enabling computers to learn automatically, without human intervention. As the software is exposed to new data, it learns to spot patterns of behavior and, crucially, anomalies in those patterns. It’s not a new science but one that’s gaining momentum in the field of fraud prevention.

Take, for example, a purchase made at an online store. If a percentage of the retailer’s past transactions were fraudulent, then it is likely that a percentage of future transactions will also be fraudulent. Machine learning technology is able to ingest gargantuan quantities of historical transaction data and synthesize them to figure out whether a new transaction displays the characteristics of an illegitimate transaction. This allows it to spot fraud patterns that would be missed by human experts.

Machine learning is not a silver bullet for fraud prevention. But it can give some high-value analysis that businesses can use to prevent scams before they happen.

Will machine learning work for expenses fraud?

Machine learning technology is very new in the field of enterprise software. Financial institutions use it to detect credit card and e-payment fraud but, at least for now, machine learning has not been used to detect expense abnormalities.

That doesn’t mean it couldn’t be used to block expenses fraud.

The secret sauce of machine learning is data: The system needs enough samples to be able to pinpoint and predict atypical patterns of behavior. In today’s interconnected world where employees constantly interact with online services, the software could mine so-called big data from such diverse sources as airline check-ins, social media and the company’s own enterprise application to spot anomalies in expenses reports.

Consider the employee who submits an expenses claim for a mileage. Modern enterprise applications already contain the functionality to scan claims and flag valuable information such as the start point, destination, time and date.

These same systems contain a wealth of data regarding employees who travel within a company. Currently, these systems are used to suggest travel plans based on project assignments, customer locations and previous travel to the same location by other team members. The technology fills out time sheets by allowing users to check in whenever they reach customer locations or offices where they need to go based on work assignments.

Using machine learning, advanced algorithms can employ the same data reactively to flag inconsistencies when final expenses claims are submitted. Mismatches between where an employee should have travelled based on work assignments and the mileage claimed would automatically be captured and referred for investigation. This capability is built into the ERP infrastructure, forming an integrated system that allows the business owner to block fraudulent expenses submissions before they happen.

More than just the bottom line

Machine learning is not just about highlighting unusual behavior to finance managers. Employees lose the expenses game too — to the tune of around $15 billion a year.

A recent multinational study by Ruigrok NetPanel found that one in five employees are regularly left short of money because they do not claim the expenses they have incurred. The findings are based on responses from almost 2,000 employees across the United States, Canada, Spain, France, Germany, Netherlands, Belgium, Sweden and the United Kingdom.

A number of reasons were cited for the low uptake, including lost receipts, forgetting to submit the expense claim or simply because the process was too frustrating and time consuming. Averaged across countries, 28 per cent of employees felt that poor expense claims processes had a negative influence on their feelings toward the company, leading to lower performance and disengagement.

Could machine learning change that? The reality is, we don’t know. No commercially available expense system offers machine learning — yet. But it is fair to suggest that introducing a more transparent, intuitive and intelligent way of processing expenses could put an end to employee frustration, as well as reducing the losses associated with fraud.

Machine learning is a useful way of providing an accurate representation of expenses. The psychological impact of the software is that employees will think twice before trying to manipulate such an intelligent system — and employers will have to play fairly, too.


Claus Jepsen, Chief Architect & Head of Unit4’s Innovation Labs for

Photo Credit: Pixabay