suspicion

Forensic accountants have long used technological tools to uncover fraud schemes. But recent advances in “big data” have provided even better, more efficient techniques for identifying suspicious activities and dishonest employees. These are three common types of data analytics used by fraud experts:

1. Association analysis

This method can help identify suspicious relationships by quantifying the odds of a combination of data points occurring together. In other words, it calculates the likelihood that if one data point occurs, another will, too.

If data point combination occurs at an atypical rate, a red flag goes up. For example, association analysis might find that a certain worker or manager tends to be on duty when inventory theft occurs.

2. Outlier analysis

Outliers are data points outside the norm for a given data set. In many types of data analysis, outliers are simply disregarded, but these items come in handy for fraud detection. Experts know how to distinguish and respond to different types of outliers.

Contextual outliers are significant in certain contexts but not others. For example, a big jump in wages on a retailer’s financial statements might be notable in April but not in December, when seasonal workers usually come aboard.

Collective outliers are a collection of data points that aren’t outliers on their own but deviate significantly from the overall data set when considered as a whole. If, for instance, several public company executives sold off substantial blocks of stock in the business on the same day, it might indicate suspicious behavior.

3. Cluster analysis

Here, experts group similar data points into a set and then further subdivide them into smaller, more homogeneous clusters. Data points within a cluster are similar to each other and dissimilar to those in other clusters. The greater the similarities within a cluster and the differences between clusters, the easier it is for an expert to develop rules that apply to one cluster but not the others.

Cluster analysis has long been used for market segmentation of consumers. But it can also detect fraud, particularly when combined with outlier analysis. Outlier clusters — those that are farthest from the nearest cluster when clusters are mapped out on a chart — generally merit extra scrutiny for suspicious activity.

Fraud experts might, for example, use cluster analysis to evaluate group life insurance claims. They then would look for clusters of large beneficiary or interest payments, or long lags between submission and payment.

Old school methods

Of course, technology alone usually doesn’t make the case against an employee. Face-to-face interviews and other “old school” methods are crucial to identifying fraud perpetrators and learning where they’ve stashed the money they’ve stolen. If you suspect fraud in your organization, contact us to investigate.

Benford’s Law is a long-standing statistical precept that remains as relevant and widely accepted in fighting fraud as ever. By wielding it effectively, experts can cut down fraudsters who unknowingly reveal their wrongdoings in dubious digits.

Historical background

The rule is named for Frank Benford, a physicist who noted that, in sets of random data, multidigit numbers beginning with 1, 2 or 3 are more likely to occur than those starting with 4 through 9. Studies have determined that numbers beginning with 1 will occur about 30% of the time, and numbers beginning with 2 will appear about 18% of the time. Those beginning with 9 will occur less than 5% of the time.

Further, these probabilities have been described as both “scale invariant” and “base invariant,” meaning the numbers involved could be based on, for example, the prices of stocks in either dollars or yen. As long as the set includes at least four numbers, the first digit of a number is more likely to be 1 than any other single-digit number.

Striking implications

Benford’s Law carries striking implications for fraud detection. To avoid raising suspicion, fraud perpetrators often use figures they believe will replicate randomness. Typically, they choose a relatively equal distribution of numbers beginning with 1 through 9.

Fraud investigators can take advantage of such errors and test data in financial documents including:

Tax returns,
Inventory records,
Expense reports,
Accounts payable or receivable, and
General ledgers.

Although complicated software programs based on Benford’s Law exist to examine massive amounts of data, the principle is simple enough to apply using basic spreadsheet programs.

Not infallible

Benford’s Law, however, isn’t infallible. It may not work in cases that involve smaller sets of numbers that don’t follow the rules of randomness or numbers that have been rounded (resulting in different digits). Also, smaller numbers are more likely to occur simply because they’re smaller and the logical place to begin a count.

Assigned numbers, such as those on invoices, are also iffy. On a similar note, uniform distributions — such as lotteries where every number painted on a ball has an equal likelihood of selection — may not suit a Benford’s Law analysis. And prices involving the numbers 95 and 99 (often used because of marketing strategies) may call for a different approach.

Still relevant

Benford’s Law isn’t appropriate in every instance. And, as advanced metrics forge new inroads into fraud detection, it could fall out of favor. But Benford’s Law is expected to remain a foundational approach to fraud detection for many years to come.

Tag Archives: suspicion

3 ways fraud experts use data analytics

Wielding Benford’s Law to find fraud