ACCT 660 Week 1 Discussion Board Forum 1 Thread Example

Discussion Board Forum 1 Thread: Data Analytics as a Fraud Detection Tool

ACCT 660 Information Technology and Fraud, Discussion Board Forum 1 Thread

Thread by [Student Name]

The thread answers a composite prompt on forensic analytics; the studies, standard and book cited are real.

Forensic analytics changes the search for fraud by letting an examiner test every transaction instead of a sample. When a company records millions of entries a year, a sample of a few dozen can easily miss the handful that were falsified. Testing the whole population closes that gap, but it brings a new risk: a computer will flag thousands of anomalies, and most of them will be innocent. The value of analytics therefore depends on two disciplines that Nigrini's first chapters emphasize, understanding the data before testing it and treating every result as a question rather than an answer.

The first discipline is the data profile. Nigrini (2011) recommends summarizing a data set by ranges before running any fraud test, counting transactions and totaling their amounts in strata such as under $10, $10 to $100, $100 to $1,000 and so on. A profile shows how the data are shaped, reveals oddities such as negative amounts or a surprising number of very small items, and tells the examiner which tests will make sense. A payables file in which a few hundred large payments account for most of the dollars calls for different tests than a file of thousands of small purchasing card charges. Profiling also checks that the data are complete, since a file that does not agree to the general ledger total cannot support any conclusion.

The second discipline concerns how anomalies are found and followed. Digit analysis is the best-known technique. In many naturally occurring financial data sets, the first digits of amounts follow a predictable pattern in which 1 appears as the first digit far more often than 9. Nigrini and Mittermaier (1997) proposed using the expected pattern as an analytical procedure, showing that departures from it can point auditors toward accounts that deserve closer examination. A departure from the expected pattern tells the examiner where to look, not what happened there. Departures can come from legitimate business features, such as a recurring payment of a fixed amount, as easily as from invented numbers.

Research on real accounting records shows both the promise and the limits. Debreceny and Gray (2010) applied digit tests and other data mining techniques to the journal entries of a set of real organizations. Many of the entity files departed from the expected digit patterns, and the authors argued that such results are most useful as starting points for audit inquiry rather than as evidence in themselves. Their study also illustrates a practical difficulty: journal entry data differ from one system to another, so the examiner must understand how each organization records its entries before interpreting what a test shows.

If the tools work, why are they not used everywhere? Survey evidence suggests that the problem is adoption rather than effectiveness. Bierstaker et al. (2006) asked accountants, internal auditors and fraud examiners to rate prevention and detection methods. Respondents rated data mining and digital analysis software as highly effective, but these methods were among the least used, while basic controls such as passwords and firewalls were used most. The gap reflects cost, training and the time needed to clean data, all barriers that have fallen as tools have improved.

For auditors of public companies, journal entry testing is not optional. Under the auditing standard on fraud, auditors must plan tests of whether the entries posted to the ledger, and other adjustments, were proper, paying special attention to entries made at odd times, by unexpected individuals or to accounts that seldom receive entries (Public Company Accounting Oversight Board [PCAOB], 2002). The standard reflects a lesson learned from past frauds: when management overrides controls, it usually does so through journal entries, often recorded at the end of a period.

Putting these ideas together suggests a practical first analysis of a company's general ledger. First, reconcile the extracted entries to the trial balance to confirm completeness. Second, prepare a data profile by amount and by month to understand the population. Third, run a first-digit test and a test for duplicate amounts on the entries, account by account rather than on the ledger as a whole. Fourth, isolate entries posted on weekends, after the close or by users who rarely post. Finally, select the entries that appear on more than one list and examine their support. That workflow uses the computer for what it does best, scanning everything, and keeps judgment where it belongs, with the examiner who must decide what each exception means.

I would be interested in classmates' experience with this. In your organizations, does anyone run tests of the full general ledger, or is the review of entries still limited to a sample and to entries above a dollar threshold?

References

Bierstaker, J. L., Brody, R. G., & Pacini, C. (2006). Accountants' perceptions regarding fraud detection and prevention methods. Managerial Auditing Journal, 21(5), 520-535. https://doi.org/10.1108/02686900610667283

Debreceny, R. S., & Gray, G. L. (2010). Data mining journal entries for fraud detection: An exploratory study. International Journal of Accounting Information Systems, 11(3), 157-181. https://doi.org/10.1016/j.accinf.2010.08.001

Nigrini, M. J. (2011). Forensic analytics: Methods and techniques for forensic accounting investigations. Wiley.

Nigrini, M. J., & Mittermaier, L. J. (1997). The use of Benford's law as an aid in analytical procedures. Auditing: A Journal of Practice & Theory, 16(2), 52-67.

Public Company Accounting Oversight Board. (2002). AS 2401: Consideration of fraud in a financial statement audit.