too technical for the typical auditor. Research on techniques is needed in two areas.
First, the field needs a better understanding of what simple techniques for outlier
detection, trending, link analysis, and fulltext
analysis are useful for fraud detection. To
date, only Benford’s Law has seen wide use, and while it is interesting, its effectiveness
for actually finding fraud is not well proven—especially on discovering frauds that have
not otherwise been discovered with other techniques. Second, a large body of advanced
statistical and computerscience
literature is available. Application of these techniques to
the fraud area has been done in a few papers, but significant work still remains. The
literature provides little clarity on which technologies can be used to discover fraud.
Both empirical and case study research is needed to determine how these techniques can
be successfully implemented.
Beyond the techniques themselves, auditors have little training in computer
programming, query languages, and statistics. In addition, they do not have sufficient
time to perform these algorithms in typical audits. Detectlets are one method of encasing
the knowledge, routine, and algorithm into a wizardlike
interface (Albrecht, 2008).
They may be able to solve both the training and time problem. Development of detectlets
or another solution is needed before the audit field can realize success with these
advanced techniques. This is a primary area that information systems and technicallysavvy
accounting researchers can add valuable knowledge to the audit field.
It is often said that the preparation of data is more difficult than the analysis of it.
Research is needed into the best tools, techniques, and methodologies that auditors can
use to prepare data for analysis from source business databases. Cutoff points, missing
values, abnormal trends, and other difficulties arise during preparation. One exciting