R&D is a primary focus of our work at Audit Data Analytics. Together with consulting, R&D provides the guiding principles for our software development. Unlike other software companies, we don’t just build software, put it out there, and hope it works. Our consulting practice provides insights into customers’ needs and pain points, helping us to focus our efforts. We then apply sound R&D based on statistical science to ensure our approaches are effective and will deliver the best results for our customers. As scientists, we challenge old notions and seek evidence to support assumptions. As a rare research institution with a focus on audit analysis techniques, Audit Data Analytics has forged new paths, pioneering cutting edge approaches and refinements in the areas of monetary misstatement sampling and AI machine learning.
Statistical Sampling
Sampling research is an ongoing effort with an array of simulations that investigate the impact of various scenarios auditors deal with in practice. We are always in pursuit of more accurate and robust sampling techniques. This research typically results in papers published in peer review journals like the Journal of the American Statistical Association.
Kurt Johnson’s latest effort is Empirical Evidence Demonstrating the Superior Performance of Monetary Unit Sampling Over Stratified Sampling Methods in 100% Error Sampling Environments. It tests traditional Classic Variable Sampling (CVS) versus Monetary Unit Sampling (MUS). The endeavor took him 2 years. During this time he built a simulator application to run Monte Carlo simulations to test his theories. The results were nothing short of extraordinary. The paper provides theoretical and simulation (empirical) evidence that MUS is unbiased and produces better estimates that CVS. This research has sparked a whole new approach to MUS along with a bootstrap estimator innovation that provides the best estimates for monetary misstatement sampling.
Big improvements are coming to the sampling modules in future versions of ADA as a result!
AI for Data Analytics
Machine learning will soon become a big part of ADA. R&D insights in the use of AI classification, topic modeling, and time series are driving new software developments with AI for audit data in mind. Monte Carlo simulations are employed to prove the approaches’ effectiveness in audit scenarios with transparent usefulness to auditors. This research will yield some awesome new features in ADA in the year 2021.
For an example, check out our blog article on Applied Machine Learning in Audit using k-Modes and k-Prototypes.