Beschreibung
Internal auditing faces multiple challenges caused by the growing amounts of data stemming from ongoing digital transformation. New techniques are therefore being evaluated for their application in auditing such as outlier detection which is able to uncover irregularities without requiring domain knowledge about a system and has already been applied in a number of auditing studies. Most identify outlier detection as only a first step, however, highlighting the key challenge of turning detected outliers into audit findings. Addressing this challenge, this work explores how outlier explanation and visualization can help auditors derive actual findings from potential findings. For this, an overview of existing outlier explanation approaches is created, requirements from internal auditing’s perspective are collected and based on these, new approaches are developed addressing two key gaps – support for mixed-type data and visualization. After a quantitative evaluation, one of the approaches is integrated into a prototype for a qualitative evaluation within the internal audit function of an international automotive manufacturer. Both quantitative and qualitative evaluations show that the developed approach can facilitate the application of outlier detection for internal auditing through outlier explanation and visualization and can, thus, help auditors to address the proliferation of data and to reduce risks by uncovering previously overlooked problems.
