Planning Audits of a Network of Retail Branches
Background
The Head of Retail Audit at a Canadian Bank was looking for a flexible solution to plan their risk-based retail branch audit program. She required to optimize the audit plan in such a way that remote branches could be audited within a time window and that the riskiest branches would be prioritized.
The Challenge
Risk in a branch network includes formal risk measure, as well as non-traditional risk indicators. This means that the risk of a particular branch can be observed through traditional metrics such as the credit rating of the loans that it issues, the variability of its income (fees and commissions), and missing cash amounts, among others. However, it should also include information about customer complaints, staff turnover, and socioeconomic indicators of the geographic region that it serves. The variety of data types that measure these risk indicators indicated that the analysis would not only be based on numerical variables, but rather it would involve geolocation, categorical and text-based data.
Our Approach
Working directly with audit team, we realized that a key component of the solution was reducing the complexity of the data set. We used feature engineering and dimensionality reduction techniques to produce three indicators that would measure risk along the financial (revenue, risk ratings, etc.), operational (missing cash, client complaints, etc.), and human (turnover, HR complaints, etc.) dimensions, respectively.
The results were displayed on charts and maps, using a business intelligence tool. As part of subsequent work, we productionalized our methodology in a self-service, web-based tool.
The Benefit
Our indicators had the dual benefit of being based on dimensionality reduction techniques that maximize the amount of information that they capture and, at the same time, they were easily interpretable by the auditors. In this sense, the audit team was able to determine which branches should be audited as a priority and, within each branch, which dimension (financial, operational or human) should be more closely examined. The map that contained the information allowed optimization by geographic location, and to observe commonalities among branches that belonged to the same region.
Contact us
