Proactive business intelligence: Discovering key performance indicators and associated business rules from historical data using data mining techniques.

Abstract

Key performance indicators have been around for a long time; organizations tend to provide their employees with metrics on strategic levels, such as marketshare or profitability. Next to the key performance indicators that serve the strategic vision of an organization, tactical and operational levels are also influenced by metrics. By using metrics and performance indicators, organizations aim to influence the awareness of employees on targets on a personal and organizational level. Performance increases are bound to occur, when employees are made conscious of their personal results by using metrics. However, an abundance of current definitions of metrics and performance indicators overwhelm employees with values which are hard to grasp. By defining key performance indicators, relating to goals and strategic aspects of the organizations; meaningful indicators can be defined on an operational and tactical level, therefore, prevent cognitive and operational overload. This research focuses on the key aspects of defining key performance indicators by studying literature containing metrics, the metrics studied are often used in the field in which an organization operates. These metrics are compared to the ones, which are currently employed within the organization’s structure. As a result of comparing data and active metrics, a selection of key metrics is provided to the organization. The availability of these key metrics, enable organizations to focus on the business, instead of defining metrics. The key performance metrics are extracted from literature and available data is used in the Rule Extraction Matrix (REM) Method. This method is constructed using several aspects of current data mining methods. One of the applicable data mining methods is the CRISPDM model, which provides a solid base to determine the nominal value of a performance metric. The method provides ways to bond the performance metrics, using business rules, and thus leaving room for an interpretation of these rules using a decision support system. The REM method enables organizations to manage their performance by using bounded metrics and extract business rules, business rules are responsible for providing thresholds to the extracted key performance indicators. Validation of the REM method is performed by an extensive case study at a large Dutch Telephone and Internet services provider and expert interviews were performed. The REM method is an addition to the performance measurement field and further enhances an organization’s ability to define their performance. Moreover, the REM method is applicable to a variety of organizations given the following requisite: historical performance measurements must be available.