Key Performance Indicators (KPIs) allow the model to implement a logic layer, so that a value can be monitored according to a rule. The outcome of the rule is a simple set of results that indicates how the value is comparing to the expectations, and usually equates to one of the three simple conditions: bad, moderate, or good.
By summarizing logic into the model in this way, the KPI improves model usability as the user is not required to interpret the value in the context of a business situation (or some layer of logic that is applied to the values); this is all managed by the KPI. Furthermore, since the KPI is represented by a state (for example, bad, moderate, or good), the KPI can cover boundary ranges of data. This effectively discretizes data into a conditional state based on the business logic.