Business-relevant metrics allow business experts to evaluate the effectiveness of the enterprise system. The metrics provide helpful insights into specific parts of the business domain. The application is responsible for gathering business-relevant metrics as part of it's use cases.

The car manufacture package, for example, performs business logic that can emit certain metrics, such as the number of cars created per hour.

From a business perspective, the relevant metrics usually originate from domain events. It's advisable to define and emit domain events, such as CarCreated, as part of the use case, as soon as a car has been successfully manufactured. These events are collected and being used to derive further information in the form of specific business metrics.

The CarCreated event is fired in the boundary as a CDI event and can be observed in a separate statistics collector. The following code snippet shows a domain event fired as part of a use case:

@Stateless
public class CarManufacturer {

    @Inject
    CarFactory carFactory;

    @Inject
    Event<CarCreated> carCreated;

    @PersistenceContext
    EntityManager entityManager;

    public Car manufactureCar(Specification spec) {
        Car car = carFactory.createCar(spec);
        entityManager.merge(car);
        carCreated.fire(new CarCreated(spec));
        return car;
    }

}

The boundary fires the CDI event that notifies about a successful car creation. The corresponding handling is decoupled from the business process and no further logic is involved in this place. The event will be observed in a separate application scoped bean. Synchronous CDI events can define to be handled during specific transaction phases. The following transactional observer therefore ensures that only successful database transactions are measured:

import javax.enterprise.event.TransactionPhase;

@ApplicationScoped
public class ManufacturingStatistics {

    public void carCreated(@Observes(during =
            TransactionPhase.AFTER_SUCCESS) Specification spec) {
        // gather statistics about car creation with 
        // provided specification
        // e.g. increase counters
    }
}

The event information is collected and processed further in order to provide the business metrics. Depending on the situation, more business-relevant data could be required.

Modeling the relevant information as domain events matches the business definition and decouples the use case from the statistics calculation.

Besides defining domain events, the information can also be collected via cross-cutting components, such as interceptors, depending on the situation and requirements. In the simplest case, the metrics are instrumented and collected in primitives. Application developers have to consider bean scopes in order not to throw away collected data with incorrect scopes.