Conclusions

There are critical issues related to the use of conventional Business Intelligence technology for decision making. Data support is an important contribution of BI, but data presented without a proper context can require more effort from decision makers and does not necessarily enable decision support. This represents a challenge that will become even more important in the future given that organizations are nowadays gathering terabytes of information.

In this research, we have provided several contributions toward a goal-oriented business intelligence-supported decision methodology, where we integrate in a novel way goal models, decision models, situations, and processes together with analytic capabilities. By integrating the decision model into the BI system, we attempt to integrate model-based decision support into a BI system thus improving the decision-making process. To do so, we extended a standard goal-oriented language, GRL, to better display relationships between Key Performance Indicators and objectives and processes, to enable formula-based evaluations of goal models involving KPI aggregation, and to integrate situations through the notion of acceptable ranges for KPIs. These extensions enable the combination of quantifiable KPIs with strategic-level softgoals in the same model, which in turn allows analysts to assess and monitor the impact of KPIs based on existing values and to explore what-if scenarios through GRL strategies.

We also proposed an enhancement to the integration architecture of the existing open source URN tool, jUCMNav. This new architecture addresses two major issues of the existing integration approach, which were considered main impediments for adaptation of the integration between jUCMNav and BI tools: dependency on an external web service and application server, as well as BI reports created, hosted, and manually maintained on the BI server.

From an implementation perspective, we also introduced a methodology lifecycle with iterative steps that support the construction of goal models (including KPIs and dimensions) even in situations where little information is available. Such models can be refined as more knowledge is gained about the organization and its context. Models can also be compared (as historical data) to validate newer models. This feature allows for assessment of the impact of past business decisions creating an ongoing system of record that permits continual adaptation. Our retail business example helped illustrate the methodology in a real context, and the results of our study suggest the feasibility of the approach, especially for middle-level managers. Other applications of parts of the methodology to information access in a large hospital and to adaptive enterprise architectures in a government's department also support this conclusion. We also believe that such a graphical, goal-oriented approach, which delivers data values used to make decisions in context, supports the comprehension of important cause-effect relationships in a way that could complement existing current BI technologies, which often lack an appropriate goal view.

The methodology is still evolving, and limitations and potential work items have been identified in our lessons learned. In particular, major future work items include:

Study of the usability of the methodology from several viewpoints, including the effort required by the iterative modelling before analysis can be done, the continuous evolution of the models, what to put in the URN model and what to leave in conventional BI systems, the usefulness of the analysis results in helping decision making, and appropriate tool support (including integration with BI systems other than IBM Cognos).

Further validation of the methodology by applying it to other real organizations and measuring its usefulness in organizations of different domains, sizes, and levels of maturity.

Improvements to the maintainability of models with formulas, and of evolving models.

Improvements to the performance of the current integration with IBM Cognos BI by exploiting the facilities provided by IBM Cognos Insight.

Still, as it stands today, this methodology brings new contributions and good value to the BI table, and we believe it has a promising future.