An Automated Quality Framework For EA Model Assessment

An Automated Quality Framework For EA Model Assessment

In Enterprise Architecture (EA) Management it has long been understood that enterprise architecture models represent a vital data stock which is the foundation for crucial planning and decision making tasks in complex businesses. But there is still a lack of metric-based quantitative approaches to assess the quality of such models that provide a holistic framework to accompany architects and practitioners in the field along their way. Furthermore there is a principle lack of research that would assess the potential of quality metrics stemming from other domains to be adapted to EA models.

The literature describes several frameworks that either rely on non-automatable qualitative analyses, or on quantitative approaches which require the model to undergo extensive prior meta data enrichment. Furthermore several quantitative techniques can be found, which only focus on limited areas of quality, remain purely theoretical in nature, or are outdated.

In this thesis we introduce a new metric-based framework for the automatable evaluation of EA model quality which offers a holistic solution. Additionally we investigate several wide-spread metric suites from the fields of software engineering, and try and adapt them to EA. We introduce a metrics elicitation procedure that is based on the theoretical foundation of a bi-directional quality and process model which links measurement needs to stakeholder-driven goals. We show the practical application of our process model by establishing a complete metric suite. This suite is incorporated into a software architecture which is implemented as a command-line tool capable of automated execution.

We present an evaluation procedure and scenario, and put our metric suite to the test using a complex model close to a real-world business context. We show our metrics elicitation technique is fit to yield artifacts that translate into a holistic metric suite, which successfully assesses the quality of the evaluation target scenario.

We expect this holistic approach and its scope to provide EA practitioners and architects with several potential benefits. Our quality and process model serve as a blueprint on how to establish own metrics. Our metrics together with the accompanying software constitute both an evaluated proof of concept for the elicitation procedure, and a suite applicable out of the box. Finally the modular software architecture might serve as the basis for own tool development and extensions.

Project information



Thesis for degree:



Michael Beyer