Architecture-Centric Project Estimation

In recent years studies have been conducted which suggest thattaking an architecture first approach to managing large softwareprojects can reduce a significant amount of the uncertainty presentin project estimates. As the project progresses, more concreteinformation is known about the planned system and less risk ispresent. However, the rate at which risk is alleviated variesacross the life-cycle. Research suggests that there exists asignificant drop off in risk when the architecture is developed.Software risk assessment techniques have been developed whichattempt to quantify the amount of risk that varying uncertaintiesconvey to a software project. These techniques can be applied toarchitecture specific issues to show that in many cases, conductingan architecture centric approach to development will remove morerisk than the cost of developing the architecture. By committing todeveloping the architecture prior to the formal estimation process,specific risks can be more tightly bounded, or even removed from theproject.The premise presented here is that through the process ofarchitecture-centric management, it is possible to removesubstantial risk from the project. This decrease in risk exceedsthat at other phases of the life-cycle, especially in comparison ofthe effort involved. Notably, at architecture, a sufficient amountknowledge is gained by which effort estimations may be tightlybounded, yet the project is early enough in the life-cycle forproper planning and scheduling. Thus, risk is mitigated through theincrease in knowledge and the ability to maintain options at anearly point. Further, architecture development and evaluation hasbeen shown to incorporate quality factors normally insufficientlyconsidered in the system design.The approach taken here is to consider specific knowledge gainedthrough the architecting process and how this is reflected inparametric effort estimation models. This added knowledge isdirectly reflected in risk reduction. Drawing on experience ofarchitecture researchers as well as project managers employing thisapproach, this thesis considers what benefits to the softwaredevelopment process are gained by taking this approach. Noting astrong reluctance of owners to incorporate solid softwareengineering practices, the thesis concludes with an outline for anexperiment which goes about proving the reduction in risk atarchitecture exceeds the cost of that development.