Sizing methods relate to quantifiable variables

Sizing methods determine the scope of a project (hard-facts) by means of different procedures and serve the transfer of specifications and expression characteristics in method-specific input variables and their metrics. There are about two dozen scientifically published or de-facto industry standard sizing methods known. Basically their will be different categories distinguished such as analogy-based, parametric or functional.

 

The best known analogy based method is the expert estimation. The acknowledged best way to implement them in a systematically way is the broadband Delphi method. The most developed sizing methods consider furthermore different boundaries in software projects, like new development via reuse of modules/components through to integration of third-party products the so-called commercial-off-the-shelf (COTS). Moreover a few support even hybrids like migration, modernization and package implementation.

 

Sizing methods only work in their specific basic conditions

Every sizing method has its own basic condition, rules and input size which are only valid in context. Sizing methods are not universally applicable, meaning not every sizing method is equally suitable for any project. Due to the permanent increasing requirements the challenge in practice is to find the optimal method, use and calibrate it to the specific environment. Many sizing methods even represent their application and purpose in the name. For example UML Use-Case Points will be used if UML is used in the project and for Web-Frontend are Internet-Points the best choice. This does not apply for Function Points which are mostly used for Backend Systems and for Feature-Points which work best in embedded and real time systems.

For new development there is usually no source code available. Furthermore you need sizing methods for determination and normalization of equivalent SLOCs as basis for further calculations. Because of that the permanent advancement of sizing methods is a mandatory requirement for reliable estimations.