Effort Estimation

by Roberto Meli

How to effectively integrate parametric models, expert judgment and Work Breakdown Structure in a flexible framework

There are two fundamental approaches to support software project’s effort and cost estimation: parametric cost models and analogy-based processes. The former are based on empirical data collected on similar projects (i.e. extracted from the ISBSG data base) fed to a human or artificial intelligence modeler (such as a neural network or other machine learning techniques), the latter are based on methods to exploit human expert judgement. Very often they are presented as alternative in the estimation endeavour. We think that they should be integrated instead.

The Work Breakdown Structure (WBS) is a planning tool, nominally known since a long time ago but also often very poorly used to support such crucial needs. Each specific project has a set of activities to be done to accomplish the final result / deliverable.

Those activities may be represented as a tree structure (the WBS) and may be related in various way to the deliverables (and their sizes) to be produced by the projects. Those activities may be strongly proportional, loosely proportional or independent by the size variables that have been chosen in any specific parametric cost model.

It is not a good idea to consider the effort/cost of activities that are independent from the chosen variables (i.e. Function Points) included in the productivity or unitary cost used for those variables. This means that we have to clearly identify in the WBS which are the activities covered by the parametric models and which should be estimated as a result of the expert judgement, taken adequate practices and constraints.

The presentation will focus on analyzing these aspects by presenting good and bad practices in the use of these support techniques.

Watch the YouTube video of the webinar

Download a PDF of the webinar