Next Generation Software Estimating Framework: 25 Years and Thousands of Projects Later

Abstract

It's about time we in the software development community revisit the assumptions, relationships, and flexibility contained in our currently available software estimating models. Most of the current models still implement fundamental relationships that are based on at least 25-years-old data and assumptions. In the meantime, data from many thousands of projects have since been collected and offer an opportunity to revisit old assumptions and relationships. This paper documents the basis, assumptions, and derivations behind a set of general software effort, duration, and defects estimating relationships that are based on the notion that software development is the cumulative effect of people laboring to do work (effort) over some duration (period of elapsed calendar time) that produces a desired software product (size or content) and unwanted byproducts (defects). This set of relationships is derived from several evidently good correlations, the primary three being 1) effort generally trends upward with increasing size, 2) duration generally trends upward with increasing effort, and 3) effort generally trends upward with increasing defects. This derivation ultimately yields three limited tradeoff relationships: one between effort and duration, one between cost and duration, and one between defects and duration.