Estimation : Universal Models or Multiple Models ?

In the field, there is a very large diversity of development processes in use, and various mixes of costs drivers, each with a different impact depending on the context. The classical approach to building estimation models in software engineering is to build a single estimation model and include within it as many cost factors (i.e. independent variables) as possible. In this paper, we do not postulate that there exists a single estimation model that is ideal in all circumstances, but rather we report on exploratory research conducted over the past few years looking at relevant concepts from the field of economics and from discussions with organizations attempting to understand the data that they have collected on their projects. The purpose of exploratory research is not to demonstrate a hypothesis, but to identify new potentially relevant concepts to develop hypotheses to be tested later on with empirical or experimental data.