Evidence-based software portfolio management

Abstract

While many studies can be found on estimating the cost (or effort) of software deliveries (Jørgensen & Shepperd, 2007), only a handful exists about the price thereof. For pricing purposes in a commercial context where software delivery is partly performed by suppliers, most companies rely heavily on expert judgment (Boehm, 1984) (Jørgensen, 2004). This is not always a successful approach (Moløkken & Jørgensen, 2003), and software development is often characterized by high cost and schedule overruns (Verhoef, 2002) (Glass, 2002), despite a huge number of tools developed over time to support project estimation (Boehm, 1984

Effort and Cost in Software Engineering

“…In both cases we look for the best fit; meaning that for each dataset different prediction models can apply. Furthermore, we evaluate the performance of both repositories from the perspective of a full software portfolio, by analyzing specific samples of software project data against the content of the EBSPM-repository [13]. Finally, we look for causes that might explain our findings, by studying existing literature on the subject of effort and cost of software projects.…”

“…statistics of the EBSPM-repository. The repository and the accompanying tool are described in more detail in a separate tool description [13].…”

“…For linear regression, we also eliminate influential observations based on Cook's distance. Besides that, we compare both repositories by mapping the project data of the ISBSG subset on the Cost Duration Matrix in the EBSPM-tool [13], analyzing the following performance indicators: 1. Cost per Function Point (FP): the weighted average cost (in Euros) per FP, where size (FP) is the weighting factor, instead of number of projects.…”

“…To visualize performance in terms of size, cost, duration, and defects, we mapped the subset of 172 ISBSG projects on the content of the EBSPM-repository, by using the EBSPM-tool [13] (see Figure 4). To analyze the overall performance of both repositories, we calculated three Performance Indicators (see Table 6).…”

“…In order to emphasize the importance of cost, especially top-level 2 decision makers highly value cost transparency, as we found in a study on pricing of software projects [12]. To explore the differences and commonalities between cost and effort, we use two large software repositories; the EBSPM repository [13] [14] [15] and the ISBSG repository [10], with the objective of gaining insights into the usefulness of historic effort and cost data for benchmarking and cost estimation purposes. This is one of the first studies in which EBSPM is used in comparison with other benchmark datasets.…”