On the Value of Project Productivity for Early Effort Estimation

TL;DR

This paper investigates the role of productivity in early effort estimation using Use Case Points, demonstrating that learning productivity from historical data improves estimation accuracy, though environmental factors are less reliable.

Contribution

It introduces empirical analysis of productivity's impact on effort estimation and proposes flexible productivity measurement methods based on available data.

Findings

Learning productivity from historical data improves effort estimation accuracy.

Predicting productivity from environmental factors is often unreliable.

Flexible productivity measurement enhances early effort estimation.

Abstract

In general, estimating software effort using a Use Case Point (UCP) size requires the use of productivity as a second prediction factor. However, there are three drawbacks to this approach: (1) there is no clear procedure for predicting productivity in the early stages, (2) the use of fixed or limited productivity ratios does not allow research to reflect the realities of the software industry, and (3) productivity from historical data is often challenging. The new UCP datasets now available allow us to perform further empirical investigations of the productivity variable in order to estimate the UCP effort. Accordingly, four different prediction models based on productivity were used. The results showed that learning productivity from historical data is more efficient than using classical approaches that rely on default or limited productivity values. In addition, predicting productivity from historical environmental factors is not often accurate. From here we conclude that productivity is an effective factor for estimating the software effort based on the UCP in the presence and absence of previous historical data. Moreover, productivity measurement should be flexible and adjustable when historical data is available