Speaker(s): Nitin Joglekar
(Boston University)
Title:
Abstract
Many software product development (PD) projects tend to outsource a
significant portion of their tasks. With PD task outsourcing, a principal PD
firm typically restricts its attention to system level problem solving and
coordination tasks, while the remainder development takes place in a distributed
manner, within and across the boundary of the principal firm. We develop and
test a set of empirical models to explore the costs of organizing distributed PD
tasks using the transaction cost economics (TCE) theory. These models compare
the aggregate transaction costs for two fundamentally different types of tasks
that constitute a PD cycle: generation and test. We present data on 71 system
level tasks from 11 software based PD projects at a medical devices firm. These
data characterize the differences in the cost structure while sourcing portions
of generation and test tasks. Analyses show that asset specificity - as
indicated by in-house technical problem solving effort - is a significant
predictor not only of the likelihood of outsourcing a portion of PD task, but
also of the associated costs of coordination during generation and test.
Contrary to conventional wisdom on software cost estimation, we show that with
distributed development, the component level complexity - measured by outsourced
problem solving effort - is not a significant predictor of the coordination
costs during the development process.