The Uncertainty Challenge: To Centralize or Decentralize Requirements Engineering Decision-Making in Open Source Software Development?

Open Source Software Development (OSSD) teams exhibit emergent diverse communication structures due to a continuously changing and largely voluntary membership. Requirements Engineering (RE) practices in OSSD are distributed, dynamic, and depend on how teams contingently organize team communications and decision-making. OSSD requirements are primarily influenced by the external environment manifesting changes in requirements volume (quantity), velocity (rate of change), and variance (knowledge diversity). The setting challenges the principle that centralizing or decentralizing RE decisions is a binary either-or choice. Rather, per Information Processing View (IPV), it depends on the level and type of uncertainty in requirements volume, velocity, and variance. Per IPV, centralizing OSSD teams’ communication structure and thus decision-making affects positively RE success when measured in terms of task completion while increase in volume and velocity dampens the positive effect of centralization. Increase in variance - a multidimensional construct - has a nonlinear quadratic U-shaped moderating effect on the centralization effect. We validate these hypotheses using data from 255 OSSD projects covering 9904 development tasks (tickets) and find support for all hypotheses. The findings inform OSSD governance such that centralizing teams will generally improve RE task completion under low levels of volume, velocity, variance, and under high levels of variance. In contrast, decentralizing improves RE under moderate to high levels of volume and velocity and moderate levels of variance. Our study contributes to RE and OSSD research by showing that—contrary to the common conception of ad hoc self-organizing, lateral collaborations in OSSD teams—centralization matters for effective RE management, especially in low levels of requirements uncertainty.