Empirical Analysis on Parallel Tasks in Crowdsourcing Software Development

Crowdsourcing has become a popular option for rapid acquisition, with reported benefits such as shortened schedule due to mass parallel development, innovative solutions based on the "wisdom of crowds", and reduced cost due to the pre-pricing and bidding effects. However, most of existing studies on software crowdsourcing are focusing on individual task level, providing limited insights on the practice as well as outcomes at overall project level. To develop better understanding of crowdsourcing-based software projects, this paper reports an empirical study on analyzing four largest projects on Topcoder platform that intensively leverage crowdsourcing throughout the product implementation, testing, and assembly phases. The analysis results conclude that: (1) crowdsourcing task scheduling follows typical patterns including prototyping, component development, bug hunt, and assembly and coding (2) budget phase distribution patterns does not following traditional patterns, and uploading task rate is not representing same budget rate associated with them as about 75% of uploaded tasks would price under 67% of total project budget, (3) Higher degree of parallelism would lead to higher demand for competing on tasks and shorter planning schedule to complete the project consequently better resource allocation.