1 The Effect of Time Separation on Coordination Costs in Global Software Teams: A Dyad Model J. Alberto Espinosa American University, Washington DC Kogod School of Business alberto@american.edu Erran Carmel American University, Washington DC Kogod School of Business carmel@american.edu Abstract Research to date has not addressed the difficulties of coordinating across time zones in global software development. We present a preliminary collaboration model to help us understand the consequences of time separation on coordination costs. The model is for a team composed of dyads and each dyad consists of a task requestor and a task producer who have a sequential workflow dependency. The model is constructed with formulas for: production, coordination, and vulnerability costs for a number of: (1) collaboration modes; (2) time overlap conditions; (3) asynchronous and synchronous communications mechanisms, each of different quality and cost; and (4) production and delay cost rates. We describe the model and evaluate it with regression analysis using randomly generated observations. Our evaluation shows that the model adequately represents time- separated work and that time-separation effects are: (1) different and more complex than distance- separation effects; (2) asymmetric, depending on whether work time overlap between the two actors occurs at the beginning or end of an actor’s day; and (3) dependent on the amount of this overlap . 1. Introduction Working across geographic distances is not easy. Coordination in these contexts is challenging because of lean communication media, fewer opportunities for spontaneous interaction, less contextual reference, and lack of other benefits of co-location (e.g., presence awareness, workspace awareness), among other things. In spite of these difficulties, team work is increasingly carried out globally for a number of reasons: transport cost for digital products is low; production costs in developing nations are very low; and companies need to access specialized software talent and technical resources [4]. However, the cost- benefit tradeoffs for Global Software Teams (GSTs) are complex and not well understood. This has led researchers to study coordination in GSTs [4, 7, 12] and geographically dispersed teams in general [14, 20]. However, distance and time separation are often correlated [6, 22, 25], making it difficult to tease out the true effects of time zone separation. One important aspect that differentiates distance and time effects has to do with the fact that geographic distance is symmetric (i.e., distance AﾆB = distance BﾆA) while time separation is not (i.e., work time overlap occurs at the beginning of the day in one site and at the end of the day in the other site). Consequently, the timing when interactions occur and when tasks are processed may not make much difference in pure distance-separated conditions, but they do make a big difference in time-separated conditions. For example, the concept of “follow-the-sun” or “round-the-clock” software development [4] takes advantage of time zone differences to speed up project work, thus offsetting some of the problems of distance separation. While there are not many successful cases of “follow- the-sun” software work, the idea that team members in other sites (e.g., India) can advance the work while members in one’s site (e.g., Eastern U.S.) are sleeping has a lot of appeal in terms of potential productivity gains, at least in principle. However, because of the asymmetric nature of time-separated work, these teams need to pay close attention to timing issues. In fact, GSTs separated by time zones often adjust their work practices to deal with temporal factors – e.g., they establish overlap work hour windows; use liaison personnel whose work hours are the same as the other site; and program work to deliver output in batches to the other site at the end of the day. The present study is motivated by our interest in learning more about working arrangements involving different time-zones, which is very typical of today’s global software organizations. Because the research literature on global software development is almost silent on the issue of time separation, we feel that it is important to start from the ground up. In a prior study [5], we formulated a model to represent coordination costs for a single dyad with a workflow dependency, Copyright 2004 IEEE. Published in the Proceedings of the Hawai’i International Conference on System Sciences January 5-8, 2004, Big Island, Hawaii.