Controlling and minimizing coordination costs has been shown to be an important factor to reduce overall project performance in distributed software development. In this research-in-progress paper we investigate the effects of software complexity, software integration, distributed labor division policies, learning effects on software coordination costs. Drawing from data collected on 130 software construction cycles in 34 large projects of a leading offshore development firm, we first present our analysis on how coordination costs relate to team organization factors and complexity of evolving software. We base our analytic model of coordination costs in offshore software development on these empirical relationships, and give an overview of our modeling approach. We apply our model of software coordination costs to develop resource allocation policies in the projects we studied. We consider both waterfall and iterative software development methodologies and also tandem and parallel integration schemes. Our modeling approach helps managers to develop a dynamic coordination policy to aid iterative software development in distributed development environments.