The functioning of complex and diverse communities of microorganisms is critical in agriculture, engineering and medicine. Yet the coming together, or assembly, of these communities is poorly understood. When the microbial communities fail to function as we hoped they might it is often attributed to the 'wrong' community microbes having assembled. But we currently have no means of ensuring that the 'right' community assembles. Thus many economically important bio technologies are engineered by a process of trial and error. The ability to predict and hence manage the process of community assembly would thus have wide set applications.  We use high resolution time series pyrosequenced amplicons from multiple bioreactors to show that a simple birth-death-immigration process can capture much of the variance in community assembly and that by controlling immigration the probability that key species form part of the community can potentially be controlled.