Satellite data have characteristics that in many respects are exact opposites of those of conventional observations, e.g. in terms of horizontal and vertical resolution, data rates, coverage, processing chains, etc. Exploiting the space-based data to the fullest extent has therefore been - and continues to be - a challenge for many operational numerical weather prediction centers. Over the last few years, significant developments have taken place at the Goddard Data Assimilation Office (DAO), both in the area of improving the use of existing satellite data from the operational meteorological satellites and in the area of evaluating proposed space-borne observing systems. In this presentation we will discuss some of the differences between conventional and space-based data and how these influence the direction of development of data assimilation systems. We will also show examples of impact studies aimed at improving the use of geostationary satellite wind observations in the DAO assimilation system.