Good scientific work requires comprehensible, transparent and reproducible research. One way to ensure this is to include all data relevant to a study or evaluation when publishing an article. This data should be at least aggregated or anonymized, at best compact and complete, but always resilient.
In this paper we present ProSA, a system for calculating the minimal necessary data set, called sub-database. For this, we combine the chase --- a set of algorithms for transforming databases --- with additional provenance information. We present the implementation of provenance guided by the ProSA pipeline and show its use to generate an optimized sub-database. In the demo, we show how the ProSA GUI looks like and present some applications and extensions.