Abstract The technical feasibility of Enhanced Geothermal Systems (EGS) has been demonstrated for the first time in fractured crystalline basement rocks at the Soultz-sous Forets project (France), thus creating a unique and vast data base. At this EGS reference site, different hydraulic and chemical stimulation procedures and experiments were performed in four wells at three different reservoir levels located between 2 and 5 km depth. These measures enhanced significantly the hydraulic yield of the three reservoirs, in some instances by about two orders of magnitude. In this compilation of hydraulic data, we summarize the achievements at Soultz during the development of three reservoirs by more than 15 major stimulations over a 20-year period between 1988 and 2007. We evaluate the efficiency of the different injection schemes used and provide details on the performance history and testing conditions. In addition to the 52 experiments described for the testing phase, this compilation includes nine tests under operational conditions conducted over the 2008–2013 period. The evolution of hydraulic yield resulting from various injection, production, and circulation experiments is a major achievement of the Soultz reservoir development. This experience points to two important results: 1) the amount of total volume circulated between wells has a very significant effect on reservoir performance and 2) given the large flow rate variation a common linear trend of pressure increase at higher fluid flow rates develops that manifests over all three reservoirs. A strong focus is on the well tests in the intermediate reservoir allowing for a characterization of productivity and injectivity indices. Our analysis showed that initial hydraulic conditions from single-well injection tests are comparable to each other in the three reservoirs, but individual fault zones may determine the stimulation behaviour. We identify progressive cyclic injection in combination with circulation between wells reaching high hydraulic yields at comparatively low pressure. The Soultz data suggest how to maximize injection and minimize induced seismicity. This unique data base illustrates the learning curve achieved in Soultz and provides a strong basis for further conceptual model developments.