In the last two decades, climate research has benefited from the continuous development of analysis systems dedicated to operational monitoring and forecasting, which opened up the possibility of exploiting the state of the art modeling and data assimilation tools to reconstruct and study the ocean during the past decades. This activity became feasible also thanks to the increasing availability of long time series of high-quality in situ and remotely-sensed observations. Retrospective analyses (or simply reanalyses or ocean syntheses), indeed combine quality controlled reprocessed ocean observations with a state-of-the-art ocean general circulation model (OGCM) using data assimilation methods to estimate the time-varying, three-dimensional state of the ocean. Ocean reanalyses benefit from data assimilation algorithms that are usually inherited from operational oceanography, although they require specific treatment in order to avoid spurious drifts stemming from instrumental or model biases. Unlike observation-only products, ocean reanalyses take advantage of time-varying atmospheric forcing, usually coming from an atmospheric reanalysis, and dynamical and physical balances implied by the OGCM. Here we give an excursus on the availability of global and regional ocean reanalyses, their applications, their strengths and weaknesses, and their future developments foreseen at the present time.