A one-to-one comparison of the significant wave height reported by Voluntary Observing Ships (VOS) with the ERA-interim database is performed. A preliminary filtering procedure allows excluding reports affected by the most evident observations’ errors and removing the systematic biases in order to clean up the database. A statistical analysis confirms the previous finding regarding the goodness of the database when analyzing average values, but also highlights a large spreading of the observations. The work identifies critical areas and conditions in which the mismatch between numerical data and observations is more significant. The pattern of the differences reveals a regional consistency with VOS overestimating the significant wave height in marginal and enclosed seas as well as in the western sides of the oceans, while showing good matches in the eastern sides and offshore areas, except in the southern oceans where VOS leads to underestimations. The causes of such differences are finally discussed.