We explore the evolution of the office market rents in Madrid by means of a spatial hedonical modelling. We estimate rental levels in a semi-annual basis by fitting a model with a set of hedonical exogenous variables as well as with the impacts of “neighbour” transactions. Our dataset comprises a new-letting contract list since 2003 to 2014 with a semi-annual structure including both a wide range of characteristics intrinsic to each leased property and the geocoding (UTM coordinates) of each transaction. This allows us to both uncover impacts on real rents of several hedonical property characteristics and outline a controlled dynamic path for real rents. The focal elements determining rents in the Madrid office market are localization with respect to the Central Business District (the farther from the CBD the lesser the rent), the age of the construction, the technical quality of the building and the international openness of the tenant, which is a proxy for its commitment with the landlord via a strong lease contract. We also prove the existence of relationships of influence of neighbour office rents on new-contract rent levels. In terms of the dynamics of the rents themselves, when estimating the rent of an archetypical office, our methodology effectively captures the downward span between 2003 and 2004 as a result of the “dot-com” bubble bust, the pick-up period of the recent Spanish economy boom (2005-2008) and the bust period of 2008-2014. Compared to the simple average and the deal-size weighted average, our methodology uncovers a lower average rent in the bust period. This may be explained by the limitation of average methodology to capture spatial feedback in prices as well as sample composition skewness that our methodology actually does not have.