We present a new state-of-the-art theoretical approach to model correlated magnetic transition metal systems, by merging a form of self-consistent \emph{GW} (QS\emph{GW}) and dynamical mean field theory (DMFT). Recent high resolution ARPES and Haas-van Alphen data of two typical transition metal systems (Fe and Ni) are used as benchmark of the methodology. (i) Properties of Fe are very well described by QS\emph{GW} alone, owing to its nonlocal character. Agreement with experimental data is excellent provided that final-state scattering is taken into account. (ii) Due to the presence of strong local spin fluctuations, QS\emph{GW} alone is not able to provide a consistent description of the Ni ARPES data. To include spin fluctuations we develop a novel form of QS\emph{GW}+DMFT where DMFT modifies the self-energy in the spin channel only. We also present a heuristic method to include spin fluctuations in QS\emph{GW} by introducing an auxiliary effective magnetic field. Remarkable agreement is obtained with experiments by using our new methodology.