We present a Principal Component Analysis for a hydrodynamic simulation and compare with CMS experimental data. While the results are reasonable for anisotropic flow, for multiplicity fluctuations they are qualitatively different. We argue that this is due to too large transverse momentum ($p_T$) fluctuations and $N-p_T$ covariance in the simulation than seen experimentally. In turn this is related to too large initial size fluctuations.