While long Doppler ensembles are, in principle, beneficial for velocity estimates, short acoustic pulses must be used in microbubble contrast-enhanced (CE) Doppler to mitigate microbubble destruction. This introduces inherent tradeoffs in velocity estimates with autocorrelators, which are studied here. A model of the autocorrelation function adapted to the microbubble Doppler signal accounting for transit time, the echo frequency uncertainty, and contrast-agent destruction is derived and validated in vitro. It is further demonstrated that a local measurement of the center frequency of the microbubble echo is essential in order to avoid significant bias in velocity estimates arising from the linear and nonlinear frequency-dependent scattering of microbubbles and compensate for the inherent speckle nature of the received echo frequency. For these reasons, broadband Doppler estimators (2-D autocorrelator and Radon projection) are better suited than simpler narrow-band estimators (1-D autocorrelator and 1-D Fourier transform) for CE flow assessment. A case study of perfusion in a VX-2 carcinoma using CE plane-wave Doppler is also shown. We demonstrate that even when considering all uncertainties associated with microbubble-related decorrelation (destruction, pulse bandwidth, transit time, and flow gradient) and the need for real-time imaging, a coefficient of variation of 4% on the axial velocity is achievable with plane-wave imaging.