Every analysis of cosmic microwave background observations faces the challenge of compressing large amounts of data into the comparatively tiny spaces of power spectra and cosmological parameters. A variety of pipelines have been used to accomplish this task, balancing the goals of optimally computing these quantities and their errors while requiring minimal computational cost. In this talk, I will introduce the XFaster pipeline, which estimates maximum likelihood bandpowers and their covariances quickly and with a relatively small ensemble of signal and noise simulations. XFaster has been used by the BOOMERanG and Planck experiments, and has been repackaged and extended for SPIDER. I will discuss the algorithm and the code, including new null test and foreground fitting features implemented for SPIDER. I will show the results of the extensive pipeline testing we have performed, and some results from its application to SPIDER data.