We offer a new method for estimating the relative size of repeating earthquakes using the singular value decomposition (SVD). This method takes advantage of the highly coherent waveforms of repeating earthquakes and arrives at far more precise and accurate descriptions of earthquake size than standard catalog techniques allow. We demonstrate that uncertainty in relative moment estimates is reduced from ±75% for standard coda-duration techniques employed by the network to an uncertainty of ±6.6% when the SVD method is used. This implies that a single-station estimate of moment using the SVD method has far less uncertainty than the whole-network estimates of moment based on coda duration. The SVD method offers a significant improvement in our ability to describe the size of repeating earthquakes and thus an opportunity to better understand how they accommodate slip as a function of time.