We discuss recent progress by the RBC & UKQCD collaborations in the lattice calculation of the measure of Standard Model direct CP violation, $\epsilon'$, with physical kinematics. We present results with improved systematic errors resulting from the use of step-scaling to reduce the renormalization error, an analysis that also takes into account the one-loop-suppressed mixing with the $G_1$ two-quark operator. We also review our progress in decreasing the dominant statistical error on our result.