Tests of lepton-universality as rate ratios in $b\to s \ell\ell$ transitions can be predicted very accurately in the Standard Model. The deficits with respect to expectations reported by the LHCb experiment in muon-to-electron ratios of the $B\to K^{(*)}\ell\ell$ decay rates thus point to genuine manifestations of lepton non-universal new physics. In this paper, we analyse these measurements in the context of effective field theory. First, we discuss the interplay of the different operators in $R_K$ and $R_{K^*}$ and provide predictions for $R_{K^*}$ in the Standard Model and in new-physics scenarios that can explain $R_K$. We also provide approximate numerical formulas for these observables in bins of interest as functions of the relevant Wilson coefficients. Secondly, we perform frequentist fits to $R_{K}$ and $R_{K^*}$. The SM disagrees with these measurements at $3.7\sigma$ significance. We find excellent fits in scenarios with combinations of $\mathcal O_{9(10)}^\ell=\bar s\gamma^\mu b_L~\ell\gamma_\mu(\gamma_5) \ell$ operators, with pulls relative to the Standard Model in the region of $4\sigma$. An important conclusion of our analysis is that a lepton-specific contribution to $\mathcal O_{10}$ is essential to understand the data. Under the hypothesis that new-physics couples selectively to the muons, we also present fits to other $b\to s\mu\mu$ data with a conservative error assessment, and comment on more general scenarios. Finally, we discuss new lepton universality ratios that, if new physics is the origin of the observed discrepancy, should contribute to the statistically significant discovery of new physics in the near future.