Particle acceleration to suprathermal energies in strong astrophysical shock waves is a widespread phenomenon, generally explained by diffusive shock acceleration. Such shocks can also amplify upstream magnetic field considerably beyond simple compression. The complex plasma physics processes involved are often parameterized by assuming that shocks put some fraction $epsilon_e$ of their energy into fast particles, and another fraction $epsilon_B$ into magnetic field. Modelers of shocks in supernovae, supernova remnants, and gamma-ray bursters, among other locations, often assume typical values for these fractions, presumed to remain constant in time. However, it is rare that enough properties of a source are independently constrained that values of the epsilons can be inferred directly. Supernova remnants (SNRs) can provide such circumstances. Here we summarize results from global fits to spatially integrated emission in six young SNRs, finding $10^{-4} le epsilon_e le 0.05$ and $0.001 le epsilon_B le 0.1$. These large variations might be put down to the differing ages and environments of these SNRs, so we conduct a detailed analysis of a single remnant, that of Keplers supernova. Both epsilons can be determined at seven different locations around the shock, and we find even larger ranges for both epsilons, as well as for their ratio (thus independent of the shock energy itself). We conclude that unknown factors have a large influence on the efficiency of both processes. Shock obliquity, upstream neutral fraction, or other possibilities need to be explored, while calculations assuming fixed values of the epsilons should be regarded as provisional.