In this work we show that localization environments can be sampled extremely efficiently with a very small impact on accuracy. We investigate the influence of the number of fingerprints used per testing point in multiple environments using different numbers of testing points, different commodity WiFi hardware, and multiple sampling methodologies and have found that in all cases using the mean of 5 fingerprints per point trained performs nearly as well as 30, and that using a single fingerprint to represent a point is a reasonable approximation. We use well-known informational measures to assess the groups of samples used and demonstrate that increasing the number of samples taken results in marginally better approximations of the histogram of signals at a point, and that additional sampling has strong and immediate decreasing returns in reducing informational divergence, well beyond what assumptions of Gaussian distributions of signal strengths would require.
