Utilities world wide are introducing advanced automation in planning, operations, and maintenance of their distribution networks. However, most of these automation solutions rely on an accurate connectivity model of the distribution network, which can easily become outdated due to system changes. This work presents novel analytics techniques that can infer the connectivity model from electrical voltage measurements. The techniques may be used to infer phase connectivity either solely from customer voltage data, or in combination with measurements from the medium voltage network and any existing but partly accurate phasing information. We report analysis results based on customer and feeder voltage measurements collected from a large distribution circuit that serves more than 1.5K single-phase customers. Our results indicate that voltage measurements spanning a few days could be utilized to infer customer phase with high accuracy (93%). To the best of our knowledge, this is the first measurement study of voltage data collected from low and medium voltage segments of a feeder circuit and its use in estimating customer phase.