The ongoing liberalization of the energy market makes energy providers increasingly look at premium services – like personalized energy consulting - as preferred methods to bind existing customers and attract new ones. Providing such services, however, requires knowledge of specific properties of the customer's household - like its size and the number of persons living in it. In this paper, we investigate how such properties can be inferred from the fine-grained electricity consumption data provided by digital electricity meters. In particular, we focus on exploring which properties are both interesting and likely to be identified using well-known classification methods. To this end, we first elicit a set of interesting properties by performing in-depth interviews with employees of three different energy providers. We then explore a large set of electricity consumption traces using a self-organizing map. This analysis allows to identify a set of household properties that are likely to be inferable from electricity consumption data using standard classification methods. For instance, our results show that the size of a household and the income of its occupants are properties that are both highly useful to energy providers as well as likely to be detectable using an automatic classification system.