Smart metering studies typically focus on quantifying behavior change. However, little is known about how users understand energy information and analyze and interpret feedback from energy data visualizations. To investigate this, we gave 13 participants from nine UK households an electricity power clamp meter. Prior to installing and using the device, we conducted interviews with participants to gauge their understanding of their home electricity consumption and found that participants varied considerably from limited to substantial energy literacy. Two weeks after the clamp meter had been installed, we conducted a contextual inquiry in which we asked participants to explain the web-based time series visualization of their recorded electricity data. We found that the visualization proved unfit: participants relied on memories and suggested likely routines, while widely being unable to reliably identify specific events in the data visualization. In follow-up interviews 3 months later, we found that participants' understanding of their home electricity consumption had hardly changed. Finally, we invited participants to generate ideas how smart electricity feedback could be optimized. They named different forms of disaggregation, higher temporal resolution, and interactivity as design requirements. In summary, these results suggest that people find home energy data very difficult to understand and link to everyday actions and behaviors.