Obtaining high-resolution energy consumption data from a large, representative sample of homes is critical for research, but low response rates, sample bias and high recruitment costs form substantial barriers. The wide-spread installation of smart meters offers a novel route to access such data, but in countries like Great Britain (GB) consent is required from each household; a real barrier to large-scale sampling. In this paper we show how certain study design choices can impact the response rate and sample bias in energy studies requesting access to half-hourly smart meter data and (optional) survey completion. We used a randomised control trial (RCT) with a 3 x 2 x 2 factorial design to test incentives, message content/structure and a ‘push-to-web’ approach in a large-scale pilot for the future 10,000+ GB sample. Up to 4 mailings were sent to 18,000 addresses, recruiting 1711 participants (9.5% response rate) from England and Wales. Our results and recommendations can be used to help future energy studies to achieve greater response rates and improved representation. UK-based researchers can apply to use our longitudinal smart meter and contextual datasets.