Purpose / Context - In the context of residential retrofit programs as an opportunity for climate change mitigation and health, the mechanisms that lead to unexpected outcomes in indoor temperature and energy consumption are not well understood. Paradoxical and surprising findings have been attributed to householder behaviour. Better knowledge of householder practices is needed for more effective designs of interventions. Methodology / Approach - This paper draws on a mixed methods study of an energy retrofit program for the homes of low-income, elderly or frail householders in Victoria, Australia. A quantitative exploration was combined with a phenomenological enquiry to explain the outcomes in warmth, heating energy consumption and householder satisfaction through householder practices and experiences. Results – This paper details three retrofit case studies that illustrated the diversity of outcomes, the take-back, the prebound and the expected effects. These cases highlighted how the interaction of the material quality of the home, householder capabilities and the meaning of heating shaped the changes in warmth and energy consumption. The householders’ evaluations of the intervention program highlighted that ‘what mattered’ was not necessarily the impact of the retrofit. Key Findings / Implications – While the effects discussed here are not new and the results lack generalisability, these case studies demonstrate that the theoretical predictions and the interpretation of the outcomes should be sensitive to contextual determinants. Interventions should include pre-study safety checks and aim for high thermal performance to be effective. Intervention studies should include a control group to assess confounding variables and bias. Originality - By identifying some contextual mechanisms that may enhance or hinder benefits in indoor temperatures and energy conservation, this study helps to better understand the effectiveness of residential energy efficiency interventions.