This paper uses a socio-technical building performance evaluation (BPE) approach to assess the pre- and post- actual performance of two discrete deep low energy retrofits in the UK - a Victorian solid-wall house and modern 1990s cavity-wall house. A 'low-energy first, then low-carbon' approach was adopted in both cases, to achieve an 80% reduction in annual CO2 emissions. Pre-retrofit, both houses had lower measured annual gas consumption as compared to predictions made by energy models, although the electricity consumption in the modern house was higher than modelled, due to occupancy pattern and occupant behaviour. Post-retrofit, it was found that the Victorian house achieved nearly 75% CO2 reduction, while the modern house achieved only 57% CO2 reduction-over the baseline emissions. Key reasons were higher than expected air permeability rates, installation issues with micro-renewable systems, lack of proper commissioning, usability of controls, occupant preferences and behaviour. Despite the gap between expected and actual carbon emissions, occupant comfort and satisfaction was significantly improved across both retrofits. This evidence-based understanding of the process and outcomes of deep low carbon retrofits is vital not only for learning and innovation, but also for scaling-up deep retrofit programmes for meeting national and international carbon targets. (C) 2016 Elsevier B.V. All rights reserved.