Integrated home garden interventions combine training in gardening practices with education about nutrition knowledge. Such interventions have been shown to improve nutrition behaviour in low income countries. However, to date rigorous evidence is lacking for their long-term impact. We test the impact of an integrated home garden intervention on vegetable production and consumption three years after the intervention ended. We analyse three rounds of survey data for 224 control and 395 intervention households in rural Bangladesh. Three years after the intervention, the average impact on vegetable production per household was 43 kg/year (+ 49% over baseline levels; p < 0.01), and the effect was not statistically different from the impact one year after the intervention, which demonstrates that impact was maintained in the long-term. The impact on the micronutrient supply for iron, zinc, folate and pro-vitamin A from home gardens was maintained in the long-term. These impacts may have been driven by the long-term improvements in women’s nutrition knowledge and gardening practices, explaining the sustainability of the behavioural nutrition change. We also identify positive impacts on women’s empowerment and women’s output market participation, highlighting how integrated programs, even if modest in scope, can be drivers of social change.