Community-based demand-generation family planning programmes have been associated with increased contraceptive use in rural areas of Ghana. However, rigorous evaluations of such programmes in urban contexts are lacking. We used a retrospective, cross-sectional with comparison group design to estimate the immediate and sustained impact of the Willows intervention on modern contraceptive use in Kumasi, Ghana. The Willows intervention is a home-based counselling and referral programme for women in low-income urban settlements. We analysed data from a cross-sectional representative survey of 1205 women of reproductive age in the intervention area and 1108 women in a matched comparison site. The main outcome was women’s reported contraceptive use at: (1) baseline (January 2013); (2) programme close (December 2016); and (3) follow-up (August to October 2018). We estimated the programme effect at the community level and for women who reported receiving a family planning counselling visit. We used coarsened exact matching to assess the impact of the intervention relative to outcomes for matched comparison women. Comparing those who reported a family planning visit in the intervention area with matched comparison area women who reported no visit, we estimated a 10.5 percentage point increase in use of modern contraceptives from baseline to close (95%CI : 6.2, 14.8; P < 0.001) and a 7.6 percentage point increase from baseline to follow-up (95%CI : 3.3, 11.9; P < 0.001). However, only 20.2% of women in the Willows intervention area reported a visit. The Willows intervention, therefore, did not achieve its aim to reach all reproductive-aged women in the community. At the community level, we found no significant effect of the intervention at either programme close or 2 years later. We recommend that similar community-based interventions strive for greater outreach and simultaneously launch robust prospective impact evaluations.