The growing concern over water scarcity has driven the need for sustainable water management in urban environments. In Brazil, reliance on centralised supply systems highlights urban vulnerability, making residential water conservation crucial. Despite extensive research on rainwater harvesting (RWH) in large residential buildings, significant gaps remain regarding compact housing typologies with low occupancy rates. This study aimed to evaluate water end-uses and the potential for potable water savings through RWH systems in fou small residential units located in Florianópolis, Brazil. The methodology involved experimental monitoring of water consumption over a 15-day period. Flow rates of continuous appliances were measured volumetrically, while occupants manually recorded the frequency and duration of water uses through daily occupant logs. Computational simulations were performed using the Netuno computer programme, considering local historical rainfall data and varying storage tank capacities. Empirical results indicated an average water consumption ranging from 82 to 232 L/capita·day. The electric shower emerged as the primary consumer of potable water, accounting for 40% to 64% of total. Conversely, the non-potable demand, consisting mainly of toilet flushing and laundry, represented a significant fraction, ranging from 23% to over 51%. The simulations demonstrated that RWH systems are highly viable, yielding potable water savings between 22% and 50% with daily reliabilities exceeding 94%. Furthermore, the analysis successfully distinguished optimal tank sizes - 2,000 to 3,000 litres, defined by a 5%/m³ marginal efficiency - from maximum technical reliability, showing that due to low absolute non-potable demands and continuous tank overflows, smaller storage tanks (500 to 1,000 litres) offer superior spatial and operational efficiency without compromising performance. Although limited by unmetered building baselines, a short 15-day monitoring window, and a sample of four households, this study confirms the technical feasibility of RWH in compact typologies and provides essential empirical data to prevent system oversizing.