Smart grids, typically driven by smart meters, enable the use of information and communication technologies to collect grid status in real-time. However, while smart meters are typically essential to the smart grid vision, many utilities globally have not installed smart meters mainly due to technical or budget constraints. In this paper we evaluate deployment strategies of GridWatch, a novel crowdsourcing system to detect electricity outages using smart phones. Using demographic, user mobility, and outage data relevant for Nairobi, Kenya, we develop an agent-based model (ABM) simulation to understand the factors that optimize the deployment strategy of GridWatch in different sub-regions of the city while maintaining high confidence of outage detection. Our results show that outage detection improves dramatically with increasing density of households per transformer, so a higher penetration of GridWatch devices is needed in areas with sparser grids.