Utilities across the world struggle to accurately measure electricity reliability on their grids; the average utility in a 109-country sample underestimates outages by a factor of 7 (relative to customers). While some utilities are addressing this challenge by installing smart meters, many utilities in emerging economies do not have the technical or budget capacity to deploy smart meters widely. In this paper, we analyze the size of deployment needed for outage detection via the GridWatch system, a novel crowdsourcing mobile application for measuring outages. Using outage data from Kenya Power and user mobility data, we consider different deployment sizes and varying levels of detection accuracy of the GridWatch app. Our results show that differences in neighborhood infrastructure and dynamics can necessitate a more than 3x difference in GridWatch deployment size to achieve the same outage detection performance, stressing the importance of deployment planning for a crowdsourced infrastructure monitoring system.