Identifying the poorest individuals for aid and development intervention is challenging. The survey approach is prone to biases by local surveyors and supervisors. Respondents usually give the wrong response if they know the data is collected for aid purposes. In an alternative approach to the survey, researchers are exploring the use of night-time satellite imagery and mobile phone data. But these approaches have limitations too. In developing countries, the rural poor are not connected to the electric grid; hence, there is no night-time satellite image. In this paper, we demonstrate how bank transaction datasets give insightful information for targeting. These datasets are collected in real time, are accurate, and cost-efficient to prioritize counties for targeting. The knowledge gained from this study also provides valuable insights to develop data sharing polices, as input to other targeting models, to develop machine learning models, and to develop a national digital poverty map.