Relying on traditional household surveys for poverty data is time consuming and expensive. What’s more, by the time the data are collected and analyzed, it is often out of date. But there are alternatives for estimating and mapping poverty with the goal of accelerating and expanding financial inclusion and helping DFS providers target the poorest. Machine learning algorithms can, for example, be trained to predict poverty based on imagery captured by satellites and from call detail records, which document mobile phone usage. For this research study, the IFC Mastercard Foundation Partnership for Financial Inclusion collaborated with the Stanford University Sustainability and Artificial Intelligence Lab to advance existing poverty prediction models to generate poverty estimates at neighborhood-level resolution, which is much more refined than macro-level estimates produced by research to date. Satellite Imagery and call detail records (CDR), validated by ground-truth surveys, were used to develop models that can predict poverty in Ghana and Uganda. The study finds that it is possible to make meaningful welfare estimates based on satellite imagery combined with geo-spatial boosting at the neighborhood-level when lower levels of precision are acceptable.