Designing a nationally representative probability based sample in developing countries can be challenging depending on availability of frames, resources, and survey timeline. Possibilities range from a large scale enumeration effort to construct a frame on one side and a random walk protocol on the other side. Our methodology represents an efficient and timely process which reduces the need for a full enumeration effort through the use a mix of Geographic Information System based information and census data. Geo- sampling also allows to avoid bias resulting from random walks. This paper outlines the steps of the methodology while highlighting its pros and cons as well as challenges faced in implementation as well as solutions developed while customizing the sampling design across different countries. Clustering and design effect are also explored and results compared to other traditional sampling methodologies.