This document discusses challenges in comparing metagenomic data from different environments and studies. It argues that when exploring a new environment, multiple methodological approaches should be used to capture natural and methodological variations. When performing global comparisons, methodological variations should be considered for all environments. Defining ecosystems precisely at the microorganism level is important. The author's vision is for projects like the Earth Microbiome Project to use flexible experimental designs informed by different experts to best represent microbial communities.