Expression of sufficient quantities of soluble protein for structural biology and other applications is often a very difficult task, especially when multimilligram quantities are required. In order to improve yield, solubility or crystallisability of a protein, it is common to subclone shorter genetic constructs corresponding to single- or multi-domain fragments. However, it is not always clear where domain boundaries are located, especially when working on novel targets with little or no sequence similarity to other proteins. Several methods have been described employing aspects of directed evolution to the recombinant expression of challenging proteins. These combine the construction of a random library of genetic constructs of a target with a screening or selection process to identify solubly expressing protein fragments. Here we review several datasets from the ESPRIT (Expression of Soluble Proteins by Random Incremental Truncation) technology to provide a view on its capabilities. Firstly, we demonstrate how it functions using the well-characterised NF-kappaB p50 transcription factor as a model system. Secondly, application of ESPRIT to the challenging PB2 subunit of influenza polymerase has led to several novel atomic resolution structures; here we present an overview of the screening phase of that project. Thirdly, analysis of the human kinase TBK1 is presented to show how the ESPRIT technology rapidly addresses the compatibility of challenging targets with the Escherichia coli expression system.