The practice of soil improvement has received renewed attention in the last two decades. As a result, existing methods were improved and new ones developed. This paper presents a knowledge-based decision support system for the selection of soil improvement methods, Improve. The system uses a knowledge representation structure based on "windows" together with a best-first search algorithm. In this context, windows are mathematical representations of the restrictions to the values variables can take, combining the knowledge and its uncertainty in a unique entity. This form of knowledge representation has many advantages and allows for procedures not available in other systems, such as the development of composite solutions, the use of different evaluation functions, the search for lacunae, and the case-based representation of knowledge.