The main quality that we ascribe to case-based reasoning is its capacity to resolve a problem even when the application area is badly formalized. The one in which we have interested doesn’t even permit to represent cases per a set of important features. The objective of our system is finding out an evolution from a given date. Matching between evolutions is realized according to a viewpoint. We have hence defined a distributed architecture in which each module "reasons per case" according to different viewpoints on the evolution. The first originality of our work is the utilization of case-based reasoning for this kind of problems. The second one, and the most important, is a case representation independent from the analysis of the important features of the case. This one has naturally consequences for the matching technic. That’s specifically on this point that our approach is related to techniques being utilized by IRS (Information Retrieval System).