Given two text schemas, how can one map each item from the first schema to an item of the second one, in order to have the best semantic correspondence? While this latter criterion has to be defined by a human, in this paper we present a new application of the Case-Based Reasoning (or CBR) methodology, that helps the user in finding a good match thanks to a score function taking into account previous results (i.e. solved cases). Such a technique makes the schema mapping easier and less error-prone. The proposed solution is in use in an industrial context, at the Multiposting firm.