title: "RCP-nDCG@10: A New Approach to Enterprise Retrieval Quality | Cohere"

Traditional evaluation metrics for information retrieval, like nDCG, are limited because they fail to recognize relevant documents not present in the original human-labeled answer key. A new standard, RCP-nDCG@10, has been introduced to address this incomplete label problem in enterprise retrieval systems. This method uses a calibrated AI judge to assess the top 10 search results, identifying relevant documents that were initially missed. By incorporating these newly found documents, RCP-nDCG@10 offers a more complete and consistent way to measure the true quality of a retrieval system.