Retrieval Augmented Generation (RAG) systems to enhance large language models (LLMs) by providing them with relevant external information. Ideally, the LLM produces the correct answer or responds with "I don't know" when certain key information is lacking. A main challenge with RAG systems is that they may mislead the user with hallucinated (and therefore incorrect) information. The authors believe that the context's relevance alone is the wrong thing to measure - they really want to know whether it provides enough information for the LLM to answer the question or not. The authors define context as "sufficient" if it contains all the necessary information to provide a definitive answer to the query and "insufficient" if it lacks the necessary information. The authors develop a way to quantify context sufficiency for LLMs and launch the LLM Re-Ranker in the Vertex AI RAG Engine. The authors show that it's possible to know when an LLM has enough information to provide a correct answer to a question. The authors use these ideas to analyze the factors that influence the performance of RAG systems and to analyze when and why they succeed or fail. The authors develop a sufficient context autorater that evaluates query-context pairs and show that they can classify sufficient context with very high accuracy. The authors use their sufficient context autorater to analyze the performance of various LLMs and datasets, leading to several key findings.