Use evaluation metrics in the Einstein Studio Retriever Playground to objectively measure the performance of your RAG application. When you test a retriever in advanced mode, the playground generates a context recall score based on a reference answer.
Where: This change applies to Data Cloud Developer, Enterprise, Performance, and Unlimited editions. Available in Government Cloud.
When: This feature is available starting in April 2026.
Why: The context recall score measures how many relevant documents or chunks of information the retriever successfully fetches based on a reference answer. A higher recall score means that the retriever missed fewer relevant documents. Testing retrievers in the advanced mode consumes flex credits.
How: To test a retriever, select the advanced mode in the retriever playground. The configuration panel shows the search index and retriever details. Enter a test question and an expected answer. After you run the test, the metrics panel displays the context recall score along with other retriever metrics, such as retrieval response time and retrieved chunks.
Retriever Playground Test Panel
Retriever Playground Metrics Panel
