Mean average precision (MAP) is an metric for evaluating a ranking model across many queries. As its name suggests, it is simply the arithmetic mean of the average precision for each of the queries. In a production context, there is typically a predefined set of test queries used to evaluate the performance of a release candidate.