Use case use db1 db.foo.insert({ "keyword" : "abc" }) db.foo.insert({ "keyword" : "acd" }) db.foo.createIndex({ "keyword" :1}) db.bar.createIndex(index) db.foo.explain( "executionStats" ).count({ "keyword" :/a/i}) Execution stats "executionStats" : { "nReturned" : 0, "totalKeysExamined" : 2, "totalDocsExamined" : 2 } The confusion is: Why examine the documents even if the index has covered the querying field? What we found is: There is still a FETCH stage is used. And I think this stage can be optimized by removing. Details: { "nReturned" : 0, "totalKeysExamined" : 2, "totalDocsExamined" : 2, "executionStages" : { "stage" : "COUNT" , "nReturned" : 0, "nCounted" : 2, "inputStage" : { "stage" : "FETCH" , "nReturned" : 2, "docsExamined" : 2, "inputStage" : { "stage" : "IXSCAN" , "nReturned" : 2, "indexName" : "keyword_1" , "keysExamined" : 2 } } } }