MainTrax is completing the analysis of another vendor's automatic sentiment scores. Ideally our client would like to replace NPS scores with the sentiment scores but so far the results vary greatly depending on numerous factors (call driver, repeat calls vs. first-time calls, length of call, etc.), making it challenging for the organization to consider the results at face value. Making it even more challenging is the fact that the vendor hides their sentiment scoring algorithms inside of a black box. Curious whether anyone has performed any thorough analysis of CallMiner's scores, what kind of success you may have had using those those and how they correlate to NPS scores.