You can use Bot Analytics to evaluate a bot's performance and thereby identify tuning opportunities. There are a few key approaches to the data: MACS and intents.


Review of a bot's Meaningful Automated Conversation Score (MACS) data is a great way to identify opportunities for bot/intent tuning.

  • For an introduction to MACS, its benefits, its scoring, and more, see the introduction in the Knowledge Center.
  • For information on using MACS within Bot Analytics, see this topic in this Developer Center.


Intent tuning is an important step in optimizing a bot for high performance.

You can readily determine that there are opportunities for intent tuning if you see a low Intent Match Rate for the account overall, or for a specific bot.

A bot on the main dashboard with a low intent match rate

A bot’s Intents view displays both “matched intents” and “unmatched phrases” for your bot’s intents, patterns and attached knowledge bases. You can see all of them together, or you can view them individually using the Source dropdown menu on the left.

The filter on the Intents page that lets you filter the view to include data for all intents or a specific intent

Above, we’re looking at the matched intents for this bot. Tapping Unmatched Phrases displays the user utterances that didn’t return any result from the bot patterns, intents or knowledge bases.

If you see utterances in the Unmatched Phrases that should be matching a particular intent or Knowledge Base article, you can add them to the training phrases for these items. Keep in mind the best practices for creating training phrases for intents and internal knowledge bases.