Analyze Chatflows

Hi @viens666

That mismatch you’re seeing is something a lot of teams have noticed, and in 2025 the reporting around chatflows still works this way: the “sessions started” metric in the banner is tied to unique conversations, while the per-action counts below can inflate if users loop back into the same branch.

It does not start a whole new session, but HubSpot still logs that action as “fired again,” which is why your greeting step shows 41 when the banner says 27. HubSpot explains the difference between conversations and actions here: (the definitions section is worth a look).

(Analyze your rule-based chatbot's performance)

Your guess is also right: many “abandoned” sessions are really just people getting what they needed and dropping before hitting the final closing action. A good 2025 trick is to add a micro-conversion step right before the “do you have other questions” block something like a feedback quick-reply or a subtle CTA (“Would you like a transcript by email?”).

That way, you turn more of those exits into measurable completions. You can also filter reports using the Conversations object to analyze by outcome, which avoids relying only on the abandoned/completed split

(Report on chat and email conversations )

So the short answer: no new session is created when they loop, the inflated action count is expected behavior, and for 2025 the best workaround is designing exit points that feel natural but also mark a “completed” state for reporting. Hope this clears the confusion.