Better observability for GPTs other than cumulative conversation counts

It’s difficult to understand where my GPTs are falling short in production. Is there a way to read the anonymized user messages between my GPT and end users? This would help a ton in understanding why users are using the GPT, what users expected when messing around, and what I can do to improve the overall user experience.

Right now, we just get the cumulative conversation counts, which isn’t really helpful in finding answers to those questions. Thanks!


Yes, agreed. I am hoping for some basic stats like

  • unique chatters
  • average msgs per session or average time per session
  • benchmarks against other GPTs


hintloop does all those things for you and it is free!
The setup takes about two minutes.

This doesn’t seem like it helps with GPTs. I skimmed through the docs and didn’t find anything; how is this relevant?

1 Like

I don’t like the ergonomics of modifying the prompt and begging the GPT to call an action each message. It’s flakey from several of my experiements.

Seems like the best solution is for GPT store/OpenAI to surface this data or have a more robust way to call analytics actions reliably.


Could you share some details about your experiments? I am curious to see if it is possible to make the action calling more consistent for extremely long instructions.

@kevo1ution I agree completely. With revenue share coming by the end of the month we’ll need way better observability of chat stats from OpenAI. Has anyone heard any details on the upcoming revenue share since the initial announcement?