Signups are happening. Users still disappear. I added an activity log and an AI support chat.

One of the more uncomfortable stages of building a SaaS is when people sign up, click around, and then disappear.

That is where Speechara.Ai is right now. We have registrations, but not enough repeat usage. The easy explanation is to blame users for not understanding the product. The more useful explanation is that we have not made the first successful path obvious enough.

So I added two things.

First, an activity log. It records important product events, not every mouse movement. I want to see where a new user starts, what they try, and the step where they stop. If many people reach Settings but never start a session, that is a different problem from people starting a session and failing to get a transcript. The log is limited to product events and does not include meeting content.

Second, an AI support chat that answers product questions in context. The goal is not to replace documentation or pretend the assistant knows everything. It is to remove the small question that can otherwise end a trial: how do I capture this audio, where did my transcript go, can I change the language, and so on.

I am still unsure whether this is the right way to learn. An activity log can tell me where users stop, but not always why. A support chat can answer questions, but it can also hide a confusing interface instead of fixing it.

For those of you running a SaaS, what signals do you use to understand why users become inactive? Do you rely on product analytics, session recordings, interviews, support conversations, or something else? And where do you draw the line between helpful instrumentation and too much tracking?