Your dashboard tells you that ten thousand people used your AI thing last week. It does not tell you which of them are quietly fed up and almost out the door. That is the question that matters, and almost no tool answers it.
Here is what we did about it.
The 11pm question
Every founder and product person ends up asking the same thing late at night: who is about to leave, and what do I do about it?
A chart of usage going up or down does not answer that. By the time the chart bends, the people who left have already left. You needed the answer two weeks earlier.
Sort people by what they're doing
Most tools sort users by their plan, their signup date, or their company size. None of those things tell you whether someone is happy.
What tells you whether someone is happy is what they're doing inside the product. Are they retrying the same thing five times? Are they hitting a tool that keeps failing? Did they used to come back every day and now they don't? That is the signal. So we sort users by that instead.
Four piles to start
Open Moda, click Users, and the people are in piles based on behavior:
- About to leave. Frustration creeping up, tool failures piling on, sessions thinning out.
- The good ones. Come back, use a lot, rarely complain.
- Stuck early. Hit a wall in the first few sessions and gave up.
- Quietly drifting. Used to be active, then weren't, never said a word.
The piles are not made by a person hand-picking who goes where. The signals already live in the chats we're watching for you. We just stack them up at the user level.
Click a person, see the receipts
Click a pile to see the people in it. Click a person to see why they're there. The exact chats. The exact moments something broke. The exact frustration we picked up and when.
Nothing is hand-wavy. Every claim points back to the chat that made us claim it.
Ask it in your terminal
The dashboard is fine. The thing we're actually excited about is that you can skip it. Open a terminal, ask in plain English:
Pull up my top 10 users about to leave and tell me why.
The CLI runs the queries against your data and hands you a ranked list with reasons attached. No clicking around.
The thread that connects them
A list of ten people about to leave is interesting. The thing that's actionable is what they have in common.
The last line of the answer is a single sentence on that: seven of the ten hit the same broken tool path since Tuesday, or four of them tried to do the same workflow and none of them finished it. That sentence is the part you can ship a fix for. The rest is context.
Why this beats sorting by plan
Plan, signup date, team size: those are useful for the marketing team. They are not useful for keeping people from leaving. The thing that tells you someone is leaving is not who they are, it's what they're doing.
Three retries on the same intent. A payment tool that fails every time. A drop-off the day a tool started returning nothing. That stuff is in your chat data already. Most teams just don't have a way to roll it up.
How to turn it on
If you already send us your chats, you already have it. Open Users in the dashboard, or grab the CLI and ask in your terminal. There is no extra wiring, no new events to send.
If you don't have a Moda tenant yet, ask for one. New teams onboard the same day, on real chats, with real piles.