Dashboard
By user
The user breakdown ranks every user by their total LLM spend. It's the fastest way to understand your unit economics and spot unusual usage patterns.
What you'll see
- User ID — the identifier you passed via the SDK's
userIdparameter - Total cost — cumulative spend attributed to this user
- % of bill — this user's share of total spend
- Calls — total LLM calls made by this user
- Avg cost / call — average spend per call for this user
- Flag — automatic flag based on spend level: High spend, Watch, or Normal
What to look for
Three signals worth acting on:
- High spend users on free plans — if your top 5% of users generate 40%+ of your LLM bill and they're on a free tier, your pricing model has a problem.
- Unusually high call counts — a single user making 10x the calls of anyone else may indicate abuse or a client-side bug causing duplicate requests.
- High avg cost / call for specific users — could mean certain users are triggering expensive code paths others aren't.
User attribution requires passing userId in every SDK call. Calls without a user ID appear as anonymous and cannot be attributed retroactively.
Enabling user tracking
await trackedCall({userId: session.userId, // your app's user identifierfeature: 'search', ...})Next: Call log →