# Real-Time Performance Panel
Symptom: "the database is slow right now" and you need the culprit in minutes.
## Diagnose
Open the RTPP (Atlas UI, cluster, Real-Time tab). It is on by default. Watch:
- Operations by namespace: which collection is hot.
- Operation execution time: what is actually slow vs just frequent.
- Scanned vs returned: the ratio that fingers missing indexes.
## Confirm
Pick the top operation by total time. Copy its shape and run `explain("executionStats")` to confirm the plan (COLLSCAN, in-memory sort, or a fine IXSCAN that is just called too often).
## Fix
- Bad plan: add the index (see the COLLSCAN and query-targeting skills).
- Good plan, too frequent: cache it, batch it, or move it off the hot path.
- One huge operation: kill it if it is a runaway (with the right role), then fix the query.
## Verify
The RTPP calms down: the offending operation drops off the top list and the scanned-to-returned ratio normalizes. Screenshot the before/after for the incident record.