Performance
Routing health
The routing health card monitors the four advanced routing features: scheduled calls, agent pools, broadcast groups, and fallback agents. The sections appear in that order, each with its own metrics, so you can verify that routing behaves the way you configured it. A colored dot at the start of each row summarizes its state — green is healthy, amber needs a look, red needs attention.
Schedule health
Each row is one enabled schedule. You can see its frequency, the last and next run times, the success rate for the selected range, the current streak of identical outcomes, and the average latency of successful runs. Expanding a row shows how many runs fired versus how many were expected for the range, plus the most recent failure message if there was one.
| Column | Description |
|---|---|
| Last / Next run | When the schedule last fired and when it is expected to fire again |
| Success | Percentage of runs in the range that succeeded |
| Streak | Consecutive runs with the same outcome — a green check for a success streak, a red mark for a failure streak |
| Avg ms | Average latency of the schedule's successful runs |
Pool health
Each row is one agent pool, showing its strategy, member count, calls in the range, a distribution bar of how those calls spread across members, the failure percentage, and utilization — the share of the owning agent's calls that went through the pool. Expanding a row reveals the per-member table.
| Member column | Description |
|---|---|
| Calls | Number of calls the pool strategy directed to this member |
| Success% | Percentage of those calls that succeeded |
| Avg ms | Average latency of the member's successful calls |
| Last call | When the member last handled a pool call |
For failover pools, the expanded view also counts failover events in the range — calls where the pool had to try more than one member before getting an answer.
Broadcast health
Each row is one broadcast group. Unlike a pool, which picks one member per call, a broadcast must reach every member — so the metric that matters here is delivery completeness. The row shows the member count, how many broadcasts fired in the range, the total deliveries they produced, the failure percentage, a distribution bar, and the number of partial broadcasts.
A partial broadcast is one that reached some members but not all of them. Even a single partial broadcast turns the row's status dot amber, because a group that silently misses members defeats the point of broadcasting. Expanding the row shows the per-member table — deliveries, success rate, average latency, and last delivery — so you can see exactly which member is missing messages.
Fallback activity
The last section tracks fallback usage for every agent that has a fallback configured. It shows the agent, its fallback agent, total calls to the agent in the range, the percentage answered by the fallback, how often the fallback itself succeeded, and when the last fallback happened. Expanding a row lists the top reasons the fallback's own calls failed.
At the bottom of the card, an amber line counts the agents that have no fallback configured. Click it to see which ones, and decide whether they need protection.
Next
Compare agents side by side with sortable leaderboard metrics. See Agent performance →