agent_ report per user and need the metricsAgent:read permission. See Permissions.
Volume
These metrics count threads, once per thread per bucket. All are time series only.
See Support volume.
Durations
Duration metrics are in minutes and acceptpercentile, an integer from 1 to 99. It defaults to 50, the median.
SLA compliance
SLA compliance is a fraction from 0 to 1, where0.95 means 95% of threads met their target. Single value only. See SLA compliance.
CSAT
CSAT metrics needfilters.surveyResponse.rating: 1 for Negative, 2 for Neutral, or 3 for Positive. Add filters.surveyResponse.surveyId to limit them to one survey. See CSAT scores.
All-time count
threads_all_time_count_done is the number of threads in Done right now. It’s single value only, and it rejects from and to.
Heatmaps
See Get a heatmap.
Per-user metrics
Per-user metrics credit each value to a user. InMETRIC mode they need groupBy: [{ dimension: ASSIGNEE }] for one result per user, or filters.userIds with another groupBy dimension. ASSIGNEE on these metrics groups by the user who gets the credit, which isn’t always the thread’s current assignee.
Status transitions only credit moves a user or machine user made, so Plain’s automatic changes don’t count toward anyone. Metrics that credit every user ever assigned count a thread once for each of those users, so per-user values don’t add up to the workspace value.

