Trends
What changed since you last looked. Trends is the Observe tab’s opening page: the trailing week against the week before it, by domain, plus what appeared for the first time and the freshest issues. Every number on it is a pointer into a deeper page.
The domain table
One row per domain, plus a No domain row for calls that address an entity directly. Each cell shows this week’s number with the movement against last week beside it — ▲ 12%, colored by whether that direction is good news for that column:
| Column | This week’s… | Up means |
|---|---|---|
| Usage | Loads | More traffic — green |
| Quality | Data completeness % (load-weighted) | More grounded payloads — green |
| Avg Assembly | Average assembly time | Slower — red |
| Errors | Error rate (errors ÷ loads) | More failed loads — red |
| Issues | Issue occurrences | More degradation — red |
Click a domain to open Usage already filtered to it.
A domain with fewer than five loads in either week shows its number without a movement, because a percentage over a handful of loads is noise. Issues is the exception: it trends on its own count, so a jump from 2 occurrences to 20 shows at any volume.
Errors and Issues measure different things. Errors are loads that failed outright; issues are degradation inside loads that mostly succeeded. See Issues.
New This Week
Domains that served their first load this week, and users who made their first call. First ever, over all recorded history: a user who has called before never appears here again.
Each new user’s row names the domains they came through. This is where a rollout shows up: a team you just enabled appears here before anyone thinks to check.
Recent Issues
The ten most recent issue rows: when it was last seen, the entity, the domain, the category, the occurrence count, and a sample message. A link opens the full Issues page, which covers 14 days but names no domain — this list is where the domain shows.
Who sees it
Admins see every domain. Domain managers see the same page, over the whole org’s traffic. Observe is org-wide on purpose: degradation anywhere can affect answers everywhere.