Trends
Real, dated counts and artifact coverage for the recorded demo calls. Show the synthetic lab instead.
Sat 26 Sep 2026 15:40 UTC (Sat 26 Sep 2026 10:40 CDT) to Tue 29 Sep 2026 15:40 UTC (Tue 29 Sep 2026 10:40 CDT), by the server's clock right now. The boundary is inclusive: a call exactly at the start of the window is included, one second earlier is not.
Not counted here: 56 older than this window.
Calls by day, in the selected range
| Date | Calls | Connected | Completed | Not connected | Recordings | Turns |
|---|---|---|---|---|---|---|
| 2026-09-28 | 1 | 1 | 1 | 0 | 1 | 25 |
| 2026-09-27 | 1 | 1 | 1 | 0 | 1 | 27 |
Artifact coverage, in the selected range
| Artifact | Present | Of |
|---|---|---|
| Recording | 2 | 2 calls |
| Stored weather briefing | 1 | 2 calls |
| Review findings | 0 | 2 calls |
Trends across the recorded calls
Worked out from the calls in this release and nothing else. Shares are shown because there are at least 30 calls. Length is how long the call lasted. It is not how fast the service answered, which these calls do not record.
| Date | Calls | Connected | Completed | Not connected | Recordings | Turns |
|---|---|---|---|---|---|---|
| 2026-09-28 | 1 | 1 | 1 | 0 | 1 | 25 |
| 2026-09-27 | 1 | 1 | 1 | 0 | 1 | 27 |
| 2026-09-23 | 2 | 1 | 1 | 1 | 1 | 45 |
| 2026-09-22 | 11 | 10 | 9 | 1 | 9 | 170 |
| 2026-09-21 | 1 | 1 | 1 | 0 | 1 | 30 |
| 2026-09-20 | 1 | 1 | 1 | 0 | 1 | 1 |
| 2026-09-19 | 16 | 13 | 13 | 3 | 13 | 324 |
| 2026-09-18 | 25 | 6 | 6 | 19 | 6 | 20 |
| Measure | Count |
|---|---|
| Reached a conversation | 34 of 58 calls (58.6%) |
| Completed | 33 of 58 calls (56.9%) |
| Did not connect | 24 of 58 calls (41.4%) |
| Completed calls with a recording | 33 of 33 completed calls (100%) |
| Completed calls with a transcript | 33 of 33 completed calls (100%) |
| Completed calls with a stored weather briefing | 16 of 33 completed calls (48.5%) |
| Turns in the transcripts of completed calls | 634 |
| Length of completed calls: middle · 90th percentile | 247 s · 441 s |
What the checks found
Local checks of what was recorded, not a judgement. Each check reads recorded text or fields (what was said, what a tool returned, whether a recording or a review finding exists). None listens to a recording and none uses a model. A check says "passed" only when the record shows it, "flagged" only when the record shows the problem, and otherwise "unknown": a call is never counted as fine because nothing failed. Version 1 of the checks. The items these raise are also listed, apart from the synthetic ones, on the Improvement queue.
| Check | Flagged | Passed | Unknown | Share of the calls it could decide | Latest flagged |
|---|---|---|---|---|---|
| The notice that this is not an official briefing was said by the agent Flagged means: The agent spoke but did not give the notice | 3 | 31 | 24 | 8.8% of 34 | call · call · call |
| The weather lookup returned a result Flagged means: The weather lookup returned an error | 1 | 15 | 42 | 1 of 16 (a share is shown from 30 calls) | call |
| A weather briefing was stored for the call Flagged means: The pilot was told the briefing was unavailable | 0 | 16 | 42 | 0 of 16 (a share is shown from 30 calls) | none |
| The archive's automatic checks Flagged means: An automatic check failed | 5 | 29 | 24 | 14.7% of 34 | call · call · call |
| A finished call has a transcript Flagged means: A finished call has no transcript | 0 | 33 | 25 | 0% of 33 | none |
| A finished call has a recording Flagged means: A finished call has no recording | 0 | 33 | 25 | 0% of 33 | none |
| Review findings are verified or dismissed Flagged means: A review finding is open, or implemented and not yet verified | 3 | 1 | 54 | 3 of 4 (a share is shown from 30 calls) | call · call · call |