Weekly support QA · 2026-11-09
Estimated serious failures outside ordinary QA · 1
All 1 conversations outside ordinary QA have a determinate grade.
Did this example miss an escalation?
Ordinary QA reviewed 1 of 2 eligible conversations. This audit covers the other 1, including cases another control already knew about.
Confirmed cases
| Case | Evidence reference | Frozen screen | Finding | Owner | Action or work item |
|---|---|---|---|---|---|
| EX-1 | synthetic/ex1 | No screen | Newly found | Fictional QA | DEMO-1 |
Grading handoff
1 of 1 assignments have a determinate grade. 0 returned with insufficient evidence. Graders recorded 3 minutes (0.05 hours).
Give the reviewer grades.csv. Keep this lead report private so the reviewer can grade each case without seeing its screen flag. Review the listed evidence revision against the question in the packet.
Enter yes, no or insufficient under serious. For a serious finding, enter new, already_known or unknown under disposition. Record an owner and action for each new finding. For other answers, use none for no and unknown for insufficient. Record total review minutes for the row. Leave unanswered rows blank. A partial return keeps all other rows open. To resolve an insufficient grade, use the same row and include all minutes spent on it.
Import the return with python3 -m telryn weekly report.html --grades returned.csv.
Sampling and screen performance
Cases were drawn without replacement within each group. Population estimates weight each grade by the inverse of its selection probability.
| Screen result | Outside ordinary QA | Assigned | Selection probability |
|---|---|---|---|
| Flagged | 0 | 0 | — |
| Unflagged | 1 | 1 | 100.00% |
Screen version · none
These weighted measures compare the frozen screen to human grades for the ordinary-unreviewed population. A measure requires completed sampling and a nonzero denominator.