27 Sep
Run 1: right answers, wrong questions
First live run on 40 mock reviews: 40/40 answered, 0.49 s average, $0.0037.
| Status | OK | Reply | Check me |
|---|---|---|---|
| Reviews | 15 | 14 | 11 |
The reply queue flagged 23 of 40 reviews, and 5★ reviews never reached "Very positive".
"Wrong colour sent… Returning it." scored 0.82 for needs a reply.
Jev followed the wording exactly. "Reports a problem that support should answer" fits every complaint, so the question was the bug.
So I rewrote the questions, not the thresholds: every level now states one idea, and every aspect gets a "not for" line. Raising the thresholds in code would have hidden a badly worded question instead of fixing it.
The reply question in questions.py (statement text):
"needs_reply": ("Needs a reply",− "Reports a problem that support should answer.",+ "Asks the store a question, asks for help or a fix, or is still waiting for an answer.",Reply flags dropped from 23 to 2, and 5★ sentiment moved from 3.10–3.92 to 3.75–3.99.