The survey quality score.
Every survey in the library carries a grade from A to F, scored out of 100. The score is the respondent risk composite made one level up: kept respondents contribute their own risk score, and deleted respondents contribute the full 100. Under the grade sits a composition strip showing the split across deleted, high, medium, low and clean respondents.
The grade bands
| Grade | Score | Read it as |
|---|---|---|
| A | 90 to 100 | Excellent sample. Little or nothing removed, mostly clean respondents. |
| B | 80 to 89 | Healthy. Normal background noise, nothing structural. |
| C | 70 to 79 | Acceptable, but look at where the weight sits before signing off. |
| D | 60 to 69 | A problem sample. Expect meaningful deletions and a conversation with the panel partner. |
| F | 0 to 59 | The field work itself needs questioning. |
What else the library row shows
- Sample health strip: the percentage split across deleted, high, medium, low and clean, worst first.
- AI cost: the total spent on AI checks for that survey. Rates are quoted per 1,000 answers, and a full survey pass is normally cents, not dollars.
- By: whose verdicts you are looking at.
A survey saved before sample health existed shows a dash: reopen and re-check the file to grade it.