The Respondent Risk tab.

Every respondent gets one risk score from 0 to 100. No respondent is ever judged on a single flag: each flag contributes up to its weight, the weights combine like independent pieces of evidence, and the total decides the tier.

Tiers

The filters

The filter row at the top is the fastest route through a sample:

• Human verdict: kept vs deleted, as decided by you.
• Tier: high, medium, low, clean.
• Flag: the major signal families (Bad open-ends, Straightliner, Speeder or skimmer, Duplicate, Pattern, Timing, Traps). A family filter alone can look alarming, so pair it with Severity.
• Severity: how hard each flag fired. Filtering Straightliners down to severe typically leaves a handful of respondents who flatlined every grid they touched.
• Type: the individual check inside a family. Fine detail; Severity is enough for most analysis.

Evidence and reflagging

The tool never just makes a claim. Expand any respondent's row and the dossier shows the evidence behind every flag:

• Click a grid in the Straightlining column and the flat answer row itself opens.
• Click an answer in the OE column to jump to that answer's flag group on the OE Checks tab, where you can reflag it on the spot.
• Duplicate, timing and pattern chips each open their own proof: the matching pair, the cluster, the profile.

If the evidence does not convince you, reflag it. A human verdict always outranks the automated one, and the risk score updates immediately.