Interview panel evidence under review

Question this record can answer

This review asks what an interview panel record can support when someone outside the original team reads it. The population is interview panel items that reached a review state during a declared window, and the unit is one item linked to its source, owner, decision, and close. The question is deliberately narrow so the answer does not drift into a general claim about assistant quality. Freeze the window and the inclusion rule before looking at results.

Case definition

Define the eligible cases before collection so the denominator cannot move toward the easiest examples. Include reopened, transferred, and closed-without-review items in the frame. Keep a stable key across the applicant tracker, interview notes, and contract repository so a case that moves between systems is not counted twice. Report every exclusion with a reason and an owner rather than dropping it quietly.

Counting rules

Write the start event, the stop event, the pause rule, and the missing-evidence rule before extraction begins. A measure such as time to close needs a numerator, a denominator, and an observation window. Without those three elements a percentage is only a mood with a decimal point. Record which clock each measure uses, because the queue clock and the calendar clock often disagree.

Evidence grades

Grade each interview panel item as verified, partial, or unsupported using the source that was actually observed. A blank field may mean unchecked, restricted, or genuinely absent, and each state leads to a different next action. The reviewer records what the source said at the time, not what the team remembers saying. Conflicting values stay visible until the agency principal chooses the controlling source and records the reason.

Selection method

Build the frame from the approved queue and the applicant tracker, interview notes, and contract repository rather than from a convenience list. Use consecutive or systematic sampling and include adverse cases such as a missing source, a conflicting instruction, an unavailable owner, and a forced deadline. Report the eligible count, the sample size, and the exact point where collection stopped. A sample chosen after the results are known cannot support the finding.

Second-reader check

Give an independent reviewer the same protected sample, the same definitions, and the same source access. Compare the classification of evidence, the owner named, and the close condition. Disagreements about policy return to the agency principal and are not settled by majority vote. The second reading is a check on the method, not a performance rating for the specialist. The principal's decision and the assistant's preparation stay on separate lines of the record.

Data protection

Use stable study keys and keep identity crosswalks in the approved system, not in a working spreadsheet. Restrict access to the minimum fields needed and remove temporary permissions when the review closes. Report what was redacted or unavailable instead of silently omitting it from the sample. Privacy frameworks inform the handling but they do not validate the record itself. The agency keeps the dated source attached to the case, not filed in a separate folder.

Limits and next use

The result depends on the window, the case mix, the system configuration, and reviewer judgment. Test the finding with a narrower window and with items closed near the end of the period. If the conclusion changes, report both readings and name the decision that depends on the definition. Use the finding to improve one control, then schedule a retest rather than treating the number as a verdict.