Structured interview records and rating evidence

A structured interview record should allow an authorized reviewer to reconstruct why a rating was selected without replaying the entire conversation or relying on an interviewer's memory. The goal is a concise evidence trail tied to the designed procedure, not a transcript and not a collection of personal impressions.

The rating chain

The most useful record keeps several linked objects distinct:

`role requirement -> question ID -> observed response evidence -> anchor -> rating rationale -> finalization event`

If any link is absent, the rating may be hard to interpret. A question ID without a note shows administration but not reasoning. A long note without an anchor can document conversation while leaving the score unexplained. A rating changed after panel discussion needs a separate event so the original assessment is not lost.

OPM describes structured interviews using predetermined questions and common rating scales. Its assessment material places interviews among selection methods that should be designed and administered appropriately. CIPD also discusses selection methods in relation to role and process. These sources support structure, but they do not make every structured form valid.

Record observable evidence, not a biography

A good note is selective. It captures response elements that matter to the assigned criterion, including a material absence after an approved neutral probe. It avoids speculation about personality, motivation, health, family, accent, appearance, or other irrelevant characteristics.

Field	Example of useful content	Avoid
Question ID	Stable identifier and version	A paraphrased mystery question
Observation	Candidate's actions, reasoning, and stated result	"Seems smart"
Anchor	Published level used at interview time	A private scoring scale
Rationale	Match and mismatch between observation and anchor	Repeating only the number
Author and time	Interviewer identity and finalization time	Shared team login
Amendment	Original value, correction, reason, approver	Silent overwrite

Verbatim quotations should be rare and necessary. Full transcripts and recordings can create additional privacy, security, access, and accommodation questions. A concise structured note may serve the assessment purpose with less data.

Separate original ratings from synthesis

Require interviewers to finalize their assigned ratings before seeing colleagues' evaluations when that matches the approved design. Locking should prevent casual editing, not legitimate correction. If an error is found, append an amendment that shows the prior value, new value, reason, actor, and timestamp.

The panel synthesis is a new record. It can identify agreements, disagreements, unresolved evidence, and the authorized recommendation. It should not rewrite every original score to manufacture consensus. A decision owner can weigh evidence under the approved rule while preserving what each interviewer recorded independently.

Reconstruct a sample

Choose completed interviews from a defined role and period. For each selected rating, ask a reviewer who was not on the panel to locate the governing question and anchor, read the observation, and explain whether the stated rationale is intelligible. Classify exceptions such as missing question version, unsupported label, anchor mismatch, copied note, or post-finalization change.

Do not ask the reviewer to rejudge the candidate from partial notes. The test concerns record integrity. Route recurring design problems to the assessment owner and suspected access or retention failures to the proper governance owner.

Legal and accessibility context

The Uniform Guidelines address validity and documentation in employee selection. EEOC guidance discusses tests and selection procedures, while its AI and Title VII resource underscores that software-assisted practices can still implicate discrimination law. If a system generates summaries or recommended ratings, retain enough information to identify that intervention and require authorized human review. Do not present generated prose as the interviewer's direct observation.

ADA.gov, DOL, and JAN provide disability and accommodation resources. The interview record should document the approved assessment conditions only to the extent necessary. Medical details should not appear in general panel notes. A format adjustment is not negative evidence, and interviewers should escalate uncertainty rather than speculate.

EEOC also publishes recordkeeping requirements for employers, but the applicable period and records depend on employer coverage and circumstances. The organization should map its own obligations and legal holds. This article intentionally does not prescribe one retention period.

Access, export, and disposal

Limit substantive records to people with an authorized assessment, decision, review, or compliance role. Coordinators may need completion status without needing note contents. Exported files should retain question version, authorship, and finalization state or be prohibited when that context cannot travel.

At disposal, account for copies in analytics stores, message attachments, backups, and vendor systems according to approved policy. Deleting the visible scorecard while retaining unrestricted exports is not meaningful lifecycle control. Conversely, deletion should not occur when an authorized legal hold or recordkeeping duty applies.

Limitations

A complete rating chain does not prove that the question measures the intended competency, that the anchor is valid, or that the interviewer applied it without bias. Notes inevitably summarize and may omit context. Audit samples can miss misconduct or off-system communication. Laws and candidate access rights vary across locations.

None of the reviewed publications supplies a universal note length, retention duration, or rating distribution. This brief makes no claim that more documentation improves hiring outcomes. Excess documentation can itself increase privacy and review risk.

Methodology and dated source notes

The editorial team reviewed the materials below on 2026-10-13. We used OPM and CIPD sources for structured-method context, the Uniform Guidelines and EEOC publications for selection and recordkeeping considerations, and disability resources for accommodation boundaries. Publisher dates and access dates are stated. We did not combine these materials into a numeric performance estimate.

EEOC, Employment Tests and Selection Procedures (2007)
NIST, AI Risk Management Framework (2023)
Uniform Guidelines on Employee Selection Procedures (current eCFR, accessed 2026-10-13)
OPM, Structured Interviews (accessed 2026-10-13)
OPM, Assessment and Selection (accessed 2026-10-13)
CIPD, Selection Methods (2024)
CIPD, Recruitment Overview (2024)
DOL, Disability Resources (accessed 2026-10-13)
EEOC, Disability Resources (updated 2024)
OPM, Disability Employment (accessed 2026-10-13)

Frequently asked questions

What belongs in a rating note?

Include concise, job-related response evidence and the reason it maps to the selected anchor. Exclude unrelated personal details, vague impressions, and medical speculation.

May a finalized rating be edited?

Permit controlled correction when justified, but retain the original value and an amendment trail. Panel synthesis should be stored as a later, separate event.

How long should interview records be retained?

Follow the employer's approved schedule based on applicable recordkeeping, litigation-hold, privacy, and business requirements. The sources here do not support one universal duration.

Use a free recruiting call to map record states, permissions, and review ownership while your authorized leaders keep all selection decisions.