# Candidate shortlist evidence record

Metix editorial template, October 28, 2026. Use in your spreadsheet or document editor without a Metix account. The example is entirely fictional. It is not a vacancy, a product demonstration, a benchmark or a candidate recommendation.

## Define the comparison before filling rows

Record the approved role ID, brief version, review date, reviewer and decision owner in your own role record. Define each criterion ID there. Keep essential requirements separate from preferences and name the evidence that would be sufficient. Obtain approval for changes, then recheck all affected candidates against the same version.

The fictional customer success opening uses version CS-1:

- K1, essential: evidence of personally owning a customer onboarding plan across multiple teams.
- K2, essential: an example of using support or product-usage evidence to investigate an adoption problem.
- K3, preference: SQL experience. This preference cannot establish either essential requirement.

This worked CSV is deliberately partial:3aliases times3criteria requires9candidate/criterion pairs, but it contains7. A/K3 and C/K3 are absent. The row checker cannot discover them. Add those two records with explicit unknowns or evidence in your own completed version, and reconcile every batch with its full approved criterion list. The sample is not a complete assessed shortlist.

The template does not establish eligibility, compensation, interest, availability, reference checks or readiness for an interview. Keep those questions in separate role/candidate records and verify them through the appropriate process. A sourced profile may not be an applicant or know about the role.

## Fill one row per candidate and criterion

Use a pseudonymous candidate alias, not personal contact details. criterion_id identifies the approved criterion; brief_version must match the version under review. criterion_kind is essential or preference.

Choose evidence_state explicitly:

- evidenced: the reviewer has a specific item to inspect. Record what it says, its private source locator and the actual observation date. This label is a claim requiring review, not proof that a requirement is met.
- unknown: the record does not establish the criterion. Write the open question, owner and next action. Do not invent a score or substitute a preference.
- conflicting: sources disagree. Keep the references, describe the disagreement, and assign someone to resolve it. Do not average conflicting claims.

source_ref can be a permitted internal record ID or source URL. Never publish a real candidate's private record or upload it to a public repository. The fictional-note handles in the example deliberately point nowhere. Keep original material accessible only to authorized reviewers. observed_on records when the material was actually checked; it is not a guarantee that the underlying information is current.

Record next_action and reviewer_decision separately. A record can contain evidence while the human decision remains pending. The example leaves every decision pending. Write decisions and reasons in your own authorized process, including any remaining role constraints and appropriate candidate communication.

## Optional local record check

Python 3, standard library only:

```sh
python3 shortlist-record-check.py candidate-shortlist-example.csv --brief-version CS-1 --review-date 2026-10-28
```

Compare stdout with candidate-shortlist-expected.csv. The checker flags duplicate candidate/criterion rows, mismatched brief versions, unsupported evidence labels, missing source/date fields, future observation dates and open questions without owners. It never opens a source, verifies a claim, calculates a fit score, ranks candidates, sends messages or makes a hiring decision. Exit success means the CSV was processed, not that the people or records passed review. Inspect every output row, including unknown/conflict flags.

record_fields_present means only that the required fields are present for that row. It cannot establish authenticity, completeness, freshness, fairness, legal compliance or candidate qualification. The checker cannot detect a completely omitted criterion or candidate; reconcile the row list with the approved brief and your actual review scope yourself. Do not use a maximum total score to bypass an essential gap.