Talent SummonerEvidence-Based Candidate Evaluation: What to Capture
A practical evidence taxonomy for candidate evaluation, with traceable sources, unknowns, contradictions, review controls and human-owned decisions.
Evidence-based evaluation is a record of what a reviewer can support, what remains uncertain and why a human-owned next action follows. It is not a scorecard, an interview script or a copyable shortlist. Use it as the evidence and governance layer that connects those documents to a frozen role brief.
Define the evidence before reviewing a person
Start with one criterion from the approved brief and its version. State what the requirement means, what evidence could satisfy it and what would need verification. Keep must-haves, nice-to-haves and genuine operating boundaries separate. A title or keyword is not evidence of ownership, scope or results.
Use this taxonomy for each captured item:
| Evidence type | Capture | Do not claim |
|---|---|---|
| Source evidence | Exact or concise wording, source, trace and date | More scope or recency than the source establishes |
| Candidate-provided evidence | CV detail, answer, portfolio or explanation supplied by the candidate, with context and date | That a claim is independently verified |
| Work-sample or interview observation | Task output, answer, behaviour relevant to the criterion, setting and observer | A personality trait or future performance |
| Reviewer interpretation | How the item relates to the frozen criterion, with confidence in that reading | Confidence that the candidate will succeed |
| Unknown | The question the available material does not answer | Absence, weakness or automatic rejection |
| Contradiction | Conflicting sources, dates or scope, with both traces | A final conclusion before a human check |
| Decision reason | Job-related reason for progress, verify, hold, revise or close | A rank, impression or unsupported causal claim |
Keep the source and interpretation in separate fields. Confidence describes the interpretation: high may mean the source is specific and clearly relevant; low may reflect vague wording, stale context or a substantial inference. It is not a candidate-quality rating, probability or prediction.
Capture a compact, reviewable record
The record should let another reviewer retrace the path from source to action without collecting a whole life history. This schema is a governance aid, not a replacement for the shortlist or scorecard:
| Field | Minimum record |
|---|---|
| Criterion and relevance | Frozen brief version; must-have, nice-to-have or boundary; why it matters to the role |
| Source trace | Source type, URL or internal ID, short passage, captured date and source date if available |
| Scope and context | Project, responsibility, setting, personal contribution, ownership and relevant outcome only where established |
| Status and interpretation | Supported, adjacent, unknown or contradicted; separate explanation and confidence in that explanation |
| Verification | Specific question or check, evidence needed, named owner and due date |
| Access and retention | Permitted viewers, access change or revocation, retention period and deletion review date |
| History and decision | Record version, reviewer and review date, changes, decision reason, next action and approver |
Source date and capture date are different. A public profile captured today may describe work from years ago. Record both when known, and flag stale or undated material. Relevance means the criterion is frozen before the evidence is read; do not add a requirement to explain a preferred candidate.
Verification should test the missing fact, not invite a general impression. Ask about scope, ownership, context or an observable result. Apply a comparable question or work sample to comparable candidates. Record who checks it, by when and what would change the status. A contradiction needs the competing traces and a human review; it is not a reason for automatic rejection.
Fictional evidence-to-decision example
The following is explicitly fictional and demonstrates record structure only. It is not a benchmark, validated score or claim about a real candidate.
Role brief v2, criterion: own production API migrations (must-have)
Source evidence: Fictional CV, "Migrated payments workloads to Kubernetes," captured 2026-08-18.
Candidate-provided context: Fictional Rowan says they planned the cutover and wrote the rollback runbook.
Observation: Fictional work discussion describes one migration decision and its constraint; observer: fictional reviewer.
Interpretation: Relevant migration evidence; production ownership is plausible but not established. Confidence: medium.
Status: Adjacent. Unknown: services in scope and who operated the release.
Verification: Ask for personal ownership, production scope and monitoring; owner: fictional engineering lead; due: 2026-08-22.
Decision reason: Verify because the criterion is material and the evidence could change the decision.The action is verify, not progress because of a polished claim and not close because a CV is brief. If the answer supports the frozen criterion, update the record to supported and retain the earlier version. If it conflicts, record the contradiction and the human decision reason. Do not turn an observation into "will be a strong operator" or any other unsupported causal or predictive claim.
Limit collection and assign ownership
Capture only job-relevant information. Exclude protected or sensitive personal traits, irrelevant proxies, personality guesses, unnecessary personal data and third-party confidential material. Do not copy health, family, ethnicity, religion, disability, gender or similar details into the evaluation record unless a separate approved process requires a strictly necessary field. Do not replace capability evidence with school prestige, postcode, accent, employer brand or vague "culture fit". Missing evidence is unknown.
Name the reviewer, decision owner and next-action owner. Limit access to people involved in the decision, record when access changes and schedule a retention or deletion review. Keep version history when the brief, source interpretation or decision reason changes; never silently overwrite an earlier record. Follow your organisation's policy and current qualified guidance for privacy or employment questions. This is general process information, not legal advice.
Where Talent Summoner fits
Talent Summoner is our product for upstream sourcing and ranking. As verified on 19 August 2026, candidate sourcing starts with a role brief and searches LinkedIn, GitHub and other public professional sources, while candidate ranking organises CVs against a supplied job description and returns ranked reasoning. Use those outputs as review inputs, then inspect sources and capture evidence yourself. Pricing covers sourcing options.
Talent Summoner does not provide interview scoring, recording, transcription or notetaking. Your people conduct interviews, review work samples, verify claims, control access and retention, and own every next action and final decision. It is not an ATS or hiring pipeline, does not automatically reject candidates and is not the decision-maker.
Wrong fit for this guide
This guide is a poor fit for collecting broad personal profiles, replacing a frozen brief, assigning an automated score or deciding from rank alone. It also does not turn a public claim into verified truth. Use a role-specific process with an accountable reviewer when the evidence is material, disputed or sensitive.
FAQ
Is reviewer interpretation evidence?
No. Evidence is the source, candidate material or observed work. Interpretation explains its relevance and should show confidence in that reading separately.
Does missing evidence mean the candidate cannot do the work?
No. Mark the criterion unknown, write a focused verification question and give comparable candidates a consistent opportunity to provide relevant evidence.
Should a contradiction close a candidate?
Not automatically. Preserve both traces, check dates and context, ask the named reviewer to resolve it and record a job-related decision reason.
Does Talent Summoner capture interviews or make hiring decisions?
No. It supports upstream sourcing and ranking. Humans own interview observations, verification, access, retention and the final decision; it is not an ATS, pipeline or auto-reject system.
For the next open role, freeze one criterion, complete the schema for each material item and assign a named reviewer and verification date before recording a human-owned action.
Related: Candidate Shortlist Template With Evidence · AI Interview Notetaking Software: What Buyers Should Compare · Candidate Interview Scorecard Template
Talent Summoner product facts read from our live candidate sourcing, candidate ranking and pricing pages, verified 19 August 2026. We re-verify this page quarterly — tell us if something changed.
