How to Write a Defensible Hiring Decision

The hiring decision is made. You want to move forward with one candidate and decline the others. Now someone asks: "Can you walk me through why?"

If your answer relies on impressions — "she seemed sharper," "he had more energy," "the panel just preferred her" — you have a problem. Not only is impression-based reasoning indefensible if challenged; it is also a mechanism for perpetuating the biases every organisation claims to be working against.

A defensible hiring decision is not a longer document. It is a more honest one: every score is traceable to a specific piece of evidence, every inference is labelled as such, and the reasoning survives scrutiny by someone who was not in the room.

This guide shows you exactly how to write one, using Verdict's six-dimension model as the evaluation frame.

What Makes a Decision Defensible

Three conditions must hold simultaneously:

1. The criteria were fixed before the candidates arrived. You cannot decide post-hoc that "cultural fit" or "executive presence" are the deciding factors — especially when those terms disproportionately penalise candidates from underrepresented groups. Criteria defined after reviewing CVs are contaminated by the candidates you happened to see.

2. Each criterion is linked to evidence, not impression. "Strong communicator" is an impression. "In the final-round interview, described restructuring a team of twelve engineers across three time zones with no increase in headcount — and articulated the sequencing in a way every panel member followed without prompting" is evidence. Evidence is quotable. Impressions are not.

3. The reasoning is separable from the person. A well-written decision record reads the same whether the candidate is named Ana or Andrew. If replacing the name changes the emotional tone of the narrative, the reasoning is contaminated.

The Anatomy of a Decision Record

A decision record has four parts:

• Role context — the specific level and scope requirements the role demands. One paragraph, written before any CV is opened.
• Dimension scores — one score per dimension, each with a verbatim evidence anchor and a one-sentence rationale for why that evidence maps to that score.
• Comparative summary — if this is a head-to-head or shortlist decision, a brief statement of where the finalists diverged on the evidence, without editorialising.
• Protected-attribute check — a sign-off confirming the decision does not rest on any factor that correlates with a protected characteristic.

None of these sections need to be long. A decision record for a mid-level role routinely fits in two pages. The discipline is in precision, not volume.

Why Decision Records Fail Scrutiny

The most common failure mode is not malice — it is language that sounds specific but is not. "Strong leadership skills," "excellent communication," and "cultural fit" appear in decision records constantly. None of them would survive a five-minute challenge: What did the candidate say or do that demonstrated strong leadership? In which context? With what outcome? When the answer is "it was just a feeling the panel had," you do not have a decision record. You have a preference dressed as an evaluation.

A secondary failure is inconsistency: applying a criterion strictly to one candidate and loosely to another without recording the reason for the asymmetry. If one candidate's employment gap is noted as a concern and another's is not mentioned at all, the disparity becomes a liability. The discipline of a decision record is precisely this consistency — the same questions asked with the same rigour for every candidate.

Evidence vs. Inference — The Key Distinction

This is where most decision records fail, because the line between evidence and inference is less obvious than it appears.

Evidence is something a candidate produced or said that can be verified. A CV line. A portfolio item. A direct quote from an interview answer. Evidence exists independently of the evaluator's interpretation.

Inference is a conclusion you draw from evidence. "This pattern of outcomes suggests she builds trust with cross-functional peers" is an inference — a legitimate one, but an inference nonetheless.

Both have a place in a decision record. The mistake is presenting inferences as if they were evidence, or omitting the evidence that would allow a reader to verify the inference.

Labelling rule: In your decision record, write the evidence first, then the inference, separated explicitly. "CV states: 'negotiated a $4.2M renewal with Telefónica's procurement team, no commercial concessions.' Inference: this signals comfort operating at enterprise commercial scale." Anyone reading that record can agree or disagree with the inference while seeing exactly what it was based on.

A Worked Decision Record — Six Dimensions

The following example uses a real-sounding candidate evaluated for a Head of Revenue Operations role at a Series B company. Each dimension follows the Verdict six-dimension model. Every evidence anchor is the kind of line you would actually encounter in a senior candidate's CV.

Role context: Head of Revenue Operations, 60-person B2B SaaS, $8M ARR, target $25M in 18 months. Requires ownership of GTM toolstack, pipeline forecasting methodology, and cross-functional alignment between Sales, Marketing, and Customer Success. No RevOps team in place — this is a builder role.

1. Capability — Score: 4 / 5

Evidence: "Designed and implemented HubSpot→Salesforce migration for 80-seat sales org, including custom attribution model; zero data loss, 6-week delivery."

Rationale: Demonstrates the technical execution depth the role requires — not just tool familiarity but end-to-end ownership of a high-stakes migration at relevant team size. Scored 4 rather than 5 because the evidence does not show prior ownership of the full forecasting-to-compensation stack.

2. Track Record — Score: 4 / 5

Evidence: "Reduced average sales cycle from 67 to 41 days over three quarters by introducing stage-exit criteria and eliminating manual approval steps. Pipeline accuracy improved from ±35% to ±12% by Q3."

Rationale: Quantified, attributed, plausible for a RevOps function at a company of comparable maturity. Owned the outcome (not "contributed to"). The improvement in forecast accuracy is particularly relevant given the stage-gating requirement of this role. Scored 4 because the dollar-return impact on ARR is not stated directly.

3. Trajectory — Score: 3 / 5

Evidence: "RevOps Analyst (2018) → Senior RevOps Analyst (2020) → RevOps Manager (2022) → Senior RevOps Manager (2024)."

Rationale: Consistent progression, two-year cadence, but each step is within the same function and the same company. No evidence of scope expansion beyond the RevOps lane — cross-functional leadership, P&L exposure, or a lateral move that expanded the frame. For a builder role requiring influence over Sales and CS leaders, this trajectory suggests competence in depth rather than the broadening that predicts success at Head level. Scored 3 — neither a red flag nor a differentiating signal.

4. Influence — Score: 4 / 5

Evidence: "Presented pipeline methodology to Board of Directors in two consecutive quarters at CFO's request; methodology subsequently adopted as the company's official forecast model."

Rationale: Evidence of upward influence at the highest organisational level, under conditions where the audience had no obligation to accept the framing. Adopted as the official model — not merely acknowledged — indicates the influence converted into a durable institutional change. Scored 4. Would score 5 if there were evidence of having shaped a peer leader's decision-making process, not only executive-level presentations.

5. Domain Edge — Score: 3 / 5

Evidence: "Certified Salesforce Administrator and HubSpot Revenue Operations Certified. Guest contributor to RevOps Co-op newsletter, two articles on pipeline hygiene methodology."

Rationale: Solid practitioner credentials and some community presence. The newsletter contributions suggest above-average investment in the field, but the pieces are practitioner-level rather than definitional work (new frameworks, empirical research, widely-cited methodology). Domain edge is a 3 — competent, current, but not a recognised authority in the field. For this role, that is sufficient; this dimension is not the differentiator.

6. Risk Surface — Score: 2 / 5 (lower score = lower risk)

Evidence: One employer gap: 8 months between Company A (departed Jan 2022) and Company B (joined Sep 2022). CV does not explain the gap. No pattern of short tenures — average tenure is 3.8 years.

Rationale: Single unexplained gap warrants a reference conversation, not a disqualifier. Average tenure is long. No scope inconsistencies across roles. Risk surface is low; the gap should be confirmed in reference rather than assumed. Scored 2 (low risk) pending reference check on the 2022 period.

Comparative summary (if shortlist): Candidate A's aggregate profile — high Track Record, high Influence, moderate Trajectory — maps well to this role's immediate demands. To illustrate how the reasoning shifts: if a second candidate scored higher on Trajectory (say, 4/5 — evidence of cross-functional leadership outside RevOps) but lower on Track Record (3/5 — outcomes less directly attributed) and Influence (3/5 — no board-level evidence), the comparative note would read: "Candidate B brings broader scope progression, but the role's primary near-term demand is credibility with existing Sales and CS leaders rather than a new career chapter. Candidate A's evidence on direct-attribution outcomes and executive influence is the stronger fit for the first 90 days; the cross-functional breadth gap can be closed in-seat." That framing is still a judgment call — the evidence does not make the decision for you — but the reasoning is now auditable by someone who was not in the room.

The Protected-Attribute Check

Before the decision record is finalised, every evaluator should answer four questions:

1. Does the rationale for any score reference, even indirectly, a factor that could correlate with a protected characteristic? (Communication "style," "polish," "cultural fit," accent, name, educational institution prestige in markets where that correlates with socioeconomic background.)

2. Were the criteria applied consistently? If you scored one candidate's career gap as "low risk" and another's identically structured gap as "high risk," what was the actual difference — and can you state it in terms of observable evidence?

3. Would the rationale read the same if the name were changed? Run the test: redact the name and reread the record cold. If the tone shifts, find the inference that is doing the work and replace it with evidence.

4. Can you explain the decision to the candidate who was not selected without referencing anything that would be unlawful to consider? In most jurisdictions this is a useful heuristic: if you could not say it in a feedback conversation, it should not be in the record.

This check is not a legal review. It is a calibration step that catches the most common failures before they are baked into a permanent record.

FAQ

How long should a decision record be? Long enough to anchor every score to evidence, short enough that someone who was not in the process can read it in ten minutes. For a mid-level role, this is typically 400–700 words of substantive content. Senior and executive roles may run longer if the evidence base is richer.

Do I need a decision record for every hire? The minimum useful case is any role where more than one person evaluated the candidate, or where the decision is likely to be challenged. In practice, documenting even early-stage rejections — at the CV screen — is valuable: a one-sentence evidence anchor ("CV lists responsibilities only; no outcomes with scope or timeline") protects against later claims that the screening was arbitrary.

What if the best candidate has a thinner CV but excelled in the structured interview? Document the interview evidence with the same rigour: the exact question asked, the candidate's response (paraphrased or quoted), and what that response evidenced. Structured interview data is valid evidence. Unstructured impressions from an informal "culture chat" are not.

Can Verdict generate a draft decision record for me? Verdict extracts verbatim CV quotes and maps them to the six dimensions with score rationales. That output is the evidence layer of a decision record — the protected-attribute check and comparative summary still require human judgment. See the free demo or evaluate a candidate now.

What is the difference between a decision record and an offer letter? Completely different purposes. The offer letter is a contractual document for the candidate you are hiring. The decision record is an internal evaluation document for every candidate evaluated — including those declined. Keep them separate.