Compare candidates on the Decision Board

Step 1: Start from the role blueprint, not the leaderboard

Before you look at a single candidate, open the role blueprint the assessment was built on and re-read the weighted competencies. Those weights are the decision you already made about what this role needs — and every role-fit number on the board is computed against them. If you skip this step, you will unconsciously re-rank candidates against your private mental model of the role instead of the one your team agreed on.

Take an inside-sales requisition as the example. Suppose the blueprint weights discovery questioning at 30 percent, written follow-up at 25 percent, objection handling at 25 percent, and CRM hygiene at 20 percent. A candidate who dazzled in objection handling but wrote sloppy follow-ups is being scored the way your team decided a rep succeeds — not the way the loudest interviewer remembers the call.

If the weights now look wrong — say, the team realizes written follow-up matters more than it did when the blueprint was drafted — resist the urge to mentally re-weight on the fly. Note it, finish this cycle against the agreed weights, and revise the blueprint for the next requisition. Changing the rules mid-comparison is exactly what a defensible process avoids.

Step 2: Read role-fit as a starting point, not a verdict

The role-fit score is a weighted roll-up of rubric scores across the blueprint's competencies. It is the right place to start reading the board because it orders candidates by the standard you set in advance. It is the wrong place to stop, because a single number hides shape: two candidates at the same role-fit can have completely different strength profiles.

In a customer-support pool, one candidate might land at a strong overall fit on the back of excellent tone and empathy with middling troubleshooting; another reaches the same number with sharp diagnostics and serviceable but flat writing. Which profile you want depends on the team they are joining — a queue full of technical escalations needs the second, a churn-sensitive account list needs the first.

Use role-fit for what it is good at: setting your reading order and spotting the obvious ends of the pool — the candidate who cleared every bar and the one who cleared none. Everything between those ends is decided by the steps that follow — the evidence, the per-competency view, the flag reviews — not by the number that opened the board.

Step 3: Drill from any score into the underlying evidence

Every score on the board is a door, not a wall. Click through from a competency score and you land on the evidence that produced it: the task the candidate was given, the output they submitted, the rubric level it was scored at, and the evaluator's notes explaining why. This is the difference between an assessment platform and a leaderboard — the number is always attached to the work.

Make drilling down a habit at three specific moments: when a score surprises you, when two candidates are close, and before any final advance-or-reject call. For a support candidate with a low judgment-within-policy score, read the actual ticket reply. You may find a candidate who made a defensible exception the rubric penalized — or one who confidently promised a refund policy that does not exist. Same score, very different hiring conversations.

Evaluator notes deserve as much attention as the outputs themselves. A note like "clear structure, but never acknowledged the customer's frustration" tells you whether the gap is coachable in onboarding or fundamental to how the candidate works. A bare number cannot. Notes also expose scoring drift: if one evaluator's comments keep praising what another's keep penalizing, that is a calibration conversation to have before this shortlist closes, not after the offer goes out.

Step 4: Compare per competency, not just by overall rank

Overall rank answers "who did best on average?" Per-competency comparison answers the question you are actually hiring against: "who is strong where this role cannot afford weakness?" Switch the board to a competency-by-competency view and read down each column before you read across any row. Reading columns first also keeps one charismatic overall profile from quietly becoming the standard every other candidate gets measured against.

For inside sales, scan the discovery-questioning column across the whole pool first. If the role's biggest failure mode is reps who pitch before they diagnose, a candidate who ranks third overall but tops that column may be your strongest hire. For support, a candidate who is merely average everywhere may beat a spiky profile if the team needs someone safe to put in front of any customer on day one.

This is also where you catch pool-level problems. If every candidate scored weakly on the same competency, the issue may be the task or the sourcing channel, not the people — worth fixing before the next cycle rather than lowering the bar for this one. The board makes those pool-level patterns visible in a single view, which a stack of individual candidate reports never would.

• Read each competency column across all candidates before comparing any two candidates head to head.
• Identify the one or two competencies where the role cannot tolerate weakness, and treat those columns as gates.
• Note spiky profiles (one standout strength, one real gap) separately from flat profiles — they suit different teams.
• If the whole pool is weak on one competency, question the task and the sourcing before questioning the candidates.

Step 5: Treat risk flags as review items, never as rejections

The board surfaces integrity signals — heavy tab switching, pasted text in a written task, a large gap between drafting time and output length — as flags for a human to review. A flag is a prompt to look closer, not a verdict. SkillCort will not auto-reject anyone on a signal, and neither should you: proportional integrity means the response matches the evidence, and the evidence starts as ambiguous.

Open the flagged candidate's timeline and read the flag in context. A support candidate who switched tabs during a troubleshooting task may have been checking the product documentation you linked in the brief — behavior you would praise on the job. Pasted text in an email task might be a candidate reusing their own draft from a scratch file. Or the same signals might sit alongside an output that reads nothing like their other responses. The timeline usually tells you which story you are in.

Resolve every flag explicitly: reviewed and cleared, reviewed and discounted the affected task, or escalated to a follow-up conversation with the candidate. An unresolved flag left hanging on the board becomes silent bias — later readers treat it as a stain without ever checking what it was. The resolution takes one sentence to record and saves a whole meeting of speculation later.

• Review every flag against the task timeline before letting it influence any comparison.
• Distinguish job-realistic behavior (checking documentation) from genuine authenticity concerns.
• When in doubt, discuss the flagged task with the candidate — a short conversation resolves most ambiguity.
• Record the resolution on the flag itself so the next reader sees your reasoning, not just the alert.

Step 6: Record the rationale for who advances — and who does not

The last step is the one most teams skip: write down why. For each candidate who advances, record which evidence carried the decision — "top of pool on discovery questioning and written follow-up; CRM task acceptable; no unresolved flags." For each candidate who does not, record the specific gap against the blueprint, not a vibe: "strong tone, but troubleshooting task showed guessing over diagnosis on both scenarios."

This takes minutes while the evidence is open in front of you and is nearly impossible to reconstruct honestly three weeks later. It is also what makes the shortlist survive challenge. When a hiring manager asks why their favorite interviewee was cut, or a rejected candidate asks for feedback, you answer from the record instead of from memory.

The rationale you record here becomes the spine of the decision file — the audit-ready package covered in the companion playbook. Write it as if a skeptical colleague will read it, because eventually one will. Keep it factual and tied to evidence — name the competencies, tasks, and flag resolutions that carried the call, and avoid adjectives a later reader cannot verify against the record.