How to Shortlist Candidates: Criteria, Matrix and Cutoff

Short answer: To shortlist candidates, write down your essential and desirable criteria before applications open, score every applicant against those criteria on a shortlisting matrix, and interview the top 10 to 12. Essential criteria are pass or fail, desirable criteria are weighted points. The discipline that makes shortlisting defensible is applying the same criteria to every applicant, including the four hundredth one, which is exactly where manual shortlisting breaks down.

Last updated July 2026.

What is shortlisting in recruitment?

Shortlisting is the step between "applications closed" and "interviews scheduled." You take the full applicant pile, measure each person against the criteria the role actually requires, and produce a small ranked group worth a hiring panel's time. It sits after screening out ineligible applicants and before any interview, assessment, or reference stage.

The word gets used loosely, so it helps to be precise about what shortlisting is not. It is not filtering by keyword, which finds people who used the right vocabulary rather than people who can do the job. It is not reading until you find enough decent candidates and stopping, which is the most common real-world practice and the least defensible. Shortlisting is a comparison exercise: every applicant measured on the same yardstick, then ranked.

How many candidates should you shortlist?

Ten to twelve per role is the standard rule of thumb, and it holds up for a practical reason rather than a theoretical one. A hiring panel running structured 45-minute interviews can handle about that many before quality erodes and the calendar stretches past the point where strong candidates have accepted something else. For high-volume frontline roles where interviews are shorter, 15 to 20 is workable. For a senior or specialist hire, 5 to 8 is often plenty.

The number matters less than where you draw the line. Rank the whole pile first, then look at where the scores drop off. If candidates 1 through 7 cluster tightly and candidate 8 falls well behind, your shortlist is 7, not 12. Padding a shortlist to hit a target number wastes panel hours on people you already know are not competitive.

Essential vs desirable criteria: the split that does the work

Every good shortlisting process starts by splitting requirements in two.

The single most valuable hour in a hiring process is the one you spend cutting your essential list down. Most job descriptions carry five to nine "requirements," of which two or three genuinely gate the work. Everything else belongs in the desirable column with a weight. Over-listing essentials shrinks your pool, disproportionately screens out career-changers and non-traditional backgrounds, and creates disparate-impact exposure you did not need to take on.

The shortlisting matrix, step by step

A shortlisting matrix is a grid: applicants down the side, criteria across the top, a score in each cell. It is the oldest tool in recruiting and still the best one, because it forces you to compare candidates on the same dimensions instead of against whichever resume you happened to read first.

1. Write the criteria before applications open. Criteria invented after you have seen the applicants tend to describe the applicants you already liked. Lock them to the job description.
2. Assign weights. A five-year track record in the core function might be worth 5; familiarity with a specific tool the team uses might be worth 2. Weights should reflect how much each factor predicts performance, not how easy it is to spot on a resume.
3. Run the knockouts first. Check essential criteria before you score anything. This is fast, it is binary, and it should be documented, because "did not hold the required license" is a defensible reason and "did not seem like a fit" is not.
4. Score the desirables on evidence. Each score should trace to something on the resume or in the screening answers. If you cannot point to the evidence, the score is an impression.
5. Rank and set the cutoff. Sort by total weighted score, look for the natural drop-off, and take everyone above it.
6. Keep the sheet. The matrix is your audit trail. If a rejected candidate or a regulator asks how the shortlist was built, the completed grid is the answer.

Why manual shortlisting quietly fails at volume

Nothing above is controversial. Recruiters know the method; most were trained on it. The failure is arithmetic. A single US job posting routinely draws 250 to 400 applications. At a genuinely careful three minutes per resume, scoring 300 applicants against six weighted criteria is roughly 15 hours of concentrated work for one role, and a recruiter carrying eight open reqs does not have 120 hours.

So the process degrades in predictable ways. Reviewers open resumes in submission order and stop once they have twelve that look fine, which means the shortlist is drawn from whoever applied fastest rather than whoever fits best. Scoring gets coarser as fatigue sets in, so the ninetieth resume of the day is judged differently from the ninth. The matrix gets filled in retroactively, if at all. The method is sound; the capacity to execute it is not there.

What automated shortlisting actually changes

Automating shortlisting does not mean handing the decision to software. It means running the matrix you already designed across the entire pile instead of across the part of the pile someone had time for.

Resume shortlisting software reads each application as it arrives, applies the essential criteria as knockouts, scores the desirable criteria against the weights you set, and ranks the full pile with the evidence behind each placement. The output is the same shortlist the matrix method would have produced with unlimited hours, plus a record of how it was built. Prescreener works this way deliberately: it ranks and flags candidates for a recruiter, never rejects anyone on its own, and shows a plain-English reason on every placement. The screening is bias-audited to support EEOC standards and NYC Local Law 144, with candidate consent and clear AI disclosure.

The practical difference shows up in who makes your shortlist. When every applicant is scored on the same criteria, the strong candidate who applied on day nine is compared as carefully as the one who applied in the first hour. That is usually where the improvement in hire quality comes from, not from the software being cleverer than a recruiter.

Can shortlisting be biased?

Yes, and both the manual and automated versions can be. Manual shortlisting carries well-documented bias in how names, schools, employment gaps, and address details influence a reviewer. Automated shortlisting carries the risk that a scoring model rewards proxies for protected characteristics rather than the job requirement itself.

The controls are the same in both cases. Score only on job-related criteria fixed before applications opened. Keep the evidence behind every score. Audit outcomes by group at the shortlist stage, not just at hire. Disclose automated tools to candidates and get consent. Keep a human making the advance and reject decisions. If you are working through the compliance side, our guides on reducing bias in AI resume screening and the current AI hiring laws by state cover what US employers are actually required to do in 2026.

Shortlisting mistakes worth avoiding

• Shortlisting on tenure alone. Years in a seat measures duration, not capability. Weight demonstrated scope and outcomes above raw years.
• Treating resume polish as a proxy for competence. Formatting quality correlates with access to career coaching more than with job performance, outside roles where writing is the job.
• Letting the first strong candidate anchor the scale. If you score candidate 1 as a 5 and quietly grade everyone against them, you are running a tournament, not a matrix.
• Rejecting on gaps without asking. Employment gaps have many causes and are a poor predictor on their own. Make it a screening question rather than a silent knockout.
• Discarding the runners-up. Candidates 13 through 20 are your pipeline for the next opening and for the moment your first choice declines. Keep the ranked list.

When you should not shortlist at all

Occasionally the pile is the wrong problem. If a role has pulled 14 applications in three weeks and six are eligible, shortlisting is not your bottleneck; sourcing is, and you need to go find candidates rather than sort the ones who found you. If the role is short-term, project-shaped, or genuinely specialist, hiring may not be the right instrument either. For scoped work with a clear deliverable, teams increasingly bring in a vetted specialist for the project instead of running a full req that will take eleven weeks to fill.

A workable shortlisting process for a busy team

If you want a version that survives a real week, it looks like this. Spend 30 minutes before the posting goes live defining two or three true essentials and four or five weighted desirables. Add two or three screening questions that capture what a resume cannot: eligibility specifics, shift availability, license status. Let screening run on every application as it arrives rather than in a batch at the close date. Review the ranked top 20, read those resumes properly, and interview the 10 to 12 above the score drop-off. Keep the ranked list for the next opening.

That process takes a couple of hours of recruiter attention per role instead of fifteen, and it produces a shortlist you can put in front of a hiring manager, a general counsel, or an auditor without flinching. If you want to see it running against your own criteria, the candidate screening software and applicant scoring system pages walk through how the knockouts and weights are set up, and the resume screening checklist covers what to look for once the shortlist is in front of you.