How to Screen CVs Without Throwing Away Good Candidates

How to Screen CVs Without Throwing Away Good Candidates

Mercedes Gordo runs delivery and customer success at Nova and spent two years at LinkedIn in Dublin. Her view on screening: most of what gets filtered out in the first pass is bad writing, not bad candidates.

Updated 6 min read

A hundred CVs land for one role. You’ve got an afternoon.

Whatever happens next determines the quality of everything downstream, and most teams do it with a method they’ve never written down: open, skim, feel, pile. By CV forty the bar has quietly moved, and by CV eighty you’re comparing people against the last person you read rather than against the job.

Here’s the structure that makes the afternoon repeatable, and the one rule that changes who survives it.

Decide the criteria before you open the first CV

Three buckets, and only one of them can reject anybody.

  • Knockouts. Hard barriers. Work authorization, the minimum language level, the salary band, the location if the job genuinely requires it. Failing one ends the evaluation, and there’s no point scoring the rest.
  • Must-haves. The core requirements, and these carry the fit. They weigh the same as each other unless you have a reason to say otherwise.
  • Nice-to-haves. They add. They break ties. They never eliminate anybody.

Writing these down before you start is what stops the bar drifting. Criteria invented while reading tend to form around whoever you read first, and then the person you opened at nine o’clock sets the standard for the whole batch without anyone deciding that.

It’s also the moment you find out whether the brief is usable. If you can’t name the knockouts, the hiring manager hasn’t told you what the job actually requires, and that’s a five-minute conversation worth having before you spend the afternoon.

The rule that changes the outcome: absence is amber

Use a traffic light rather than a yes or no.

  • 🟢 Advance. The evidence is there.
  • 🟡 Maybe. Something is missing, unclear, or needs a conversation.
  • 🔴 Reject. A knockout failed, or there’s clear evidence the person doesn’t meet the bar.

And then the rule that matters most: what the CV doesn’t prove goes amber, never red.

A red means you saw something below the bar. An absence is not that. If a CV doesn’t mention whether someone has led a team, that’s a question for a ten-minute call, not a rejection.

This sounds small and it isn’t, because the most common reason a strong candidate gets filtered out in a first pass is that they wrote a vague CV. Writing a CV is a distinct skill from doing the job, and the correlation between them is weaker than the hiring process assumes. The people who write the best CVs are often the people who’ve been job-hunting most recently, which is not the signal you’re looking for.

So the amber pile is not the maybe-pile-of-shame. It’s the list of people whose evidence you haven’t seen yet, with the question you’d ask written next to each one.

Pedigree, and the distinction worth getting right

Two things get merged here that are genuinely different, and merging them leads people either into discrimination or into a kind of performative blindness that doesn’t help candidates either.

Personal characteristics never enter the scoring. Age, date of birth, gender, photo, nationality, marital or family status, religion, health, address. They don’t predict performance, and filtering on them is, depending on where you are, illegal. If a CV contains them, they stay out.

The brand of a company or a university is a different kind of thing. It’s a market signal, it’s often job-related, and employers and clients visibly value it. Pretending otherwise doesn’t make screening fairer, it just makes the reasoning unstated. The workable rules:

  • ✅ It adds, as a nice-to-have. Coming from a strong firm counts in someone’s favour.
  • 🚫 It is never a knockout. It doesn’t reject someone who demonstrates the same competency another way.
  • ⚠️ It doesn’t cover for missing evidence. A recognisable logo with no achievements underneath it doesn’t win on the badge.
  • 📌 Exception: if the role explicitly requires it, it’s a must-have, because the client set that criterion rather than you assuming it.

And the same reframing applies to the brief itself. “Maximum 35 years old” is seniority and adaptability. “Native speaker” is usually a C1, which is a language level you can actually test. “Good presentation” is how someone handles clients, which a CV doesn’t show either way. Screen on the legitimate requirement underneath, and flag the original to whoever wrote it.

Gaps don’t lower the score

A break, a sector change, a career switch: these get flagged as something to explore, not as a shortfall.

There are a thousand legitimate reasons for a gap and most of them are nobody’s business. Penalising them systematically filters out precisely the non-linear careers that tend to bring the most distinctive experience, which is the same structural problem keyword search has and for the same reason: the filter rewards the conventional shape rather than the capability.

Unless the role has an explicit criterion that a gap fails, it’s a conversation, not a deduction.

What the output should look like

One row per candidate, and four columns:

CandidateLightFitWhy, with the evidence
…🟢 / 🟡 / 🔴%The line from the CV that supports it, and for amber, the question to ask

The why column is the part that earns the method its keep. A screen that produces a pile and no reasons can’t be checked by anyone, can’t be handed over, and can’t be defended if a candidate asks. A screen with the evidence written next to each call can be reviewed in five minutes by the hiring manager, which is usually when you discover you and they mean different things by “senior”.

And the fit percentage ranks the batch. It doesn’t decide it.

The decision stays with you

The tool classifies, scores and shows its reasoning. The decision about a person is made by a person.

That isn’t only an ethical position, although under the EU AI Act the employer using the tool carries responsibility for the outcome regardless of who built it. It’s also better screening. The cases where a tool is least certain are reliably the ones most worth a human minute, and a process that auto-rejects the uncertain ones is optimising away exactly the candidates worth looking at.

For any team handling more than fifty applications a role, writing the three buckets before you start and treating absence as amber will change who reaches your shortlist more than any tool you buy. If you screen in smaller batches, the single highest-return change is the reasons column: once every call has evidence next to it, the rest of the discipline follows.

If you’d rather not do the first pass by hand, this method is one of eleven free AI skills our customer success team packaged up. You give it the job description and the CVs, it confirms the criteria with you before screening anything, and it returns the table with the evidence and the questions to ask. It doesn’t reject anybody.

The half before this one matters too, though. A screening process applied to the wrong hundred people produces a well-documented disappointment, which is why the shortlist is ranked by merit rather than by who wrote the right words.

Frequently asked questions

How do you screen CVs fairly?

Decide the criteria before you open the first CV, sort them into knockouts, must-haves and nice-to-haves, and apply the same ones to everybody in the batch. Score only on things related to the job: skills, requirements, experience and results. Anything in a CV about age, gender, nationality, marital status, health or a photo stays out of the scoring entirely. And treat missing evidence as a question rather than as a failure.

What should you do when a CV doesn't show whether someone meets a requirement?

Mark it amber and write down the question you'd ask. A red is for evidence that the person doesn't meet the bar; an absence is not that. This single rule changes screening outcomes more than anything else, because the most common reason a strong candidate gets filtered out in the first pass is that they wrote a vague CV, not that they can't do the job.

Is it biased to consider where someone worked or studied?

It's worth separating two things that often get merged. Personal characteristics like age, origin, gender or a photo never predict performance and never enter the scoring. The brand of a company or a university is different: it's a market signal, employers and clients genuinely value it, and it can be job-related. The rules are that it adds as a nice-to-have, it is never a knockout, and it never covers for a profile with no demonstrated achievements. The exception is when the role explicitly requires it, in which case it's a must-have because the client set it, not because you assumed it.

Do career gaps count against a candidate?

Not by default. A break, a sector change or a career switch gets flagged as something to explore in conversation, not as a shortfall in the score. Gaps have a thousand legitimate explanations, most of which are nobody's business, and penalising them systematically filters out exactly the non-linear careers that often bring the most distinctive experience.

Can an AI tool reject candidates for you?

It shouldn't, and under the EU AI Act the employer using the tool carries the responsibility for the outcome regardless. The workable arrangement is that the tool classifies, scores and shows its reasoning with the evidence, and a person makes the decision about a person. That's also simply better screening: the cases where a tool is least confident are usually the ones most worth a human minute.

Sources and method

The method described here is the one behind Nova's CV Screening skill, built and published by the customer success team. It's free to download as part of the Inbound pack.

Mercedes Gordo
Mercedes Gordo

Head of Delivery and Customer Success

Mercedes Gordo is Head of Delivery and Customer Success at Nova. Mercedes spent two years at LinkedIn in Dublin as a relationship manager and account director, and previously ran Customer Alliance's Spanish operation.

Connect on LinkedIn →
Sourcing and outreach The Interview Scorecard: How to Score Candidates on the Same Yardstick Sourcing and outreach How to Write a Job Description That Doesn't Shrink Your Pool