Back to Resources
Hiring 7 min read

How to screen resumes without losing good people

Screening is where most hiring processes lose their consistency. A defined pass, a fixed set of criteria and a written reason for every no will do more for quality than any tool.

Knowing how to screen resumes is mostly about deciding what you are looking for before you start looking. Screening goes wrong when it becomes reactive: you open the pile, form impressions, and let the impressions become criteria. Two hundred applications later the bar has drifted, nobody can reconstruct why the first forty were rejected, and the candidates who got the most attention are the ones who applied when you were fresh.

The fix is unglamorous and it works. Define the criteria, run one consistent pass, and record a reason for every decision.

Write the criteria before you open the pile

Take the must-haves from the job description and turn them into four or five screening criteria that can be answered from a resume. Not assessed, not inferred. Answered.

"Has owned a monthly close cycle" is a screening criterion. "Strong attention to detail" is not, because you cannot establish it from a document, and trying to will just launder your instincts into something that looks like evidence.

Write them down before you read anything. Five minutes of work that changes the whole exercise, because now every candidate is being measured against the same fixed thing rather than against the last resume you read.

Decide your bar at the same time. Must-haves are pass or fail. Everything else is context that informs a borderline call, not a reason to reject.

The first pass: fast, and the same every time

The first pass is not an evaluation. It is a sort into yes, no and maybe against the criteria you just wrote. Ninety seconds a resume is realistic, and going slower does not make it more accurate.

Read in a fixed order so you are looking for the same things in the same sequence:

Read the experience before the education, and read what they owned before you read where they did it. Both habits pull you away from the proxies that feel like signal and mostly are not.

Write one line of reason for every no. This is the discipline that matters most. It takes seconds, it makes you name the actual criterion, and it stops the drift. It is also the only thing that lets you answer a question about a rejection six months later.

  • Most recent role and what they actually owned in it
  • Evidence against each must-have criterion
  • Progression and the shape of the career
  • Anything genuinely unclear that a screening call would resolve

The second pass is a comparison

The first pass produces a yes pile and a maybe pile. The second pass compares them with each other rather than against the bar.

Here you are looking at the shape of a shortlist. Do these candidates cover the range of what the role needs. Are you holding a group that all look like the same person. Is there a maybe whose evidence is thinner but whose trajectory is steeper.

This is the right moment to revisit the maybes, because now you have calibration. The first three resumes you read were judged without context. The thirtieth was judged against twenty-nine others. Re-reading the early maybes at the end corrects for that, and it regularly surfaces someone good.

Where automation helps and where it does not

Software is useful for narrowing and ordering. It is not useful for deciding.

Reasonable: filtering on a genuine hard requirement such as a licence or a right to work, surfacing applications that match criteria you defined, putting the pile into a sensible reading order, flagging duplicates.

Not reasonable: keyword matching as a proxy for capability, scoring candidates on inferred traits, or any automatic rejection. Keyword filters reject people who did the job and describe it differently, and they reward people who know to mirror the advert's wording back. That is a test of resume literacy, not of ability.

The rule: automation orders the pile, a person decides. HRM8 does not auto-reject candidates, and any disqualification routes to the employer to sign off. When something is genuinely hard to establish from a resume, that is an argument for assessment later in the process, not for a cleverer filter earlier.

Screen with two people when the role matters

For a senior or business-critical hire, have two people screen the same pile independently against the same criteria, then compare.

The disagreements are the valuable part. They show you where the criteria were ambiguous, and they catch the candidate that one reader would have dismissed on a pattern the other reads differently. It costs an hour. On a role you will live with for three years, that is nothing.

Reply to everyone

Not a nicety, a practical matter. Every unanswered applicant is someone who tells other people about it, and some of them are people you will want to hire later, or customers.

A short, clear rejection sent promptly costs you almost nothing, and an applicant tracking system will send it for you. Avoid promising a specific timeframe you may not meet. Say what happens next and then do it.

Common questions

How long should screening a resume take?

About ninety seconds for the first pass. The goal is a consistent sort against fixed criteria, not a full evaluation. Depth belongs at the screening call and the interview.

Should you screen blind?

Removing names, photographs and graduation dates reduces some bias and is cheap to do. It is worth doing, but it is not sufficient on its own, because bias also enters through employer prestige, career gaps and phrasing. Fixed criteria and written reasons do more.

What is the biggest resume screening mistake?

Changing the bar as you read. Everything else follows from it. Fixed written criteria prevent it.

Can AI screen resumes for you?

It can order and narrow the pile against criteria you set. It should not reject anyone. Automated employment decisions carry specific obligations in a growing number of jurisdictions, and a score you cannot explain is a decision you cannot defend.

The takeaway

Write the criteria before you open the pile, run a ninety-second first pass in a fixed order, record one line of reason for every no, and re-read the early maybes at the end when you finally have calibration.