Any role with real visibility pulls in more resumes than anyone can read closely. A hundred, three hundred, sometimes far more. Screening is what happens at that moment, when someone has to turn the stack into a short list of people worth a conversation. Done on the fly, the person deciding ends up filtering by the order resumes arrived, by five o'clock fatigue, and by impressions that have little to do with competence.
You can do better, and it isn't complicated. What follows are the steps that keep screening fair while it still makes sense, the criteria worth looking at, and the point where reading resumes stops helping.
In one sentence
Screening well means deciding against a written yardstick set before you open the stack, weighing what a resume proves rather than what it implies, and treating the read as the first cut instead of the verdict. The sections below unpack each of those three.
The 5 steps of a screen that holds up
| Step | What to do | Why it matters |
|---|---|---|
| 1. Yardstick before the stack | Write the criteria and the weight of each before you see the first resume | The first strong resume stops becoming the benchmark for the rest |
| 2. Requirement or nice-to-have | Mark what is disqualifying and what is only a bonus | You stop cutting good people over a "nice-to-have" |
| 3. Rubric field by field | Score everyone on the same scale, one criterion at a time | Kills judgment by overall impression |
| 4. Blind first read | Hide name, photo, age and school on the first pass | Cuts gender, origin and prestige bias before it acts |
| 5. Short list to a real stage | Send whoever passed to a practical test, not to another read | A resume says what a person did, a test shows what they do |
1. Write the yardstick before you open the stack
Before you look at the first application, decide on three to five things that genuinely predict performance in that role and give each a weight. For a senior backend role, that might be experience with the kind of problem (not the framework of the month), evidence the person shipped something end to end, and signs of autonomy. Without that yardstick, the third resume of the morning gets judged against the first two rather than against the role. Arrival order becomes a criterion and nobody notices.
2. Separate requirement from nice-to-have
One column holds what actually disqualifies, like work authorization, a mandatory credential, or the language the team works in. The other holds what would be great but can be learned, like a specific stack or experience in a similar sector. The most common screening mistake is pushing a nice-to-have into the requirement column and narrowing the funnel too early. You quietly discard the person who would pick up the tool in a week.
3. Use a rubric, field by field
Scoring "relevant experience" from 0 to 2 leads to decisions that resemble each other far more than "I liked the profile" does. The difference shows up when someone asks, three weeks later, why one candidate advanced and another didn't. With a rubric you have the answer in writing. Without it you have only memory, which has already rewritten the story in favor of whoever got hired.
4. Hide name, photo and school on the first read
A name signals gender and often origin. A photo and an age open the door to a dozen mental shortcuts. The school name flips the prestige autopilot on. None of those fields measure competence, and all of them fire before you reach the part that does. On the first pass, cover them. The detail on each of these biases, with evidence, is in the 11 types of bias in hiring.
5. Stop reading at the short list
Screening resumes is the start of the funnel. A resume reports what the person says they did, in the wording they chose, and the next stage has to show what they can actually do. The sooner the short list meets a practical test, the less the decision depends on who writes a good resume, which is a different skill from writing good code.
What to weigh, and what to ignore
Not everything on a resume carries the same value. What earns weight is what the person delivered with context and result ("cut the build time from 12 minutes to 3"), the real hands-on time with that kind of problem, and anything you can verify from the outside, like a repository, a public project, or a portfolio.
What earns little is whatever is easy to inflate. The big-company name at the top, the "10 years in the market" with no hint of what those ten years held, the list of technologies anyone can paste in the footer. The biggest screening risk is ranking by that second list, because it is the most visible and the cheapest to write.
The mistakes that turn into bias
Filtering by exact keyword is the costliest one. A filter that demands "React" drops the senior frontend engineer who wrote "rebuilt the SPA from scratch" and never spelled the word out. You don't see who you lost, so you assume the filter worked.
Treating a gap in the resume as a negative is another. A parental leave, a year caring for someone ill, or a sabbatical becomes a penalty with no bearing on what the person delivers today.
Then there is "culture fit," which is sometimes just "social fit" in a better outfit. "They'll get along with us" tends to mean "they're like us." A culture fit that holds up is sharing the values the company put in writing. The rest is affinity dressed up as a criterion.
By hand, with an ATS, or with assessment
There are three ways to shrink the stack, and they don't solve the same problem.
Reading by hand works when there are twenty resumes and falls apart at three hundred. It is consistent only with a rubric, and expensive in senior time. An ATS with keyword filtering handles the volume, but it inherits the bias of the optimized resume and rewards whoever knows how to please the filter. It is there to organize applications, not to measure who is better. Assessment by test swaps reading what the person claims for measuring what they do, and it is the only one of the three that produces a signal the candidate cannot simply rewrite.
AI resume screening sits between the last two. It scales like the ATS but is still reading a resume, so it gets the structured parts right and the potential wrong.
How Nort handles it
At Nort, the stack arrives already filtered by competence, not by document. Everyone in the pool went through a single assessment of technical skill, languages and behavioral profile, on open frameworks like Big Five and the CEFR. Instead of reading resumes one by one, the company searches a ranking of already-assessed people and filters by what was measured. The bias of the well-written resume never enters the math, and the time spent reading the stack becomes time talking to people who have already shown their work.
Reading a resume isn't useless. It just answers a weaker question than the one you need. It shows what the person wrote about themselves, when the decision calls for what they can deliver.
Frequently asked questions
What is the difference between screening and sorting resumes?
In practice, the same step. "Screening" leans toward cutting whoever doesn't meet the requirement, and "sorting" leans toward ranking by fit. A decent process does both with the same rubric.
How many resumes can you screen well in a day?
With a rubric and a structured read, a few dozen before fatigue starts deciding for you. Past that, the hour and the order weigh more than the candidate, and it is worth trading part of the reading for an objective filter.
Is AI resume screening fair?
It depends on what the AI reads. If it scores keywords and formatting, it inherits the bias of whoever optimizes their resume. If it measures competence with a validated instrument, it tends to be fairer than a tired human read. The tool doesn't guarantee fairness on its own.
Do US laws limit automated screening?
Rules like NYC Local Law 144 require an annual bias audit and candidate notice for automated employment decision tools, and the EEOC still applies Title VII disparate-impact analysis to any screening tool. A screen that ranks candidates for a recruiter to decide is on safe ground. The risk lives with automation that rejects on its own, with nobody in the loop.
What should you hide on the first read?
Name, photo, age, address and the name of the school. Those are the fields that switch on gender, origin and academic-prestige bias before any read of competence.
TL;DR
- Write the yardstick and the weights before you open the stack, or arrival order decides for you
- Separate requirement from nice-to-have so you don't cut good people over a bonus
- Score with a rubric and hide name, photo and school on the first read
- Weigh what a resume proves, distrust what is easy to inflate
- A resume is the start of the funnel, so send the short list to a practical test, not to another read
- Nort filters by competence already measured across an assessed pool, not by document
Related resources
- How AI resume screening works and where it fails
- The 11 types of bias in hiring
- How to evaluate a tech candidate without interviewing
- What an ATS is
- What Smart Match is
Reviewed July 24, 2026. Send comments or corrections to [email protected].