Good resume screening criteria are a short list of the things a role genuinely requires, split into must-haves and nice-to-haves, each one concrete enough that a CV can give evidence for it and each one actually predictive of doing the job. The criteria are the most important part of screening, and the part teams spend the least time on: most screening problems, slow, unfair, or inconsistent, trace back not to how the CVs were read but to criteria that were vague, too many, or quietly standing in for something other than ability. Decide the criteria well, before you open a single CV, and the screening itself becomes fast and defensible. Decide them badly and no tool or technique will save the result.
The discipline is to write down two or three real must-haves and a handful of nice-to-haves, in language a colleague could apply without asking what you meant, and to test each one against a single question: does this predict performance, or does it just predict a type of person?
Must-haves versus nice-to-haves
The first and most useful move is to sort every requirement into two buckets, because collapsing them is how screening goes wrong in both directions.
| Must-have | Nice-to-have | |
|---|---|---|
| Definition | The role cannot be done without it | Separates good from great |
| In screening | A pass-or-fail filter | Scored, never disqualifying |
| How many | Two or three, no more | A handful |
| The test | A brilliant candidate could not do the job without it | A brilliant candidate could, but it helps |
The failure modes are symmetrical. Treat a nice-to-have as a must-have and you filter out strong candidates on an optional box, the classic false negative. Treat a must-have as a nice-to-have and you let through people who genuinely cannot do the job. The single test that keeps the buckets honest is the one in the table: if a genuinely excellent candidate could still do the role on day one without a given requirement, it is a nice-to-have, and it must not be a hard filter.
What makes a criterion good
A requirement earns a place on the list only if it is both evidenced and predictive, and most weak criteria fail one of those two tests.
Evidenced means a CV or an interview can actually show it. “Has led a team of a stated size,” “holds the specific certification,” “has shipped work in the relevant domain” are evidenced; “strong communication skills” scored from a hunch is not, because it dresses an impression as a measurement. Predictive means it actually relates to doing the job well, which is where the research matters: decades of selection studies find that structured, job-relevant criteria predict performance far better than unstructured judgement, with structured evaluation reaching validities well above intuition. A criterion that is neither evidenced nor predictive is not a standard, it is a preference wearing one.
| Good criterion | Weak criterion | Why the difference |
|---|---|---|
| Three years in a B2B sales role | ”A proven track record” | One is checkable, the other is a vibe |
| Holds the required professional licence | ”A strong academic background” | One is a fact, the other a prestige proxy |
| Has shipped a product in the domain | ”A team player” | One predicts the work, the other predicts nothing |
The criteria that quietly cause bias
The most important thing to check is whether a criterion predicts the job or predicts a type of person, because the second kind is where screening becomes discriminatory without anyone intending it.
Some proxies are obvious: a photo, an age, a name have no bearing on capability, and the evidence that they change outcomes is strong, since identical CVs get different callbacks by name alone, which is why a fair first pass blinds them. Others hide inside plausible-looking requirements. A specific university as a must-have is prestige bias, not a performance predictor; an unbroken employment history filters out carers and the ill; a years-of-experience cliff rules out capable career-changers; “native-level English” where the job needs professional fluency screens on background rather than ability. The test for each is the same: strip the criterion to what the job actually needs. If the role needs the degree subject, require the subject, not the brand of university. If it needs professional English, require professional English, not “native.” Naming the underlying need, rather than a convenient signal for it, is what keeps criteria both fairer and more accurate. It also aligns with the GDPR’s data-minimisation principle: a criterion that names only what the job requires tends to need less personal data to assess than a proxy does.
Review your criteria against who they let through
Criteria are not set once and trusted forever; the useful check is to look at who your current criteria actually select and ask whether that matches who does the job well. If your shortlists keep skewing toward one background, one set of employers, or one type of career path, the criteria are probably encoding something beyond capability, even if each one looked reasonable in isolation. This is the quiet way a screening standard drifts into a filter for a type of person: not through one obviously biased rule, but through several plausible ones that happen to point the same way.
A simple, honest review is to take a few people who have done the role well and a few who struggled, and ask whether your criteria would actually have distinguished them. Often a requirement everyone assumed was essential, a particular degree, a specific prior employer, turns out not to separate the strong from the weak at all, while something not on the list, evidence of a specific skill, does. That is the signal to drop the non-predictive requirement and add the predictive one. Criteria improved this way get shorter and sharper over time, which is the opposite of the usual drift toward an ever-longer wish-list nobody can meet.
The point is that good criteria are a living judgement about what predicts performance in your context, checked against real outcomes, not a static list copied from an old job ad. Reviewing them is cheap and it is where most of the fairness and accuracy gains actually come from.
Turn the criteria into filters
Once the criteria are right, expressing them is straightforward, and it is worth writing them as plain-language rules rather than a query string. “Five years in B2B marketing, has owned a campaign budget, speaks professional Dutch” is readable, checkable, and reviewable by a hiring manager without decoding syntax, which is the argument for plain-language filters over a boolean string. Mark the must-haves as pass-or-fail and the nice-to-haves as soft, so a candidate who misses a nice-to-have is flagged for a second look rather than dropped. That soft treatment is the safeguard against the false negatives that good criteria still produce when applied as hard cliffs.
A worked example
A team is hiring a customer success manager and drafts its first list: a degree, five-plus years of experience, experience at a SaaS company, strong communication, a track record of retention, and familiarity with a named tool. Six requirements, several of them proxies, and applied as filters they would cut the pile to almost nobody while quietly favouring a particular background.
Reworked against the tests, the list shrinks and sharpens. The genuine must-haves turn out to be two: demonstrable experience owning customer renewals, and at least a couple of years in a customer-facing role in software. The degree comes off entirely, because the role does not need one and it was a prestige proxy. “Strong communication” becomes a thing to assess at interview, not a CV filter, because a CV cannot evidence it. The named tool and the SaaS-specifically background drop to nice-to-haves, scored but not disqualifying, so a strong candidate from an adjacent industry survives the first pass. The reworked list is shorter, fairer, and more predictive, and the screening that follows is faster because every criterion is now checkable.
Where a tool fits
A screening tool applies your criteria, it does not choose them, and that division is the honest one. Zen Job CV lets you write your must-haves and nice-to-haves as plain-language criteria and then ranks every candidate against them with the met-and-missed reasons attached, keeping near-misses visible so a soft-filter miss does not drop someone strong. What it cannot do is fix a bad list: feed it vague, proxy-laden, or too-many criteria and it will faithfully and consistently screen against exactly those. The judgement about what the role requires stays with you, which is why the criteria deserve more of your time than the tool. Done well, the tool makes good criteria fast and consistent across a large pile; it cannot make bad criteria good.
How to write resume screening criteria
Decide the criteria before you open any CV. Write two or three genuine must-haves and a handful of nice-to-haves, keep each one concrete enough that a CV can evidence it, and test every requirement against two questions: does it predict performance, and does it name the job’s real need rather than a proxy for a type of person? Sort ruthlessly into pass-or-fail must-haves and soft nice-to-haves, express them as plain-language rules, and apply the nice-to-haves softly so near-misses survive. The criteria are the part of screening worth the most care, because everything downstream, the speed, the fairness, and the defensibility, is decided by them.
Quick answers
What are good resume screening criteria? A short list of the things a role genuinely requires, split into two or three must-haves and a handful of nice-to-haves, each concrete enough that a CV can evidence it and each actually predictive of doing the job. Good criteria are checkable facts, not vibes like “a proven track record,” and they name the job’s real need rather than a proxy for a type of person.
How many screening criteria should I use? Two or three genuine must-haves and a handful of nice-to-haves. More than that, applied as filters, cuts the pile to almost nobody and usually means several requirements are really proxies or preferences rather than real needs. The discipline is to keep only what the role cannot be done without as a hard filter, and score everything else softly so it separates good from great without disqualifying anyone.
What is the difference between must-have and nice-to-have criteria? A must-have is something the role genuinely cannot be done without, applied as a pass-or-fail filter; a nice-to-have separates good from great and is scored but never disqualifying. The test is simple: if a brilliant candidate could still do the job on day one without a requirement, it is a nice-to-have. Treating nice-to-haves as must-haves is the most common way strong candidates get filtered out.
Can screening criteria be discriminatory? Yes, when a criterion predicts a type of person rather than the job. A specific university as a must-have, an unbroken employment history, a years-of-experience cliff, or “native-level” language where professional fluency is needed all screen on background rather than ability. The safeguard is to strip each criterion to what the job actually requires and to blind identity signals like name, photo and age in the first pass.
What is the biggest mistake when writing screening criteria? Treating a nice-to-have as a must-have, which filters out strong candidates on an optional box, and stuffing the list with proxies that predict a type of person rather than the job. Both cut the pile to almost nobody while quietly favouring one background. Keep only what the role cannot be done without as a hard filter, name the real need rather than a signal for it, and score everything else softly.